
 ChangeLog for LAME
-==================-

 - All dates/times are in GMT.
 - Generated from svn log entries (svn2cl)

-- 

2026-11-06 07:53  aleidinger

	* libmp3lame/version.h:
	  Enter beta1 phase

2026-11-06 07:47  aleidinger

	* doc/maintainer-check-dist.md, maintainer/check-dist.ps1,
	  maintainer/check-dist.sh:
	  Expect the beta name in the distribution check
	  
	  A beta carries its number in the package version, as in 4.1beta1.
	  The check
	  takes that number from LAME_PATCH_VERSION, so the name and the
	  version that
	  the program prints have to agree.

2026-11-05 11:51  aleidinger

	* testcase.mp3:
	  Update testcase.mp3

2026-11-05 11:34  aleidinger

	* ChangeLog:
	  Update the ChangeLog

2026-11-05 11:31  aleidinger

	* configure, configure.ac:
	  Switch to beta1

2026-11-05 11:12  aleidinger

	* maintainer/check-dist.ps1, maintainer/check-dist.sh:
	  Count the patch level of a release in the distribution check

2026-11-05 11:12  aleidinger

	* doc/maintainer-check-dist.md, maintainer/check-dist.ps1:
	  Pass the library folders and platforms from check-dist.ps1 to the
	  matrix
	  
	  The tarball carries none of the optional libraries, so the
	  Windows check
	  skipped the decoder, libsndfile and DirectShow cells whatever was
	  installed.

2026-11-05 11:11  aleidinger

	* doc/maintainer-build-matrix.md, maintainer/gen-build-matrix.ps1:
	  Honour the library folders in every Windows matrix cell
	  
	  The folders given to the matrix reached only the nmake cells; the
	  MSBuild and
	  client cells looked under vc_solution regardless. The mpg123 and
	  libsndfile
	  options now name a folder with one subfolder per platform, as the
	  setup script
	  lays them out, and each cell gets the path for its platform.

2026-11-05 11:11  aleidinger

	* maintainer/check-optdoc.py:
	  Read the long options from the parser as it is written now
	  
	  The check still looked for the macros that the parser lost when
	  it was split
	  into one function per group, so it found no option at all and
	  failed.

2026-11-05 11:11  aleidinger

	* doc/man/lame.1:
	  Document --athaa-sensitivity in the manual page

2026-11-05 11:11  aleidinger

	* doc/maintainer-gen-usage.md:
	  Name usage.html as the source of the examples in USAGE

2026-11-05 11:10  aleidinger

	* USAGE, doc/html/usage.html, maintainer/gen-usage.py:
	  Number the notes in the generated USAGE
	  
	  The generator printed the notes as bullets, so nothing matched
	  the [1] and
	  [2] in the text, and a hand edit had added them a second time. It
	  now numbers
	  them under a Notes heading, separates table columns and keeps the
	  text in
	  front of a table. USAGE is generated again, with the option
	  descriptions that
	  changed since the hand edit.

2026-11-01 20:48  aleidinger

	* test/unit/test_gain_analysis.c:
	  Compare title gains by histogram step in the ReplayGain test
	  
	  The test compared the title gain with its reference for exact
	  equality. On
	  32-bit x86 without SSE, both lie on the same histogram step, but
	  the
	  reference keeps more precision than a float, so the test failed.
	  A gain now
	  fails when it is half a step or more from the reference.

2026-11-01 20:47  aleidinger

	* test/unit/test_quantize_lines_vector.c:
	  Give the SSE control register accesses of a vector test the SSE
	  attribute
	  
	  The quantization vector test reads and sets MXCSR itself. On
	  32-bit x86, where
	  the compiler does not enable SSE by default, it did not compile,
	  and make
	  check stopped before any test ran. The two accesses now sit in
	  functions
	  marked like the library's own SSE code.

2026-11-01 20:47  aleidinger

	* doc/maintainer-build-matrix.md, maintainer/gen-build-matrix.sh:
	  Build the internal options in a matrix cell of their own
	  
	  The debug cell enabled the internal options, which configure
	  accepts only in
	  an alpha version. So the distribution check of a beta or a
	  release tarball
	  failed in that cell. The internal options now have their own
	  cell, and the
	  matrix skips it outside an alpha.

2026-11-01 20:47  aleidinger

	* include/lame.h, libmp3lame/lame.c, libmp3lame/set_get.c:
	  Return -2 for every failed allocation in an encode call
	  
	  An encode call returned -10 when the resampler could not allocate
	  its
	  buffers, and -2 when any other allocation failed. The
	  documentation of the
	  encode and flush functions now lists each return value as a
	  number.

2026-11-01 20:47  aleidinger

	* doc/doxygen/vector-dispatch.dox:
	  Remove project-internal notes from the vector dispatch page
	  
	  The page named the CPU of a test machine and listed lessons from
	  how an
	  experiment was run. Neither describes the code.

2026-11-01 11:38  aleidinger

	* ChangeLog:
	  Update the ChangeLog.

2026-11-01 11:31  aleidinger

	* libmp3lame/version.h:
	  Enter beta phase.

2026-10-31 16:56  aleidinger

	* test/unit/test_albumart_pipe.sh:
	  Stop the album art pipe test from leaving a process behind
	  
	  The test opened the pipe once more to release a writer that might
	  still be waiting. When lame had already read the pipe, the writer
	  was
	  gone and that reader waited forever, so a run could leave a
	  process
	  behind. The test now ends the writer itself.

2026-10-31 16:23  aleidinger

	* maintainer/coverage-mp3x.py:
	  Check for the mp3x driver's GObject modules without unused
	  imports
	  
	  The probe now loads the Atspi and Gio modules with importlib. It
	  no
	  longer binds names it never uses, so pyflakes stops reporting
	  them.

2026-10-31 16:23  aleidinger

	* maintainer/coverage-mp3x.py:
	  Step the mp3x coverage driver past an early pause
	  
	  mp3x shows the frame READ_AHEAD frames behind the newest one.
	  When the
	  pause comes before that frame, Back stays disabled for up to
	  READ_AHEAD + 2 steps. The driver now steps until Back is enabled.

2026-10-28 16:53  aleidinger

	* frontend/lame_main.c, test/unit/test_nogap_failures.sh:
	  Treat a --nogap file that cannot be opened like one that cannot
	  be read or written

2026-10-28 16:52  aleidinger

	* Makefile.am, Makefile.in, doc/maintainer-coverage.md,
	  maintainer/coverage-mp3x.py, maintainer/coverage-run.sh,
	  maintainer/coverage-workload.txt:
	  Drive the mp3x analyzer in the coverage harness

2026-10-28 16:52  aleidinger

	* frontend/mp3x_ui.c:
	  Update the menu and the buttons of mp3x when the analysis pauses,
	  plays or advances

2026-10-28 16:52  aleidinger

	* Makefile.am, Makefile.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/doxygen/internal.md,
	  doc/doxygen/maintainer.md, doc/maintainer-analyze-msvc.md,
	  maintainer/analyze, maintainer/analyze-msvc.ps1,
	  maintainer/analyze/baseline.txt:
	  Add a harness for the MSVC static analyzer

2026-10-28 16:52  aleidinger

	* test/clients/acm_test.cpp:
	  Uninitialize COM in the ACM test only after it started

2026-10-28 16:51  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/ADbg/ADbg.cpp,
	  ACM/AEncodeProperties.cpp, ACM/main.cpp:
	  Print pointers with %p in the ACM codec's debug output

2026-10-28 16:51  aleidinger

	* doc/maintainer-build-matrix.md, maintainer/gen-build-matrix.sh:
	  Add a debug build to the build matrix

2026-10-28 16:51  aleidinger

	* doc/maintainer-abi.md, maintainer/abi/libmp3lame.abignore:
	  Do not count the FILE of the C library in the ABI check

2026-10-28 16:50  aleidinger

	* libmp3lame/takehiro.c:
	  Report the quantizer's debug note on the debug channel

2026-10-28 16:50  aleidinger

	* libmp3lame/lame.c:
	  Do not divide by a free format bitrate of 0 before refusing it

2026-10-28 16:50  aleidinger

	* configure, configure.ac:
	  Stop the compiler from raising floating-point exceptions of its
	  own in a debug build

2026-10-28 16:49  aleidinger

	* frontend/get_audio.c:
	  Count samples above full scale without raising an exception for a
	  NaN

2026-10-28 16:49  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in:
	  Fix make dist, which a literal \n in an EXTRA_DIST line of the
	  unit tests broke

2026-10-24 12:16  aleidinger

	* doc/html/history.html, libmp3lame/VbrTag.c, libmp3lame/util.h,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_lametag_room.c:
	  Give a stream of 4 GiB or more the largest size the tag fields
	  can hold, instead of a wrapped one

2026-10-24 12:16  aleidinger

	* doc/html/history.html, libmp3lame/VbrTag.c,
	  test/unit/test_encode_clipgain.c:
	  Write the LAME tag's peak for input louder than an int can scale,
	  instead of a wrong value

2026-10-24 12:15  aleidinger

	* libmp3lame/set_get.c, test/unit/test_set_get.c:
	  Refuse an ATH formula outside 0 to 5, which the LAME tag cannot
	  hold

2026-10-24 12:15  aleidinger

	* doc/html/history.html, frontend/parse.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_albumart_pipe.sh:
	  Report album art whose size cannot be read, instead of a lack of
	  memory

2026-10-24 12:14  aleidinger

	* frontend/get_audio.c, test/unit/test_get_audio_wav.c:
	  Compute the byte rate of a WAV header without overflowing an int

2026-10-24 12:14  aleidinger

	* doc/html/history.html, frontend/parse.c,
	  test/unit/test_classify_suffix.c:
	  On Windows and OS/2, accept forward slashes in the paths that
	  output names are built from

2026-10-24 12:14  aleidinger

	* libmp3lame/l3side.h, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/takehiro.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h, libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/xmm_quantize_sub.c:
	  Use the name of the granule size in the quantizer

2026-10-24 12:14  aleidinger

	* libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/fft.c, libmp3lame/lame.c,
	  libmp3lame/newmdct.c, libmp3lame/psymodel.c,
	  libmp3lame/set_get.c, libmp3lame/util.c:
	  Name the size of a granule, and use the name in the framing and
	  the psychoacoustic model

2026-10-17 16:23  aleidinger

	* doc/html/history.html, include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/lame.c, libmp3lame/set_get.c, libmp3lame/util.c,
	  libmp3lame/util.h, test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_free_format.c:
	  Bound a free format bitrate to 640 kbit/s, and compare bitrates
	  without overflow

2026-10-17 16:23  aleidinger

	* include/lame.h, libmp3lame/mpglib_interface.c:
	  Document the room hip_decode() needs, and the call for a fixed
	  buffer

2026-10-17 16:22  aleidinger

	* doc/html/history.html, frontend/main.c, frontend/main.h,
	  frontend/parse.c, test/unit/test_parse_id3.c:
	  Convert tag text from the character set it arrives in

2026-10-17 16:22  aleidinger

	* frontend/get_audio.c, frontend/get_audio.h, frontend/lame_main.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_short_mp3.sh:
	  Finish the output of an input that is shorter than its LAME tag
	  says

2026-10-17 16:21  aleidinger

	* frontend/lame_main.c, frontend/parse.c, frontend/parse.h,
	  test/unit/test_classify_suffix.c:
	  Build the output file name within the size of its buffer

2026-10-17 16:21  aleidinger

	* libmp3lame/VbrTag.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_lametag_room.c:
	  Check that the LAME tag is as long as the room reserved for it

2026-10-17 16:21  aleidinger

	* libmp3lame/lame.c, libmp3lame/util.h:
	  Move the encoder's input window instead of the samples

2026-10-17 16:21  aleidinger

	* frontend/get_audio.c:
	  Read input files through a 64 KiB buffer

2026-10-17 16:20  aleidinger

	* frontend/get_audio.c, test/unit/test_get_audio_float.c:
	  Read input samples into the encoder's buffers in one pass

2026-10-17 16:20  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  test/unit/test_get_audio_wav.c:
	  Read unsigned 8-bit samples without an offset

2026-10-14 11:42  aleidinger

	* Makefile.am, Makefile.in, doc/maintainer-abi.md,
	  maintainer/abi/libmp3lame.abignore, maintainer/abicheck.sh:
	  Do not count the opaque encoder and decoder state in the ABI
	  check

2026-10-14 11:41  aleidinger

	* doc/html/detailed.html, doc/html/history.html, doc/man/lame.1,
	  frontend/get_audio.c, test/unit/test_decode_swap.sh:
	  Warn that a libsndfile build ignores -x for a file with a header

2026-10-14 11:41  aleidinger

	* doc/html/history.html, libmp3lame/bitstream.c,
	  test/unit/test_set_get.c:
	  Do not report an error from lame_get_size_mp3buffer()

2026-10-14 11:41  aleidinger

	* doc/html/history.html, frontend/lame_main.c, frontend/parse.c,
	  include/lame.h, libmp3lame/id3tag.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_id3tag.c,
	  test/unit/test_replaygain_pad_max.sh:
	  Limit the ID3v2 padding in the library to the frontend's maximum
	  
	  On a 32-bit build, a padding near 4 GB wrapped the tag size
	  before it was
	  compared with the 28-bit limit, and the tag was written past the
	  buffer.

2026-10-14 11:40  aleidinger

	* doc/maintainer-build-matrix.md, frontend/mp3rtp.c,
	  maintainer/gen-build-matrix.sh:
	  Build mp3rtp in the build matrix, and fix the warning it shows
	  
	  The level meter compared its column with an unsigned size; the
	  bound is now
	  the last column before the line end.

2026-10-14 11:40  aleidinger

	* test/clients/acm_test.cpp:
	  Let the ACM test drive a dialog only once it is shown
	  
	  The tab and focus checks failed whenever the timer fired before
	  the dialog
	  appeared.

2026-10-14 11:40  aleidinger

	* libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/l3side.h, libmp3lame/newmdct.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c, libmp3lame/vector/xmm_quantize_sub.c,
	  test/unit/test_quantize_lines_vector.c:
	  Copy only the part of gr_info that a quantization search changes
	  
	  gr_info is split into what is fixed for the granule, the Huffman
	  division
	  and the quantization. The searches keep their best result as
	  copies of the
	  last two, and xrpow is saved only in the modes that restore it.

2026-10-14 11:39  aleidinger

	* frontend/get_audio.c, frontend/main.h,
	  libmp3lame/gain_analysis.h, libmp3lame/lame_global_flags.h,
	  libmp3lame/util.h:
	  Reorder struct members to remove the padding between them

2026-10-14 11:38  aleidinger

	* frontend/get_audio.c:
	  Name --mp3input when --decode cannot tell the input is an MP3

2026-10-14 11:38  aleidinger

	* USAGE:
	  Add the notes that USAGE refers to

2026-10-14 11:38  aleidinger

	* INSTALL:
	  Bring INSTALL up to date with the build

2026-10-14 11:38  aleidinger

	* doc/man/lame.1:
	  Fix the warnings groff and mandoc give for lame(1)

2026-10-14 11:37  aleidinger

	* USAGE, doc/html/detailed.html, doc/man/lame.1:
	  Name libmpg123 as the decoder in the --decode documentation

2026-10-14 11:37  aleidinger

	* frontend/parse.c, test/unit/test_help_exit.c:
	  Report the missing decoder for the options that need it

2026-10-14 11:37  aleidinger

	* vc_solution/vs_lame_enc_dll.vcxproj,
	  vc_solution/vs_lame_enc_dll.vcxproj.filters:
	  Give the Blade DLL built with MSBuild its version resource

2026-10-14 11:37  aleidinger

	* config.h.in, configMS.h, configure, configure.ac,
	  mac/Precompile_Common.h:
	  Remove the type size checks and the fixed-width integer type
	  fallbacks

2026-10-14 11:36  aleidinger

	* ACM/ACM.h, ACM/ACMStream.h, ACM/ADbg/ADbg.cpp, ACM/ADbg/ADbg.h,
	  ACM/AEncodeProperties.h, ACM/DialogFont.h, ACM/SettingText.h,
	  ACM/adebug.h, Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h,
	  Makefile.MSVC, configure, configure.ac, dshow/Encoder.h,
	  dshow/REG.H, dshow/SystemFontPage.h, dshow/iaudioprops.h,
	  frontend/Makefile.am, frontend/Makefile.in, frontend/parse.c,
	  libmp3lame/lame.rc, libmp3lame/set_get.h, libmp3lame/util.c,
	  libmp3lame/version.c, libmp3lame/version.h, mpglib/mpglib.h,
	  test/clients/acm_test.cpp, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_set_get.c,
	  vc_solution/vs_lame_lame.vcxproj,
	  vc_solution/vs_lame_mp3rtp.vcxproj,
	  vc_solution/vs_lame_mp3x.vcxproj,
	  vc_solution/vs_libmp3lame_dll.vcxproj:
	  Rename the identifiers that are reserved to the implementation

2026-10-14 11:35  aleidinger

	* dshow/resource.h, frontend/brhist.c, frontend/lame_main.c,
	  frontend/mp3x_ui.c, frontend/rtp.c, frontend/timestatus.c,
	  libmp3lame/bitstream.c, libmp3lame/machine.h,
	  libmp3lame/set_get.c, libmp3lame/set_get.h, libmp3lame/tables.c,
	  libmp3lame/takehiro.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c, test/clients/blade_test.cpp:
	  Fix printf types, hidden prototypes, linkage, const casts and
	  dead code

2026-10-14 11:35  aleidinger

	* libmp3lame/id3tag.c:
	  Bound the parts of the encoder version text in the ID3v2 tag

2026-10-14 11:35  aleidinger

	* frontend/get_audio.c:
	  Split the reading of one frame by the kind of input

2026-10-14 11:35  aleidinger

	* frontend/get_audio.c, frontend/get_audio.h, frontend/lame_main.c,
	  frontend/mp3rtp.c, frontend/mp3x_core.c, frontend/mp3x_core.h,
	  frontend/mp3x_session.c, frontend/mp3x_session.h,
	  frontend/mp3x_ui.c, test/unit/test_get_audio_aiff.c,
	  test/unit/test_get_audio_float.c,
	  test/unit/test_get_audio_truncated.c,
	  test/unit/test_get_audio_wav.c:
	  Give the input reader an object instead of file-scope state

2026-10-14 11:34  aleidinger

	* configure.ac, frontend/Makefile.am, frontend/Makefile.in:
	  Compile only the frame analyzer with the GTK flags

2026-10-14 11:34  aleidinger

	* libmp3lame/quantize_pvt.c:
	  Show 0 in the frame analyzer for bands the encoder does not
	  evaluate, not unset values

2026-10-14 11:33  aleidinger

	* libmp3lame/lame.c:
	  Compute the default VBR lowpass in one place and drop a repeated
	  default

2026-10-14 11:33  aleidinger

	* libmp3lame/mpglib_interface.c:
	  Name the decoder's output format instead of passing a flag and
	  byte pointers

2026-10-14 11:33  aleidinger

	* libmp3lame/id3tag.c:
	  Classify the ID3v2 frames once for the size and the writing

2026-10-14 11:32  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/util.h:
	  Write the Xing frame's bitrate once and the LAME tag's 16-bit
	  fields with one helper

2026-10-14 11:32  aleidinger

	* libmp3lame/psymodel.c:
	  Move the model's history, attack thresholds and line count check
	  out of its setup

2026-10-14 11:31  aleidinger

	* libmp3lame/psymodel.c:
	  Move the pre-echo control and the hand-back of the previous
	  granule into functions

2026-10-14 11:31  aleidinger

	* libmp3lame/vbrquantize.c:
	  Split the VBR frame encoder into helpers along its stages

2026-10-14 11:31  aleidinger

	* libmp3lame/encoder.c:
	  Split the frame encoder into helpers along its stages

2026-10-14 11:31  aleidinger

	* libmp3lame/fft.c:
	  Write out the windowed products of the FFT instead of one-use
	  macros

2026-10-14 11:30  aleidinger

	* libmp3lame/encoder.h, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c:
	  Move the search for the highest coded line out of calc_xmin()

2026-10-14 11:30  aleidinger

	* libmp3lame/quantize.c:
	  Write the outer loop's gain search once

2026-10-14 11:29  aleidinger

	* libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/util.h:
	  Remove the noise shaping refine mode, which nothing selects

2026-10-14 11:29  aleidinger

	* libmp3lame/encoder.c, libmp3lame/encoder.h, libmp3lame/newmdct.c:
	  Name the window tables and the subband constants of the MDCT

2026-10-14 11:29  aleidinger

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/vbrquantize.c:
	  Name the global gain of the unit step and the largest global gain

2026-10-14 11:28  aleidinger

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  Drop stores nothing reads, and store the noise cache's gain once

2026-10-14 11:28  aleidinger

	* libmp3lame/machine.h, libmp3lame/quantize_pvt.h:
	  Turn the POW20 and IPOW20 table lookups into inline functions

2026-10-14 11:28  aleidinger

	* libmp3lame/vbrquantize.c:
	  Drop a branch that the minimum above it makes unreachable

2026-10-14 11:27  aleidinger

	* libmp3lame/quantize.c, libmp3lame/reservoir.c,
	  libmp3lame/takehiro.c, libmp3lame/util.h:
	  Name the bits of the substep shaping setting

2026-10-14 11:27  aleidinger

	* libmp3lame/bitstream.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c:
	  Name the limits of the region counts

2026-10-14 11:27  aleidinger

	* libmp3lame/bitstream.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c:
	  Name the two scalefactor marker values

2026-10-14 11:27  aleidinger

	* libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c:
	  Use the scalefactor band count names instead of 39, 38 and 21

2026-10-10 15:25  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in:
	  Continue the unit tests' file list onto its last line, so that
	  make dist works

2026-10-10 15:22  aleidinger

	* doc/html/history.html:
	  Sum up the changes of the ACM codec and the DirectShow filter in
	  the history

2026-10-10 15:22  aleidinger

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  dshow/Encoder.cpp, dshow/Mpegac.cpp, dshow/PropPage_adv.cpp,
	  dshow/Property.rc, dshow/iaudioprops.h, dshow/resource.h,
	  test/clients/acm_test.cpp, test/clients/dshow_test.cpp:
	  Stop offering dual channel, which LAME does not implement

2026-10-10 15:22  aleidinger

	* ACM/ACM.cpp, ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  ACM/SettingText.h, ACM/lame_acm.xml, dshow/Encoder.cpp,
	  dshow/Mpegac.cpp, dshow/PropPage.cpp, test/clients/acm_test.cpp,
	  test/clients/ctest.h, test/clients/dshow_test.cpp:
	  Use LAME's defaults in the ACM codec and the DirectShow filter

2026-10-10 15:21  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  ACM/DialogFont.cpp, ACM/DialogFont.h, ACM/Makefile.am,
	  ACM/Makefile.in, ACM/SettingText.cpp, ACM/SettingText.h,
	  ACM/TODO, ACM/acm.rc, ACM/lame_acm.xml, ACM/resource.h,
	  dshow/Encoder.cpp, dshow/Encoder.h, dshow/Makefile.am,
	  dshow/Makefile.in, dshow/Mpegac.cpp, dshow/Mpegac.h,
	  dshow/PropPage.cpp, dshow/PropPage.h, dshow/PropPage_adv.cpp,
	  dshow/PropPage_adv.h, dshow/Property.rc,
	  dshow/SystemFontPage.cpp, dshow/SystemFontPage.h,
	  dshow/aboutprp.cpp, dshow/aboutprp.h, dshow/elogo.ico,
	  dshow/iaudioprops.h, dshow/resource.h, test/Makefile.am,
	  test/Makefile.in, test/clients/acm_test.cpp,
	  test/clients/dialogcheck.h, test/clients/dshow_test.cpp,
	  test/common/mp3frame.h, vc_solution/vs_lame_acm.vcxproj,
	  vc_solution/vs_lame_acm.vcxproj.filters,
	  vc_solution/vs_lame_acm_test.vcxproj,
	  vc_solution/vs_lame_dshow.vcxproj,
	  vc_solution/vs_lame_dshow.vcxproj.filters,
	  vc_solution/vs_lame_dshow_test.vcxproj:
	  Redesign the settings of the ACM codec and the DirectShow filter
	  
	  The ACM codec offers CBR, ABR and VBR formats, each family with
	  its own
	  range. Its settings dialog has two tabs, Formats and Encoding,
	  and adds
	  the encoding quality, VBR bitrate limits and the advanced
	  settings of
	  the filter.
	  
	  The DirectShow filter drops voice mode, no short blocks and mode
	  fixed,
	  and adds the private bit, the bit reservoir and ABR, also through
	  the
	  new IAudioEncoderProperties2 interface.
	  
	  Both use the desktop font, the same names and texts for the same
	  settings, and the same About box.

2026-10-10 15:21  aleidinger

	* ACM/ACMStream.cpp, test/clients/acm_test.cpp:
	  Count the frames the bit reservoir holds back in the ACM codec's
	  output size

2026-10-10 15:21  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h, ACM/AEncodeProperties.cpp, ACM/main.cpp,
	  ACM/tinyxml/tinyxmlparser.cpp, dshow/Mpegac.cpp, dshow/Mpegac.h,
	  dshow/PropPage.cpp, dshow/PropPage.h, dshow/PropPage_adv.cpp,
	  dshow/PropPage_adv.h, dshow/REG.CPP:
	  Fix the warnings that MinGW reports in the ACM codec and the
	  DirectShow filter

2026-10-10 15:20  aleidinger

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  test/clients/acm_test.cpp:
	  Keep the ACM codec's settings when its dialog opens and after a
	  failed load

2026-10-10 15:20  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h, ACM/main.cpp, test/clients/acm_test.cpp:
	  Use the ACM codec's own module when its driver has none

2026-10-10 15:20  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h, ACM/TODO,
	  ACM/acm.rc, ACM/adebug.h, ACM/lameACM.def, ACM/main.cpp,
	  ACM/resource.h:
	  Remove dead code from the ACM codec and define each settings name
	  once

2026-10-10 15:19  aleidinger

	* frontend/lame_main.c:
	  Reserve the ReplayGain tag frames after lame_init_params()
	  
	  The check whether LAME builds its own ID3v2 tag calls
	  lame_get_id3v2_tag(), which needs lame_init_params() first.

2026-10-10 15:19  aleidinger

	* doc/doxygen/api.dox, include/lame.h, libmp3lame/id3tag.c,
	  libmp3lame/lame.c:
	  Document which calls belong before lame_init_params() and which
	  after it

2026-10-10 15:18  aleidinger

	* doc/html/detailed.html, frontend/parse.c, libmp3lame/set_get.c:
	  Describe the values of --substep as the bits they are

2026-10-10 15:18  aleidinger

	* libmp3lame/mpglib_interface.c:
	  Return -1 from hip_decode1_unclipped() without a decoder, as its
	  siblings do

2026-10-10 15:18  aleidinger

	* doc/html/history.html, libmp3lame/lame.c, libmp3lame/util.c,
	  libmp3lame/util.h, test/unit/test_decoder_init_fail.c:
	  Undo a failed lame_init_params(), so that a second call starts
	  afresh

2026-10-10 15:18  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/lame_main.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_get_audio_truncated.c, test/unit/test_stubs.c:
	  Stop at an input that ends inside a sample frame, finish the file
	  and fail

2026-10-10 15:17  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in:
	  Link the two unit tests that use internal library functions
	  statically

2026-10-08 14:29  aleidinger

	* doc/html/history.html, frontend/lame_main.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_output_complete.sh:
	  Exit with an error when the end of the output cannot be written

2026-10-08 14:28  aleidinger

	* ACM/TODO:
	  Drop the ACM TODO's lame_close() bug, which no test reproduces

2026-10-08 14:28  aleidinger

	* ACM/ACM.cpp, doc/html/history.html, test/clients/acm_test.cpp:
	  Read fdwFlags of an MP3 format as the padding mode in the ACM
	  codec

2026-10-08 14:28  aleidinger

	* ACM/ACMStream.cpp, ACM/ACMStream.h, doc/html/history.html,
	  test/clients/acm_test.cpp, vc_solution/vs_lame_acm_test.vcxproj:
	  Encode all of the input of DirectShow's ACM Wrapper in the ACM
	  codec

2026-10-08 14:27  aleidinger

	* Dll/BladeMP3EncDLL.c, doc/html/history.html,
	  test/clients/blade_test.cpp:
	  Keep the CBR preset of lame_enc.dll within the bitrate presets,
	  and check it

2026-10-08 14:27  aleidinger

	* doc/html/history.html, libmp3lame/presets.c,
	  libmp3lame/set_get.c, libmp3lame/set_get.h,
	  test/unit/test_set_get.c:
	  Return -1 from lame_set_preset() for a value that is no preset

2026-10-08 14:27  aleidinger

	* doc/html/history.html, dshow/Encoder.cpp, dshow/Encoder.h,
	  dshow/Mpegac.cpp, test/clients/dshow_test.cpp:
	  Deliver the first MP3 frame of the DirectShow filter's frame
	  output as preroll, and the last

2026-10-08 14:27  aleidinger

	* doc/html/history.html, dshow/Mpegac.cpp, dshow/Mpegac.h,
	  test/clients/dshow_test.cpp:
	  Discard the encoder's data on a DirectShow flush and start a new
	  encoder after it

2026-10-08 14:27  aleidinger

	* frontend/lame_main.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_preserve_modtime.sh:
	  Give an output whose writing failed no times from
	  --preserve-modtime

2026-10-08 14:26  aleidinger

	* USAGE, doc/html/history.html, frontend/get_audio.c,
	  frontend/main.h, frontend/parse.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_decode_swap.sh,
	  test/unit/test_get_audio_wav.c:
	  Swap only the output bytes with -x when decoding

2026-10-08 14:26  aleidinger

	* doc/html/history.html, dshow/Mpegac.cpp,
	  test/clients/dshow_test.cpp:
	  Stop the DirectShow filter's encode when it gets no output buffer

2026-10-08 14:26  aleidinger

	* test/clients/dshow_test.cpp:
	  Test the DirectShow filter's frame output and its time stamps
	  after a gap

2026-10-08 14:26  aleidinger

	* ACM/ACMStream.cpp, Dll/BladeMP3EncDLL.c, doc/html/history.html,
	  dshow/Encoder.cpp, test/clients/acm_test.cpp,
	  test/clients/blade_test.cpp, test/clients/ctest.h,
	  test/clients/dshow_test.cpp:
	  Keep LAME's messages of the Windows clients off the stderr of
	  their host

2026-10-08 14:25  aleidinger

	* doc/html/history.html, include/lame.h:
	  Give the report function type of lame.h the calling convention of
	  the library
	  
	  A program built with __stdcall as its default, as the DirectShow
	  filter is,
	  declared its report functions with that convention, and the
	  library called
	  them with __cdecl.

2026-10-08 14:25  aleidinger

	* dshow/Encoder.cpp, test/clients/dshow_test.cpp:
	  Start no DirectShow encode with a setting that LAME rejects

2026-10-08 14:24  aleidinger

	* Dll/BladeMP3EncDLL.c, doc/html/history.html,
	  test/clients/blade_test.cpp:
	  Refuse a lame_enc.dll stream with a setting that LAME rejects

2026-10-08 14:24  aleidinger

	* ACM/ACMStream.cpp, test/clients/acm_test.cpp:
	  Open no ACM stream with a setting that LAME rejects

2026-10-08 14:24  aleidinger

	* doc/html/history.html:
	  Describe the refactoring in the history

2026-10-08 14:24  aleidinger

	* frontend/get_audio.c, frontend/lame_main.c, frontend/main.h,
	  frontend/mp3rtp.c, frontend/mp3x_core.c, frontend/mp3x_session.c,
	  frontend/mp3x_session.h, frontend/mp3x_ui.c, frontend/parse.c,
	  frontend/timestatus.c, test/unit/parse_test_stubs.c,
	  test/unit/test_get_audio_float.c, test/unit/test_get_audio_wav.c,
	  test/unit/test_stubs.c:
	  Keep the frontend settings in one configuration struct with one
	  set of defaults

2026-10-08 14:24  aleidinger

	* doc/html/history.html, dshow/REG.CPP,
	  test/clients/dshow_test.cpp:
	  Close the registry key that the DirectShow filter creates to save
	  its settings

2026-10-08 14:23  aleidinger

	* doc/html/history.html, dshow/Mpegac.cpp,
	  test/clients/dshow_test.cpp:
	  Return a rejected delivery from the DirectShow filter's Receive()
	  
	  When the downstream filter rejected a sample, the filter went on
	  encoding
	  and delivering. It now stops and returns the result, so the
	  upstream filter
	  stops sending.

2026-10-08 14:23  aleidinger

	* dshow/Encoder.cpp, dshow/Encoder.h, dshow/Mpegac.cpp,
	  dshow/Mpegac.h:
	  Split the DirectShow filter's output by mode, and name its time
	  and frame values

2026-10-08 14:23  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h, doc/html/history.html,
	  test/clients/acm_test.cpp:
	  Suggest no format for an MP3 source from the ACM codec
	  
	  The codec does not decode, and an MP3 to PCM stream does not
	  open, so a
	  suggestion for one only made the application fail later.

2026-10-08 14:22  aleidinger

	* Dll/BladeMP3EncDLL.c:
	  Split beInitStream() of lame_enc.dll into its setup steps

2026-10-08 14:22  aleidinger

	* ACM/ACM.cpp, doc/html/history.html, test/clients/acm_test.cpp:
	  Return the format tag index asked for and the largest format from
	  the ACM codec

2026-10-08 14:22  aleidinger

	* ACM/ACM.cpp:
	  Return from the ACM codec's format tag query where the result is
	  known

2026-10-08 14:22  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  doc/html/history.html, test/clients/acm_test.cpp:
	  Fix the ACM codec's stream close, size query and used source
	  length
	  
	  A close after a query open deleted whatever the application had
	  put into
	  the stream's instance data; the codec keeps its stream in the
	  driver's own
	  field now. The codec also returns a source size for a destination
	  size, and
	  does not count a partial sample frame at the end of the source as
	  used. A
	  conversion that ends the MP3 stream checks that the flush fits
	  before it
	  encodes anything.

2026-10-08 14:21  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  ACM/AEncodeProperties.h, doc/html/history.html,
	  test/clients/acm_test.cpp:
	  Keep one encoder per ACM stream, from open to close
	  
	  Each prepared header started a new encoder, which leaked the
	  previous one
	  and restarted the MP3 stream in the middle. The encoder now
	  starts when the
	  stream opens, is flushed by a conversion with the END flag, or by
	  the last
	  conversion of a client that does not send it, and starts again,
	  with the
	  settings as they are then, with the next conversion or on START.

2026-10-08 14:21  aleidinger

	* doc/html/history.html, frontend/lame_main.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_nogap_failures.sh:
	  Go on after an unreadable file in a --nogap run and stop on an
	  output or encoder error

2026-10-08 14:21  aleidinger

	* doc/html/history.html, frontend/brhist.c, frontend/timestatus.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_brhist_cbr.sh:
	  Show the bitrate of a CBR encode as one row of the bitrate
	  histogram

2026-10-08 14:21  aleidinger

	* frontend/get_audio.c, test/unit/test_get_audio_wav.c:
	  Write the samples of --decode in 16-bit units

2026-10-08 14:20  aleidinger

	* frontend/lame_main.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_nogap_files.sh:
	  Allocate the --nogap file name buffers of lame on the heap

2026-10-08 14:20  aleidinger

	* libmp3lame/encoder.c, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h:
	  Return nothing from the psychoacoustic model, which cannot fail

2026-10-08 14:20  aleidinger

	* libmp3lame/lame.c:
	  Split lame_print_internals() into its three sections and name
	  functions

2026-10-08 14:19  aleidinger

	* libmp3lame/lame.c:
	  Describe the sample layout of each encode entry point in one
	  constant

2026-10-08 14:19  aleidinger

	* ACM/ACMStream.cpp, ACM/ACMStream.h, ACM/AEncodeProperties.cpp,
	  ACM/AEncodeProperties.h, dshow/Encoder.cpp, dshow/Encoder.h:
	  Fix the const qualifiers of the ACM codec and DirectShow encoder
	  classes

2026-10-08 14:19  aleidinger

	* dshow/Encoder.cpp, dshow/Encoder.h:
	  Hold the DirectShow encoder's buffers in members and containers

2026-10-08 14:19  aleidinger

	* ACM/ACM.cpp, dshow/Encoder.cpp, dshow/Mpegac.cpp:
	  Name the format values of the ACM codec and the DirectShow filter

2026-10-08 14:18  aleidinger

	* ACM/ACM.cpp, ACM/ADbg/ADbg.cpp, ACM/ADbg/ADbg.h,
	  ACM/AEncodeProperties.cpp, Dll/BladeMP3EncDLL.c, Dll/Example.cpp,
	  dshow/Mpegac.cpp, dshow/REG.CPP, dshow/REG.H, dshow/aboutprp.cpp:
	  Use bounded string calls in the Windows clients
	  
	  The DirectShow filter's registry class keeps only the members the
	  filter
	  calls, and takes value names as constant strings.

2026-10-08 14:18  aleidinger

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h, Dll/Makefile.mingw32,
	  vc_solution/vs_lame_acm.vcxproj:
	  Check the lame_enc.dll export prototypes against their
	  definitions

2026-10-08 14:18  aleidinger

	* ACM/ACM.h, ACM/ACMStream.h, ACM/ADbg/ADbg.h:
	  Include what the ACM codec's headers use

2026-10-08 14:18  aleidinger

	* Dll/BladeMP3EncDLL.c:
	  Remove dead code from lame_enc.dll
	  
	  Each encode and flush call gets the output buffer size of its own
	  stream, not
	  that of the stream initialised last. Debug output goes to the
	  debugger in
	  debug builds only.

2026-10-08 14:17  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h, ACM/main.cpp,
	  configMS.h, test/clients/acm_test.cpp:
	  Remove dead code from the ACM codec
	  
	  The codec uses the standard operator new and checks for a failed
	  allocation
	  with std::nothrow. When the configuration dialog is cancelled,
	  the codec
	  returns DRVCNF_CANCEL to the host.

2026-10-08 14:17  aleidinger

	* Dll/BladeMP3EncDLL.c, doc/html/history.html,
	  test/clients/blade_test.cpp:
	  Close the stream in lame_enc.dll when beInitStream() fails

2026-10-08 14:17  aleidinger

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  test/clients/acm_test.cpp, test/clients/ctest.h:
	  Stop leaking elements when the ACM codec saves its configuration

2026-10-08 14:17  aleidinger

	* dshow/PropPage.cpp:
	  Stop the DirectShow sample rate lookup at the end of its table

2026-10-08 14:16  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h, ACM/AEncodeProperties.cpp,
	  test/clients/acm_test.cpp, vc_solution/vs_lame_acm_test.vcxproj:
	  Fit the whole LAME version into the ACM codec's version texts
	  
	  In alpha and beta builds the version includes the build date and
	  time, which
	  overflowed the buffer of the configuration dialog.

2026-10-08 14:16  aleidinger

	* ACM/ACM.cpp, ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  doc/html/history.html, test/clients/acm_test.cpp:
	  Stop the ACM codec's ABR list at its minimum
	  
	  With a minimum below the step, the list wrapped around and grew
	  without end,
	  in the codec's format list and in the configuration dialog.

2026-10-08 14:16  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_pipe_length.sh:
	  Treat an MP3 input length that libmpg123 does not know as unknown

2026-10-08 14:15  aleidinger

	* INSTALL, frontend/get_audio.c:
	  Document that lame needs the libmp3lame of the same build

2026-10-08 14:15  aleidinger

	* frontend/parse.c:
	  Look up the option flags, the ID3 text options and the developer
	  flags in tables

2026-10-08 14:15  aleidinger

	* frontend/parse.c:
	  Parse the long options in one function per group and the short
	  options in another

2026-10-08 14:15  aleidinger

	* frontend/parse.c:
	  Keep the option parser's state between options in one struct

2026-10-08 14:14  aleidinger

	* frontend/rtp.c:
	  Build the RTP header of mp3rtp byte by byte

2026-10-08 14:14  aleidinger

	* frontend/lame_main.c:
	  Write the ID3v2 tag and the --flush step of the encode loop
	  through helpers

2026-10-08 14:14  aleidinger

	* frontend/parse.c, test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_preset_range.sh:
	  Name 8 kbps as the lowest --preset bitrate in its error message

2026-10-08 14:13  aleidinger

	* frontend/parse.c:
	  Look up the --preset aliases in a table

2026-10-08 14:13  aleidinger

	* frontend/get_audio.c:
	  Read the WAV and AIFF chunk sizes through two helpers and name
	  three header constants

2026-10-08 14:13  aleidinger

	* frontend/get_audio.c, test/unit/test_stubs.c:
	  Print the input reader's error messages through one function

2026-10-08 14:13  aleidinger

	* frontend/console.c, frontend/get_audio.c, frontend/lametime.c,
	  frontend/mp3x_plot.c, frontend/mp3x_ui.c, frontend/parse.c,
	  frontend/timestatus.c, test/unit/Makefile.am,
	  test/unit/Makefile.in:
	  Mark unused parameters in the frontend with LAME_UNUSED

2026-10-08 14:12  aleidinger

	* frontend/get_audio.c, frontend/get_audio.h, frontend/lame_main.c,
	  frontend/mp3rtp.c, frontend/mp3x_core.c, frontend/timestatus.c:
	  Name the frontend's frame buffer size and its unknown-length
	  value

2026-10-08 14:12  aleidinger

	* frontend/console.c, frontend/console.h, frontend/get_audio.c,
	  frontend/main.c, frontend/parse.c:
	  Remove dead code from the frontend and name the MP3 decoder delay

2026-10-08 14:12  aleidinger

	* frontend/parse.c:
	  Keep the terminator when the OS/2 build renames UTF-16 to UCS-2
	  
	  Not compiled: the code is built only with the OS/2 compiler.

2026-10-08 14:12  aleidinger

	* frontend/get_audio.c, test/unit/test_get_audio_float.c:
	  Return a read error when the frontend's sample buffer cannot grow

2026-10-08 14:12  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/lame_main.c,
	  test/unit/test_get_audio_wav.c:
	  Write the correct size into the header of a decoded WAV file
	  above 2 GiB

2026-10-08 14:11  aleidinger

	* doc/html/history.html, frontend/mp3rtp.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_mp3rtp_status.sh:
	  Stop mp3rtp with an error status on a rejected option or a read
	  error

2026-10-08 14:11  aleidinger

	* doc/html/history.html, include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/util.h, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_bitstream_reservoir.c:
	  Fail the encoder when the bits it wrote disagree with its bit
	  reservoir

2026-10-08 14:10  aleidinger

	* doc/html/history.html, libmp3lame/lame.c,
	  test/unit/test_encode_api.c:
	  Name an MPEG-2 stream as MPEG-2 in lame_print_internals()

2026-10-08 14:10  aleidinger

	* libmp3lame/VbrTag.h, libmp3lame/gain_analysis.h,
	  libmp3lame/id3tag.h, libmp3lame/lame_global_flags.h,
	  libmp3lame/machine.h, libmp3lame/util.c:
	  Make four library headers include what they use, and include
	  windows.h only where it is used

2026-10-08 14:10  aleidinger

	* libmp3lame/quantize.c, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c:
	  Mark unused parameters with LAME_UNUSED in the encoder core

2026-10-08 14:10  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/lame.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/util.h:
	  Decode the nspsytune fields once, keep the frame header bytes
	  unsigned, and name the reservoir flag

2026-10-08 14:09  aleidinger

	* libmp3lame/lame.c, libmp3lame/presets.c:
	  Set preset options through one macro and map V0-V9 by arithmetic

2026-10-08 14:09  aleidinger

	* libmp3lame/id3tag.c, libmp3lame/id3tag.h:
	  Name the ID3v2 text encodings and the two tag error results

2026-10-08 14:09  aleidinger

	* libmp3lame/id3tag.c:
	  Bound the formatted text of the ID3 tags by its buffer

2026-10-08 14:09  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/VbrTag.h:
	  Remove the Xing header reader that nothing calls

2026-10-08 14:08  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/gain_analysis.c,
	  libmp3lame/gain_analysis.h, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/mpglib_interface.c, libmp3lame/util.c,
	  libmp3lame/version.h:
	  Remove unused and debug-only code from the library
	  
	  The album histogram of the ReplayGain analysis is gone, so each
	  encoder
	  instance is 48 KB smaller.

2026-10-08 14:08  aleidinger

	* libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/tables.h,
	  libmp3lame/vbrquantize.c:
	  Remove dead and debug-only code from the encoder core

2026-10-08 14:07  aleidinger

	* doc/html/history.html, libmp3lame/lame.c,
	  test/unit/test_encode_api.c:
	  Check the instance in lame_print_config() and
	  lame_print_internals()

2026-10-08 14:07  aleidinger

	* libmp3lame/lame.c, test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_decoder_init_fail.c:
	  Fail lame_init_params() when the decoder that measures the output
	  cannot start

2026-10-08 14:07  aleidinger

	* doc/html/history.html, libmp3lame/id3tag.c,
	  test/unit/test_id3tag.c:
	  Check ID3 genre, year and track numbers before narrowing them to
	  int

2026-10-08 14:07  aleidinger

	* libmp3lame/id3tag.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_id3tag_alloc.c:
	  Return an error when a UTF-16 genre cannot be copied for lack of
	  memory

2026-10-08 14:07  aleidinger

	* doc/html/history.html, libmp3lame/lame.c, libmp3lame/util.c,
	  libmp3lame/util.h, test/unit/test_free_format.c:
	  Keep a free format bitrate above 65535 kbit/s whole when choosing
	  the lowpass

2026-10-08 14:06  aleidinger

	* doc/html/detailed.html, doc/html/history.html, doc/man/lame.1,
	  frontend/get_audio.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_swap_channel.sh:
	  Reject --swap-channel for an input with one channel
	  
	  With the option, a mono input was encoded as silence.

2026-10-08 14:06  aleidinger

	* frontend/mp3x_ui.c:
	  Save the mp3x screenshot and graphs through one action

2026-10-08 14:06  aleidinger

	* frontend/get_audio.c:
	  Write the deinterleave loops and the frame reader of the frontend
	  once

2026-10-08 14:06  aleidinger

	* doc/Doxyfile-internal.in, test/Makefile.am, test/Makefile.in,
	  test/clients/mp3frame.h, test/common, test/common/mp3frame.h,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_bitstream_headers.c, test/unit/test_encode_api.c,
	  vc_solution/vs_lame_acm_test.vcxproj,
	  vc_solution/vs_lame_blade_test.vcxproj,
	  vc_solution/vs_lame_dshow_test.vcxproj:
	  Share the MPEG frame reader of the client tests with the unit
	  tests
	  
	  The header moves to test/common. It finds the LAME tag through
	  the flags of
	  the Xing header instead of assuming all four optional fields.

2026-10-08 14:03  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_bitstream_headers.c,
	  test/unit/test_decode_scale.c, test/unit/test_decoder_api.c,
	  test/unit/test_encode.h, test/unit/test_encode_api.c:
	  Share the encode and collect loop of four unit tests

2026-10-08 14:03  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_encode_mono_interleaved.c,
	  test/unit/test_encode_nonfinite.c, test/unit/test_fixture.h,
	  test/unit/test_float_bits.h, test/unit/test_set_get.c:
	  Share the float bit patterns and the encoder setup of the unit
	  tests

2026-10-08 14:03  aleidinger

	* libmp3lame/psymodel.c:
	  Interpolate the spreading normalization of the long and the short
	  blocks with one function
	  
	  Built with clang and fast floating point math, a few encodes
	  change in
	  rounding only: the compiler folds the constants of the shared
	  function in a
	  different way. The perceptual quality is unchanged.

2026-10-08 14:02  aleidinger

	* doc/html/history.html, libmp3lame/psymodel.c:
	  Compute the tonality index and the band energies of the long and
	  the short blocks with one function each
	  
	  Built with fast floating point math, a few encodes change in
	  rounding only:
	  the shared code adds the same terms in a different order. The
	  perceptual
	  quality is unchanged.

2026-10-08 14:02  aleidinger

	* test/clients/acm_test.cpp:
	  Check that the ACM codec keeps a CBR stream at its bitrate

2026-10-08 14:01  aleidinger

	* include/lame.h:
	  Say in lame.h that the report functions are set before
	  lame_init_params()

2026-10-08 14:01  aleidinger

	* doc/html/history.html, frontend/mp3rtp.c:
	  Stop mp3rtp with an error when the final flush fails

2026-10-08 14:01  aleidinger

	* dshow/Encoder.h, dshow/Mpegac.cpp, dshow/Mpegac.h:
	  Remove the DirectShow filter's layer setting, which nothing can
	  set and the encoder does not read

2026-10-08 14:00  aleidinger

	* test/clients/dshow_test.cpp:
	  Check that the DirectShow filter's saved settings reach a new
	  filter

2026-10-08 14:00  aleidinger

	* Makefile.am, Makefile.in, doc/maintainer-gen-usage.md,
	  maintainer/check-optdoc.py:
	  Ship the check that the documentation names every long option

2026-10-08 14:00  aleidinger

	* doc/html/history.html, frontend/mp3x_session.c,
	  frontend/mp3x_session.h, frontend/parse.c, frontend/parse.h,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_classify_suffix.c:
	  Read the input format hint of lame and mp3x from one table of
	  file name suffixes

2026-10-08 13:59  aleidinger

	* doc/html/history.html, frontend/lame_main.c, frontend/mp3rtp.c,
	  frontend/mp3x_core.c, frontend/parse.c, frontend/parse.h,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_encode_error_text.c:
	  Print one message per encode error in lame, mp3rtp and mp3x

2026-10-08 13:59  aleidinger

	* ACM/ACMStream.cpp, Dll/BladeMP3EncDLL.c:
	  Write the settings into the debug logs of lame_enc.dll and the
	  ACM codec through lame_print_config(), and take their frame size
	  from lame_get_framesize()

2026-10-08 13:59  aleidinger

	* Dll/BladeMP3EncDLL.c, doc/html/history.html,
	  test/clients/blade_test.cpp, test/clients/mp3frame.h:
	  Round the ABR bitrate of lame_enc.dll's ABR preset to the nearest
	  kbit/s

2026-10-08 13:58  aleidinger

	* doc/html/history.html, frontend/parse.c,
	  test/unit/test_parse_pathlen.c:
	  Take a --highpass-width value below 16 as kHz, like --highpass

2026-10-08 13:58  aleidinger

	* libmp3lame/psymodel.c, libmp3lame/quantize_pvt.c:
	  Compute the absolute threshold of a band in one place

2026-10-08 13:58  aleidinger

	* libmp3lame/vbrquantize.c:
	  Share the budget split, the rebalancing and the bisection of the
	  VBR quantization

2026-10-08 13:58  aleidinger

	* libmp3lame/quantize.c:
	  Share the band amplification, the masking adjustment and the
	  budget scaling of the quantization loops

2026-10-08 13:57  aleidinger

	* libmp3lame/vector/Makefile.am, libmp3lame/vector/Makefile.in,
	  libmp3lame/vector/avx2_quantize_lines.c,
	  libmp3lame/vector/xmm_calc_sfb_noise.c,
	  libmp3lame/vector/xmm_quant_kernel.h,
	  libmp3lame/vector/xmm_quantize_lines.c,
	  libmp3lame/vector/xmm_quantize_sub.c,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Share the quantization steps of the SSE2 and AVX2 kernels

2026-10-08 13:57  aleidinger

	* libmp3lame/Makefile.am, libmp3lame/Makefile.in, libmp3lame/fft.c,
	  libmp3lame/fft_pvt.h, libmp3lame/vector/xmm_quantize_sub.c,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Share the rotation table and the first butterflies of the FFT
	  between its C and SSE2 forms

2026-10-08 13:57  aleidinger

	* libmp3lame/psymodel.c:
	  Share the threshold limits and the spectrum code of the long and
	  the short blocks in the psychoacoustic model

2026-10-08 13:56  aleidinger

	* test/clients/dshow_test.cpp, test/unit/test_bitstream_headers.c,
	  test/unit/test_encode_api.c, test/unit/test_encode_nonfinite.c,
	  test/unit/test_pkgconfig.sh,
	  test/unit/test_quantize_lines_vector.c,
	  test/unit/test_rtp_socket.c:
	  Share the repeated steps inside single test files

2026-10-08 13:56  aleidinger

	* test/clients/acm_test.cpp, test/clients/blade_test.cpp,
	  test/clients/ctest.h, test/clients/dshow_test.cpp:
	  Generate the test tone in one place in the client tests

2026-10-08 13:56  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_bitstream_headers.c, test/unit/test_encode_api.c,
	  test/unit/test_free_format.c, test/unit/test_report.h:
	  Collect the library's reports in one place in the unit tests

2026-10-08 13:56  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_bytes.h, test/unit/test_get_audio_aiff.c,
	  test/unit/test_get_audio_float.c, test/unit/test_get_audio_wav.c:
	  Give the audio reader tests one set of byte writers

2026-10-08 13:55  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_choose_table_vector.c, test/unit/test_cpu_probe.h,
	  test/unit/test_quantize_lines_vector.c:
	  Probe the AVX2 and AVX-512 tiers in one place in the vector tests

2026-10-08 13:55  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_choose_table_neon.c,
	  test/unit/test_choose_table_vector.c, test/unit/test_esc_cases.h:
	  Run the same escape-count cases in the x86 and the NEON vector
	  tests

2026-10-08 13:54  aleidinger

	* test/clients/acm_test.cpp, test/clients/dshow_test.cpp,
	  test/clients/mp3frame.h:
	  Scan MPEG frames in one place in the client tests

2026-10-08 13:54  aleidinger

	* test/clients/acm_test.cpp, test/clients/blade_test.cpp,
	  test/clients/ctest.h, test/clients/dshow_test.cpp:
	  Find the component under test in one place in the client tests

2026-10-08 13:54  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_encode_api.c, test/unit/test_id3tag.c,
	  test/unit/test_mem.h, test/unit/test_parse_id3.c:
	  Give the unit tests one set of byte search helpers

2026-10-08 13:54  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_fixture.h, test/unit/test_get_audio_aiff.c,
	  test/unit/test_get_audio_wav.c, test/unit/test_help_exit.c,
	  test/unit/test_id3tag.c, test/unit/test_parse_id3.c,
	  test/unit/test_set_get.c, test/unit/test_vector_routines.c:
	  Give the unit tests one encoder fixture

2026-10-08 13:53  aleidinger

	* dshow/Mpegac.cpp, dshow/Mpegac.h, test/clients/dshow_test.cpp:
	  Read and write the DirectShow filter's settings through one
	  accessor pair and one registry table

2026-10-08 13:53  aleidinger

	* frontend/main.c, frontend/parse.c:
	  Share the tag option table and the plain getenv of the frontend

2026-10-08 13:53  aleidinger

	* Dll/BladeMP3EncDLL.c, test/clients/blade_test.cpp,
	  test/clients/mp3frame.h:
	  Share the result mapping and the ABR bitrate range of
	  lame_enc.dll

2026-10-08 13:53  aleidinger

	* frontend/mp3x_plot.c, frontend/mp3x_plot.h, frontend/mp3x_ui.c:
	  Draw and export the mp3x graphs from one table

2026-10-08 13:53  aleidinger

	* frontend/console.c, frontend/console.h, frontend/lame_main.c,
	  frontend/mp3rtp.c, frontend/mp3x_session.c, frontend/parse.c,
	  frontend/parse.h:
	  Share the encoder setup of the frontend programs

2026-10-08 13:52  aleidinger

	* frontend/get_audio.c:
	  Share the sample width and byte order rules of the frontend's
	  input readers

2026-10-08 13:52  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h:
	  Build the ACM codec's format list with one function

2026-10-08 13:52  aleidinger

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h:
	  Look up the ACM codec's configuration file in one place

2026-10-08 13:51  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  ACM/DecodeStream.cpp, ACM/DecodeStream.h, ACM/Makefile.am,
	  ACM/Makefile.in, ACM/readme.txt, vc_solution/vs_lame_acm.vcxproj,
	  vc_solution/vs_lame_acm.vcxproj.filters,
	  vc_solution/vs_lame_acm_test.vcxproj:
	  Read the ACM codec's debug file setting in one place, and remove
	  its unbuildable decoder

2026-10-08 13:51  aleidinger

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  dshow/Encoder.cpp, dshow/Encoder.h, dshow/Mpegac.cpp,
	  dshow/PropPage.cpp, test/clients/acm_test.cpp,
	  test/clients/dshow_test.cpp:
	  Take the bitrates and sample rates of the ACM codec and the
	  DirectShow filter from the library

2026-10-08 13:51  aleidinger

	* Dll/BladeMP3EncDLL.c, Dll/Makefile.am, Dll/Makefile.in,
	  Dll/lametag_scan.h, dshow/Encoder.cpp,
	  vc_solution/vs_lame_dshow.vcxproj,
	  vc_solution/vs_lame_dshow.vcxproj.filters,
	  vc_solution/vs_lame_enc_dll.vcxproj,
	  vc_solution/vs_lame_enc_dll.vcxproj.filters:
	  Share the LAME tag frame search of lame_enc.dll and the
	  DirectShow filter

2026-10-08 13:51  aleidinger

	* libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/id3tag.c, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/obsolete_api.h,
	  libmp3lame/set_get.c, libmp3lame/util.c,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Write small repeated pieces of libmp3lame once

2026-10-08 13:50  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/id3tag.c, libmp3lame/id3tag.h,
	  libmp3lame/util.h:
	  Share the ID3 tag frame lookup, routing, writers and sizers

2026-10-08 13:50  aleidinger

	* libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c, libmp3lame/vbrquantize.c:
	  Compute the quantizer step of a scalefactor band in one function

2026-10-08 13:50  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h:
	  Write the MP3 frame header and the granule side information once

2026-10-08 13:49  aleidinger

	* libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/tables.c,
	  libmp3lame/tables.h, libmp3lame/util.c, libmp3lame/util.h:
	  Write the sample rate and bitrate ladders once in libmp3lame

2026-10-08 13:49  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  doc/html/history.html, test/clients/acm_test.cpp,
	  test/clients/mp3frame.h:
	  Make the ACM codec resample to the rate that smart output picks

2026-10-08 13:49  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h, ACM/acm.rc,
	  ACM/resource.h, doc/html/history.html, test/clients/acm_test.cpp:
	  Let the ACM codec encode stereo input as mono on request

2026-10-08 13:48  aleidinger

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  doc/html/history.html, test/clients/acm_test.cpp,
	  test/clients/mp3frame.h:
	  Make joint stereo the ACM codec's default channel mode

2026-10-08 13:48  aleidinger

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  test/clients/acm_test.cpp:
	  Remove ACM codec settings that have no effect

2026-10-08 13:48  aleidinger

	* ACM/ACMStream.cpp, ACM/AEncodeProperties.cpp,
	  ACM/AEncodeProperties.h, ACM/acm.rc, ACM/resource.h,
	  doc/html/history.html, test/clients/acm_test.cpp,
	  test/clients/mp3frame.h:
	  Add a bit reservoir switch to the ACM codec

2026-10-04 17:14  aleidinger

	* libmp3lame/util.c, test/unit/test_vector_ladder.c:
	  Test that every vector set the build compiles has a known name

2026-10-04 17:13  aleidinger

	* test/unit/test_get_audio_aiff.c:
	  Make the AIFF FORM size test fail when the size check is missing

2026-10-04 17:13  aleidinger

	* doc/Doxyfile-internal.in, test/clients/acm_test.cpp,
	  test/clients/blade_test.cpp, test/clients/ctest.h,
	  test/clients/dshow_test.cpp, test/clients/mp3frame.h:
	  Add the client tests to the internal reference, in plain language

2026-10-04 17:13  aleidinger

	* test/unit/parse_test_stubs.c, test/unit/test_bitstream_headers.c,
	  test/unit/test_calc_sfb_noise_vector.c,
	  test/unit/test_choose_table_neon.c,
	  test/unit/test_choose_table_vector.c,
	  test/unit/test_decode_scale.c, test/unit/test_decoder_api.c,
	  test/unit/test_encode_api.c, test/unit/test_encode_clipgain.c,
	  test/unit/test_encode_mono_interleaved.c,
	  test/unit/test_encode_nonfinite.c, test/unit/test_exit_status.sh,
	  test/unit/test_file_times.c, test/unit/test_free_format.c,
	  test/unit/test_gain_analysis.c, test/unit/test_get_audio_aiff.c,
	  test/unit/test_get_audio_float.c, test/unit/test_get_audio_wav.c,
	  test/unit/test_help_exit.c, test/unit/test_huffman_tables.c,
	  test/unit/test_id3tag.c, test/unit/test_lib_info.c,
	  test/unit/test_merge_argv.c, test/unit/test_parse_id3.c,
	  test/unit/test_parse_pathlen.c, test/unit/test_pkgconfig.sh,
	  test/unit/test_quantize_lines_vector.c,
	  test/unit/test_quantize_pe.c, test/unit/test_resample_decision.c,
	  test/unit/test_rtp_socket.c, test/unit/test_sanity.c,
	  test/unit/test_set_get.c, test/unit/test_stubs.c,
	  test/unit/test_totalframes.c, test/unit/test_vector_ladder.c,
	  test/unit/test_vector_routines.c:
	  Rewrite the unit test documentation in plain language

2026-10-04 17:12  aleidinger

	* Dll/BladeMP3EncDLL.c, dshow/Encoder.cpp, dshow/Encoder.h,
	  dshow/Mpegac.cpp, dshow/PropPage.cpp, dshow/PropPage_adv.cpp,
	  dshow/aboutprp.cpp:
	  Document the DirectShow filter methods in plain language

2026-10-04 17:12  aleidinger

	* ACM/ACM.cpp, ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h:
	  Rewrite the ACM codec documentation in plain language

2026-10-04 17:12  aleidinger

	* frontend/console.h, frontend/get_audio.c, frontend/lame_main.c,
	  frontend/lametime.c, frontend/lametime.h, frontend/main.c,
	  frontend/mp3rtp.c, frontend/mp3x_core.c, frontend/mp3x_core.h,
	  frontend/mp3x_plot.c, frontend/mp3x_plot.h,
	  frontend/mp3x_session.c, frontend/mp3x_session.h,
	  frontend/mp3x_ui.c, frontend/parse.c, frontend/timestatus.c:
	  Rewrite the frontend documentation in plain language

2026-10-04 17:11  aleidinger

	* doc/maintainer-abi.md, doc/maintainer-build-matrix.md,
	  doc/maintainer-check-dist.md, doc/maintainer-coverage.md,
	  doc/maintainer-gen-api-docs.md, doc/maintainer-gen-usage.md,
	  doc/maintainer-perf.md, doc/maintainer-quality.md:
	  Rewrite the maintainer guides in plain language

2026-10-04 17:11  aleidinger

	* doc/doxygen/internal.md, doc/doxygen/maintainer.md,
	  doc/doxygen/mp3x-internals.dox, doc/doxygen/unit_tests.dox:
	  Rewrite the internal overview pages in plain language

2026-10-04 17:11  aleidinger

	* configure.ac, doc/doxygen/internal.md,
	  doc/doxygen/vector-dispatch.dox:
	  Split the vector dispatch page into the current tiers and the
	  measurement record

2026-10-04 17:11  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/fft.c, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/mpglib_interface.c, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/tables.c,
	  libmp3lame/tables.h, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vector/avx2_choose_table.c,
	  libmp3lame/vector/avx2_quantize_lines.c,
	  libmp3lame/vector/avx512_quantize_lines.c,
	  libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/neon_choose_table.c,
	  libmp3lame/vector/xmm_calc_sfb_noise.c,
	  libmp3lame/vector/xmm_choose_table.c,
	  libmp3lame/vector/xmm_quantize_lines.c,
	  libmp3lame/vector/xmm_quantize_sub.c, libmp3lame/version.c:
	  Rewrite the internal library documentation in plain language

2026-10-04 17:10  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/id3tag.c, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/tables.c,
	  libmp3lame/version.c:
	  Rewrite the remaining API documentation in plain language

2026-10-04 17:10  aleidinger

	* libmp3lame/set_get.c:
	  Rewrite the parameter API documentation in plain language

2026-10-04 17:09  aleidinger

	* doc/doxygen/api.dox, doc/doxygen/mainpage.md:
	  Rewrite the API reference overview pages in plain language

2026-10-04 17:09  aleidinger

	* include/lame.h:
	  Rewrite the comments in lame.h in plain language

2026-10-04 17:09  aleidinger

	* ACM/ACM.cpp, ACM/AEncodeProperties.h, ACM/main.cpp,
	  doc/doxygen/internal.md, doc/doxygen/unit_tests.dox,
	  doc/maintainer-abi.md, doc/maintainer-build-matrix.md,
	  doc/maintainer-quality.md, frontend/mp3x_core.h,
	  frontend/mp3x_session.h, frontend/parse.c, libmp3lame/tables.c,
	  test/unit/parse_test_stubs.c,
	  test/unit/test_calc_sfb_noise_vector.c,
	  test/unit/test_encode_api.c, test/unit/test_encode_nonfinite.c,
	  test/unit/test_lib_info.c,
	  test/unit/test_quantize_lines_vector.c, test/unit/test_set_get.c,
	  test/unit/test_stubs.c, test/unit/test_vector_routines.c:
	  Correct wrong statements in the internal documentation

2026-10-04 17:09  aleidinger

	* doc/doxygen/api.dox, doc/doxygen/mainpage.md,
	  doc/doxygen/vector-dispatch.dox, include/lame.h,
	  libmp3lame/id3tag.c, libmp3lame/set_get.c:
	  Correct wrong statements in the API documentation

2026-10-04 10:57  aleidinger

	* doc/html/history.html, frontend/console.c, frontend/console.h,
	  frontend/lame_main.c, frontend/mp3rtp.c, frontend/mp3x_session.c,
	  frontend/parse.c, frontend/parse.h, test/unit/test_help_exit.c,
	  test/unit/test_parse_pathlen.c:
	  Pass the output directory out of parse_args so --nogapout sets it
	  in all cases, and keep terminal control strings whole

2026-10-04 10:57  aleidinger

	* frontend/parse.c, test/unit/test_help_exit.c:
	  Reject an empty argument list

2026-10-04 10:57  aleidinger

	* configure, configure.ac, test/unit/test_pkgconfig.sh:
	  Link the math library only where the link needs it

2026-10-04 10:57  aleidinger

	* configure, configure.ac, test/unit/Makefile.am,
	  test/unit/Makefile.in:
	  Run the frontend's exit status test only when the frontend is
	  built

2026-10-04 10:56  aleidinger

	* test/clients/mp3frame.h, test/unit/test_file_times.c:
	  Move the source's times explicitly in the file time test, and
	  make the client tests' frame helpers inline

2026-10-03 16:37  aleidinger

	* Makefile.MSVC, doc/maintainer-build-matrix.md,
	  maintainer/gen-build-matrix.ps1, maintainer/smoke-clients.ps1,
	  vc_solution/Makefile.am, vc_solution/Makefile.in,
	  vc_solution/README.vs.txt, vc_solution/vs_lame.slnx,
	  vc_solution/vs_lame_acm.vcxproj,
	  vc_solution/vs_lame_acm_test.vcxproj,
	  vc_solution/vs_lame_clients.slnx,
	  vc_solution/vs_lame_dshow.vcxproj,
	  vc_solution/vs_lame_enc_dll.vcxproj,
	  vc_solution/vs_lame_test.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame_files.props,
	  vc_solution/vs_libmp3lame_nodec.vcxproj:
	  Build the Blade DLL, the ACM codec and the DirectShow filter
	  without the decoder

2026-10-03 15:01  aleidinger

	* doc/html/history.html, libmp3lame/gain_analysis.c,
	  test/unit/test_gain_analysis.c:
	  Filter both channels of stereo input in one loop in the
	  ReplayGain analysis

2026-10-03 15:01  aleidinger

	* doc/html/history.html, libmp3lame/gain_analysis.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_gain_analysis.c:
	  Filter mono input once in the ReplayGain analysis

2026-10-03 15:01  aleidinger

	* doc/html/history.html, frontend/mp3rtp.c:
	  Measure only the samples read in mp3rtp's level display

2026-10-03 15:00  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/lame_main.c, frontend/mp3rtp.c,
	  frontend/mp3x_core.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_get_audio_float.c:
	  Hand floating point input to libmp3lame as floats

2026-10-03 15:00  aleidinger

	* include/lame.h:
	  Give the flush buffer size for free format in lame.h

2026-10-03 14:59  aleidinger

	* USAGE, doc/html/detailed.html, doc/html/history.html,
	  frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/set_get.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  test/unit/test_encode_api.c, test/unit/test_set_get.c:
	  Refuse input louder than 4096 times full scale, and report a
	  frame the encoder cannot fit instead of exiting

2026-10-03 14:59  aleidinger

	* doc/html/history.html, dshow/Mpegac.cpp,
	  test/clients/dshow_test.cpp:
	  Pad an aligned DirectShow stream's end with zeros, and refuse an
	  alignment the filter cannot fill

2026-10-03 14:59  aleidinger

	* doc/html/history.html, include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/lame.c, test/unit/test_encode_api.c:
	  Refuse a negative sample count, and answer a missing output
	  buffer as too small

2026-10-03 14:59  aleidinger

	* libmp3lame/encoder.c:
	  Scale the perceptual entropy in CBR and ABR only by a positive
	  smoothed sum

2026-10-03 14:58  aleidinger

	* doc/html/history.html, libmp3lame/lame.c,
	  test/unit/test_encode_api.c:
	  Encode a frame left over from a call whose output did not fit
	  before filling the next

2026-10-03 14:58  aleidinger

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_quantize_pe.c:
	  Clamp a bit count from perceptual entropy before converting it to
	  int

2026-10-03 14:58  aleidinger

	* ACM/AEncodeProperties.cpp, doc/html/history.html,
	  test/clients/acm_test.cpp:
	  Keep the ACM codec's default ABR range when its configuration
	  file holds one it cannot walk

2026-10-03 14:57  aleidinger

	* doc/html/history.html, dshow/Encoder.cpp, dshow/Mpegac.cpp,
	  test/clients/dshow_test.cpp:
	  Fail the DirectShow filter's start when LAME refuses its settings

2026-10-03 14:57  aleidinger

	* ACM/ACMStream.cpp, test/clients/acm_test.cpp:
	  Compute the ACM codec's smart output ratio without overflowing
	  int

2026-10-03 14:57  aleidinger

	* doc/html/history.html, dshow/Encoder.cpp, dshow/Encoder.h,
	  dshow/Mpegac.cpp, test/clients/dshow_test.cpp:
	  Accept any output type in the DirectShow filter when its output
	  rate is 0

2026-10-03 14:56  aleidinger

	* Dll/BladeMP3EncDLL.c, doc/html/history.html,
	  test/clients/blade_test.cpp:
	  Keep lame_enc.dll from using a stream it has already released

2026-10-03 14:56  aleidinger

	* Dll/BladeMP3EncDLL.c, doc/html/history.html,
	  test/clients/blade_test.cpp:
	  Refuse a VBR method lame_enc.dll does not have instead of
	  asserting

2026-10-03 14:56  aleidinger

	* doc/html/history.html, libmp3lame/bitstream.c,
	  test/unit/test_set_get.c:
	  Size the frame buffer as free format only for a free format
	  stream

2026-10-03 14:55  aleidinger

	* ACM/ACMStream.cpp:
	  Keep the ACM codec's debug trace within the output it reports

2026-10-03 14:55  aleidinger

	* ACM/ACM.cpp, doc/html/history.html, test/clients/acm_test.cpp:
	  Refuse to suggest an ACM format for a sample rate the codec
	  cannot encode

2026-10-03 14:55  aleidinger

	* Dll/BladeMP3EncDLL.c, doc/html/history.html,
	  test/clients/blade_test.cpp:
	  Keep lame_enc.dll's output within the buffer it advises

2026-10-03 14:54  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, doc/html/history.html,
	  test/clients/acm_test.cpp:
	  Keep the ACM codec's end-of-stream output within the
	  application's buffer

2026-10-03 14:54  aleidinger

	* libmp3lame/lame.c, test/unit/test_set_get.c:
	  Cap the bitrate a compression ratio asks for before converting it

2026-10-03 14:54  aleidinger

	* libmp3lame/lame.c, libmp3lame/util.c, test/unit/test_set_get.c:
	  Keep the bitrate and lowpass arithmetic within int for extreme
	  settings

2026-10-03 14:53  aleidinger

	* doc/html/history.html, frontend/parse.c,
	  test/unit/test_parse_pathlen.c:
	  Refuse an option value that is not a usable number

2026-10-03 14:53  aleidinger

	* doc/html/history.html, libmp3lame/mpglib_interface.c,
	  test/unit/test_decoder_api.c:
	  Describe the stream on a decode call that needs more input

2026-10-03 14:53  aleidinger

	* doc/html/history.html, libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/set_get.c, test/unit/test_set_get.c:
	  Refuse NaN and infinity in the floating point setters

2026-10-03 14:52  aleidinger

	* doc/html/history.html, libmp3lame/lame.c, libmp3lame/set_get.c,
	  test/unit/test_set_get.c:
	  Refuse a variable bitrate minimum above the maximum

2026-10-03 14:52  aleidinger

	* include/lame.h:
	  Give the encode buffer sizes in lame.h as they hold

2026-10-03 14:52  aleidinger

	* doc/html/history.html, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/util.h, test/unit/test_set_get.c:
	  Refuse to resample to more than 128 times the input sample rate

2026-10-03 14:51  aleidinger

	* libmp3lame/vector/xmm_quantize_sub.c,
	  test/unit/test_quantize_lines_vector.c:
	  Keep the SSE xr^(3/4) finite for coefficients far below
	  audibility

2026-10-03 14:51  aleidinger

	* doc/html/history.html, libmp3lame/set_get.c, libmp3lame/util.c,
	  libmp3lame/util.h, test/unit/test_set_get.c:
	  Keep lame_get_maximum_number_of_samples() within the buffer it is
	  asked about

2026-10-03 14:50  aleidinger

	* frontend/get_audio.c:
	  Copy decoded MP3 samples only when the decoder returned some

2026-10-03 14:50  aleidinger

	* doc/html/history.html, libmp3lame/id3tag.c,
	  test/unit/test_id3tag.c:
	  Write ID3v2.4 frame sizes as synchsafe integers

2026-10-03 14:50  aleidinger

	* libmp3lame/id3tag.c, test/unit/test_id3tag.c:
	  Leave an empty ID3v2 URL frame out of the tag size as well as the
	  tag

2026-10-03 14:49  aleidinger

	* doc/html/history.html, libmp3lame/id3tag.c,
	  test/unit/test_id3tag.c:
	  Refuse an ID3v2 frame identifier shorter than four characters

2026-10-03 14:49  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, configure, doc/Makefile.in, doc/html/Makefile.in,
	  doc/html/history.html, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/id3tag.c,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in, test/Makefile.in, test/unit/Makefile.in,
	  test/unit/test_id3tag.c, vc_solution/Makefile.in:
	  Pad a short ID3v2 comment language instead of reading past it

2026-09-19 07:32  aleidinger

	* USAGE, doc/html/detailed.html, doc/man/lame.1,
	  libmp3lame/gain_analysis.c, libmp3lame/gain_analysis.h,
	  maintainer/gen-usage.py:
	  Replace the dead replaygain.org reference in the ReplayGain
	  documentation
	  
	  The manual page, the HTML documentation and the gain_analysis
	  headers sent
	  readers to www.replaygain.org for what ReplayGain is and how the
	  gain is
	  computed. That domain has served unrelated content since at least
	  2017
	  (Debian bug #859636). The specification is kept on the
	  Hydrogenaudio wiki
	  now, and the page there describes the algorithm libmp3lame
	  implements, down
	  to the filter coefficients.
	  
	  What ReplayGain is, how LAME measures it and where the value goes
	  is said
	  once: in a REPLAYGAIN section of the manual page, and as the
	  introduction
	  of a ReplayGain group in the HTML documentation, each giving the
	  specification's address. The option entries say what each switch
	  changes
	  and point there. gen-usage.py no longer breaks a word longer than
	  the line
	  width, which the address is.

2026-09-17 11:43  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  config.h.in, configure, configure.ac, doc/Makefile.in,
	  doc/html/Makefile.in, doc/html/history.html, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/vector/Makefile.in, libmp3lame/version.c,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in,
	  test/Makefile.in, test/unit/Makefile.in, vc_solution/Makefile.in:
	  Name the working copy's revision in the version string of an
	  alpha or beta build
	  
	  Subversion first, git where the sources are tracked there, none
	  from a tree
	  that is neither; asked at build time.
	  
	  Based on a patch by Elio Blanca, SF patch #85.

2026-09-17 11:43  aleidinger

	* Makefile.am, Makefile.in, configure, configure.ac,
	  doc/html/history.html, lame.spec.in:
	  Drop the RPM spec file; the distributions carry their own
	  
	  SF patch #31.

2026-09-17 11:43  aleidinger

	* doc/html/history.html, dshow/Encoder.cpp, dshow/PropPage.cpp,
	  dshow/PropPage.h, dshow/PropPage_adv.cpp, dshow/PropPage_adv.h,
	  dshow/aboutprp.cpp, dshow/aboutprp.h,
	  test/clients/dshow_test.cpp, test/clients/mp3frame.h,
	  vc_solution/vs_lame_dshow_test.vcxproj:
	  Make the DirectShow filter's keep-all-frequencies switch reach
	  the encoder
	  
	  With the advanced property page's "Keep all frequencies" set, the
	  encoder
	  runs without its lowpass and highpass filters. The DirectShow
	  test sets
	  the switch, has the LAME tag written, and reads off that tag that
	  no
	  lowpass was applied. Two comparisons of unsigned values against
	  zero on
	  the basic page go, and the property pages' message handlers
	  return the
	  type the base class declares.

2026-09-17 11:42  aleidinger

	* frontend/parse.c, libmp3lame/bitstream.c, libmp3lame/id3tag.c:
	  Drop three conditions the code around them already decides

2026-09-12 15:27  aleidinger

	* frontend/brhist.c, frontend/console.h, frontend/lame_main.c,
	  frontend/mp3rtp.c:
	  Mark the frontend's printf-style functions as such, and fix what
	  that flags
	  
	  console.h declares the printf format of console_printf(),
	  error_printf(),
	  report_printf() and the four va_list reporters, so a compiler
	  that checks
	  formats checks these too. That finds two error messages printing
	  a size_t
	  with %d and two calls handing a built string in as the format;
	  the sizes
	  are printed as unsigned long and the strings go through "%s".

2026-09-12 15:26  aleidinger

	* test/clients/dshow_test.cpp:
	  Assert the bitrate the DirectShow test set is the one its stream
	  carries

2026-09-12 15:23  aleidinger

	* ACM/ACM.cpp, ACM/acm.rc, doc/html/history.html,
	  dshow/Property.rc, libmp3lame/VbrTag.c,
	  test/clients/acm_test.cpp, test/unit/test_encode_api.c:
	  Carry the LAME version in the Windows ACM codec and DirectShow
	  filter
	  
	  Both components take their file version from
	  libmp3lame/version.h, and
	  the codec reports the same number as its driver version and in
	  its long
	  name, so there is one version to read off an installed copy.
	  
	  SF bug #416.

2026-09-12 08:19  aleidinger

	* ACM/ACM.cpp, ACM/ACM.h, doc/html/history.html,
	  test/clients/acm_test.cpp:
	  Suggest a well-formed MP3 format from the ACM codec, and name one
	  
	  An out-of-range format index is refused rather than reported as a
	  success.

2026-09-09 19:49  aleidinger

	* doc/html/history.html:
	  doc: sort and group the 4.1 changelog, and even out its wording

2026-09-09 12:05  aleidinger

	* ACM/ADbg/ADbg.cpp, ACM/ADbg/ADbg.h:
	  Spell the AS IS in the ADbg licence with plain double quotes
	  
	  This fixes doxygen output.

2026-09-09 12:05  aleidinger

	* include/lame.h:
	  Document the public enums in the API reference
	  
	  The reference showed one enumerator of one enumeration, because
	  it hides
	  members that nothing documents.
	  The bitrate strategies, channel modes, presets, bit reservoir
	  constraints and
	  status codes are now described where they are declared.

2026-09-09 12:05  aleidinger

	* frontend/get_audio.h:
	  Drop two frontend declarations that never had a definition

2026-09-09 12:05  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/lame_main.c,
	  test/unit/test_get_audio_float.c:
	  Report input samples clipped on read
	  
	  The encoder takes integer samples, so a floating point sample
	  beyond full
	  scale cannot be carried in and the reader clamps it. The
	  conversion happens
	  before any scaling the user asks for, so no later stage can
	  recover it.
	  
	  The reader counts those samples and the frontend reports the
	  count at the end
	  of the encode, so the file can be re-encoded at a lower input
	  level. This is
	  separate from the clipping report beside it, which measures the
	  decoded output
	  at the current gain and needs decoding on the fly.
	  
	  A sample of exactly full scale is clamped but not counted: -1.0
	  becomes the
	  smallest integer either way. -32768/32768 is exactly -1.0, so a
	  16 bit
	  recording touching full scale becomes a floating point file made
	  of such
	  samples, and counting them would report an intact file as
	  clipped.
	  
	  The unit test asserts the count rises for clipped samples and for
	  nothing
	  else, including at that boundary.
	  
	  SF bug #512.

2026-09-09 12:04  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  maintainer/gen-usage.py, make_changelog.sh, misc/lameid3.pl,
	  misc/mk_mp3.sh, misc/mugeco.sh, test/cvscheck.sh,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_get_audio_float.c:
	  Encode floating point input at the level the file stores it
	  
	  The frontend reads floating point audio through two readers,
	  libsndfile
	  and the LAME internal reader.
	  
	  The libsndfile path asked libsndfile to convert the samples to
	  integers, which
	  scales a file by its own peak. A quiet file was encoded at full
	  scale, and the
	  same file encoded differently depending on how LAME was
	  configured.
	  
	  Both readers now take the samples as they are and convert them in
	  one place.
	  Files holding double precision samples take the same path.
	  
	  The unit test pins that conversion in both file reading backends.
	  What it
	  cannot see is a whole file being scaled before it is called, so
	  the harness
	  that compares two builds end to end has the check for that.
	  
	  SF bug #510.

2026-08-27 15:21  aleidinger

	* frontend/parse.c:
	  Restore the null-character constant in the tag conversion loop

2026-08-27 14:10  aleidinger

	* libmp3lame/util.c, libmp3lame/util.h,
	  test/unit/test_choose_table_vector.c,
	  test/unit/test_quantize_lines_vector.c:
	  Ask __builtin_cpu_supports() only about names the compiler knows
	  
	  Whether the builtin exists and whether it knows a given feature
	  name are
	  separate questions, and a name it does not know is a compile
	  error: GCC 5
	  has the builtin without the AVX-512 names, so it cannot build
	  LAME.
	  Reported by Elio Blanca.

2026-08-27 14:10  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in, configure,
	  configure.ac, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.am,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/vector/Makefile.am, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, misc/Makefile.am, misc/Makefile.in,
	  mpglib/Makefile.am, mpglib/Makefile.in, test/Makefile.in,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  vc_solution/Makefile.in:
	  Migrate the deprecated INCLUDES variable to AM_CPPFLAGS

2026-08-27 14:09  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in:
	  Declare the test-driver list where automake can read it
	  
	  Automake reads the list of test-driver extensions before any
	  condition
	  has an answer, and everything in the unit-test directory sits
	  inside
	  one.

2026-08-27 14:09  aleidinger

	* configure.ac, frontend/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/vector/Makefile.in, misc/Makefile.in,
	  test/unit/Makefile.in:
	  Place objects built from another directory's sources under the
	  build tree
	  
	  Two of the unit tests compile a frontend source in directly.
	  Automake
	  puts such an object next to the makefile that builds it unless
	  told
	  otherwise, and warns that a future release will stop doing so;
	  the
	  option it names puts the object in the matching subdirectory of
	  the
	  build tree instead.

2026-08-27 14:09  aleidinger

	* acinclude.m4, config.h.in, configMS.h, configure, configure.ac,
	  mac/Precompile_Common.h:
	  Remove the 80-bit float check nothing asks for
	  
	  The check defines a floating point type and a feature macro that
	  no
	  source file in the tree reads.

2026-08-26 14:41  aleidinger

	* maintainer/gen-build-matrix.ps1, test/Makefile.am,
	  test/Makefile.in, test/clients/blade_test.cpp,
	  vc_solution/Makefile.am, vc_solution/Makefile.in,
	  vc_solution/README.vs.txt,
	  vc_solution/vs_lame_blade_test.vcxproj,
	  vc_solution/vs_lame_clients.slnx:
	  Exercise the Blade encoder DLL from the test harnesses
	  
	  The DLL was built and smoke-tested for its exports, and then
	  never run:
	  unlike the ACM codec and the DirectShow filter, no component test
	  drove
	  it through its own interface.
	  
	  lame_blade_test loads the built lame_enc.dll and makes the calls
	  a Blade
	  host makes - encoding from both sample layouts, draining the
	  stream, and
	  rewriting the reserved tag frame over a bare file and over an
	  ID3v2 tag -
	  then walks the frames that come out against the configured rate.
	  The
	  build matrix gains one cell per architecture that builds the DLL
	  and its
	  test together and requires both, the 64-bit build included.

2026-08-26 14:40  aleidinger

	* Dll/BladeMP3EncDLL.c, vc_solution/README.vs.txt,
	  vc_solution/vs_lame.slnx, vc_solution/vs_lame_enc_dll.vcxproj:
	  Build the Blade encoder DLL for x64 as well
	  
	  Its project file only ever carried 32-bit configurations, and
	  nothing in
	  the interface asks for that: the entry points are cdecl and
	  exported
	  undecorated, and the configuration structure has the same layout
	  and size
	  on both architectures. The static library it links has built for
	  x64 since
	  the solution was converted.
	  
	  The 64-bit build turns up three narrowing conversions on the way
	  into
	  fseek(), whose offset is a long on Windows whatever the word
	  size.

2026-08-26 14:40  aleidinger

	* DEFINES, libmp3lame/newmdct.c:
	  Remove a build switch that cannot be switched on
	  
	  A conditional around the polyphase analysis window suppressed
	  that table
	  when a macro was defined, to make room for a substitute the same
	  commit
	  expected. The substitute never arrived: it was to be
	  GOGO-no-coda's
	  assembler subband filter, and the only assembler that ever came
	  from there
	  was the FFT, since replaced by intrinsics.
	  
	  Nothing else in the library defines the table, so defining the
	  macro
	  deleted it and the file stopped compiling. The list of build
	  definitions
	  described the switch with a question mark, and that entry goes
	  with it.

2026-08-26 14:39  aleidinger

	* TODO:
	  Retire six entries from the list of things to do
	  
	  Six of the nineteen have been overtaken, some of them long ago.
	  
	  The Visual Studio solutions build every target the project-files
	  entry asks
	  for, and the compiler versions it names have not shipped this
	  century.
	  
	  Standard input works in a build configured for libsndfile, which
	  is not asked
	  to read it - the reader that already handles a pipe does.
	  
	  The decoder whose intensity-stereo handling one entry reports
	  against is no
	  longer part of this tree, so the report has nowhere to go here.
	  
	  The threshold of hearing is evaluated over six partitions of the
	  topmost
	  scalefactor band, where the entry hoped for two or three, and
	  over the
	  short-block band beside it as well.
	  
	  The assembler entry names a project whose last release is twenty
	  years old,
	  and the work that answered the question it was asking went to
	  intrinsics
	  instead, on four instruction sets with a run-time choice between
	  them.
	  
	  The last entry asks for an extension named nowhere else in the
	  source, the
	  change log or the history, and a name on its own cannot be acted
	  on.

2026-08-26 14:39  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/TODO:
	  Read the ACM codec's rate tables from the library
	  
	  The codec carried its own copy of the standard's bitrate and
	  sampling-frequency tables, and its own routine for picking the
	  nearest
	  valid sample rate - a duplicate of a library function, with a
	  note beside
	  it asking for the library's. libmp3lame publishes both, so the
	  codec asks
	  for them instead of restating them.
	  
	  The order the codec presents is unchanged, because it is not
	  presentation:
	  a host asks for a format by number and that number is a subscript
	  into
	  these tables.

2026-08-26 14:39  aleidinger

	* Dll/BladeMP3EncDLL.c:
	  Harden the Blade encoder DLL's debug log and tag skip
	  
	  The debug helper wrote into a fixed buffer with a call that
	  leaves no
	  terminator when the text does not fit, and rewrote the last three
	  bytes of
	  the module path at an offset computed from a length that is zero
	  when the
	  path lookup fails, putting the stores far outside the buffer. The
	  ID3v2
	  check read ten bytes from the file and compared them as a string.

2026-08-26 14:39  aleidinger

	* maintainer/smoke-clients.ps1, vc_solution/README.vs.txt,
	  vc_solution/vs_lame.slnx:
	  Build the Blade encoder DLL from the Visual Studio solution
	  
	  The project file for lame_enc.dll shipped in the distribution but
	  belonged to
	  no solution, so opening either one gave no way to build the
	  component and the
	  project list in README.vs.txt did not name it. It stayed outside
	  the warning
	  level and the build matrix that cover the rest of the tree.
	  
	  The target is 32-bit, so it is selected for Win32 and left
	  unbuilt in the
	  solution's x64 configurations, as mp3rtp already is.

2026-08-26 14:39  aleidinger

	* ACM/ACM.cpp, ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h,
	  ACM/DecodeStream.cpp, doc/doxygen/unit_tests.dox:
	  Build the internal documentation without warnings
	  
	  The internal Doxygen configuration reported nine warnings. None
	  of them
	  stopped anything: the reference is generated on request, and a
	  page that
	  comes out wrong costs nothing at build time.
	  
	  Three blocks of disabled code in the ACM codec are commented out
	  with
	  /* ... */ and each holds another /* inside it. A compiler ends
	  such a
	  block at the first */ and is content; Doxygen counts the inner
	  one and
	  then runs off the end of the file looking for a close that is not
	  there.
	  Those blocks now sit behind #if 0, which says what is meant of
	  disabled
	  code and does not depend on how a scanner counts comment markers.
	  The
	  conversion has to close the inner comment properly, because
	  inside #if 0
	  it is a comment again rather than text within one. A fourth block
	  of the
	  same shape goes with them.
	  
	  The rest is documentation that no longer describes what it sits
	  on: two
	  parameter lists that name one argument and leave the other, a
	  parameter
	  whose name was taken from the first word of its own description,
	  blocks
	  left behind on declarations that are commented out, whose
	  parameter lists
	  were then carried down to the next function that still exists,
	  and an
	  angle-bracketed placeholder in prose that was read as markup.
	  
	  Rendering the two document sets before and after gives the same
	  pages and
	  the same entries, so this is the warnings and nothing else.

2026-08-26 14:38  aleidinger

	* doc/html/history.html, dshow/Mpegac.cpp,
	  test/clients/dshow_test.cpp:
	  Answer the DirectShow encoder's capability list without crashing
	  
	  IAMStreamConfig::GetStreamCaps() describes what the encoder can
	  produce, and
	  it built each answer by copying the output pin's own media type
	  and writing
	  an MP3 format into the copy. That type carries no format block
	  before the pin
	  connects, which is the state an application enumerating formats
	  finds it in,
	  and none afterwards either when the pin has connected as a stream
	  - which is
	  what a file writer downstream asks for. In both cases the write
	  went through
	  a null pointer and the application taking inventory of the
	  encoder crashed.
	  
	  The entry is now built from the capability table, so what the pin
	  happens to
	  be connected as no longer decides whether the call works.
	  
	  The index one past the last entry was accepted as well. It read
	  the zeroed
	  slot after the table and divided by its 0 Hz sample rate; it is
	  refused now,
	  as is the same index on the side that fills the table in, where
	  the bound was
	  tested after the store rather than before it.
	  
	  The filter-graph test asks for the capability list in both pin
	  states, and
	  for the boundary indices. Reported by an anonymous user, bug
	  ticket #424.

2026-08-26 14:38  aleidinger

	* USAGE, doc/html/detailed.html, doc/html/history.html:
	  Describe the ID3 tag options in the settings reference
	  
	  The nine tag options had a row apiece in the reference's index
	  and nothing
	  in the body for the row to link to, as did the two options that
	  choose how
	  tag text is encoded - the third of those three was described and
	  they were
	  not. A twelfth link named the decoding sense of -t, whose entry
	  exists under
	  another name.
	  
	  Each now has an entry. What they say is what the built-in help
	  and the
	  manual page already said, so the three accounts agree; --tv also
	  gains what
	  neither of them mentions, that setting the same frame twice
	  replaces it
	  rather than adding a second one.
	  
	  USAGE is generated from this file, so a link with nothing behind
	  it came out
	  there as a heading with nothing under it. There are no such links
	  left.

2026-08-26 14:38  aleidinger

	* doc/html/history.html, doc/man/lame.1:
	  Describe the options the manual page left out
	  
	  Fourteen of the seventy-five long options the parser accepts had
	  no entry in
	  lame(1), --nogaptags and --flush among them - the first of those
	  is the
	  complaint the bug report about this documentation opens with.
	  Each now has
	  one, in the section it belongs to rather than appended at the
	  end, and each
	  says what the built-in help and the HTML reference already said
	  about it, so
	  the three accounts agree.
	  
	  The page renders with no warnings from groff, and every long
	  option is now
	  the subject of an entry rather than merely mentioned somewhere in
	  the prose,
	  which is the weaker thing a search for the name would have shown.

2026-08-26 14:37  aleidinger

	* Makefile.am, Makefile.in, USAGE, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/doxygen/maintainer.md,
	  doc/html/detailed.html, doc/html/history.html,
	  doc/html/usage.html, doc/maintainer-gen-usage.md,
	  maintainer/gen-usage.py:
	  Generate USAGE from the documentation
	  
	  USAGE is the plain-text form of the documentation, and it was
	  kept beside
	  the HTML by hand. It had drifted: it named 61 of the 75 long
	  options the
	  parser accepts - two of the missing ones were added in the round
	  that found
	  this - and misspelled --resample in a sentence explaining how to
	  switch
	  resampling off, which had been reported and left for years.
	  
	  It is now written by a script from doc/html/usage.html and
	  doc/html/detailed.html, both of which have to be edited for a new
	  option or
	  a new example anyway. The file says where it comes from.
	  
	  The examples that used to sit at the top of it now sit on the
	  Usage page,
	  which is where the rest of the introduction is, with two more
	  beside them:
	  raw PCM through a pipe, and filtering explicitly at a low
	  bitrate. The list
	  of the scripts in misc/ moves there too, and gains the one that
	  was never
	  mentioned.
	  
	  Every example in the documentation has now been run, which none
	  of them ever
	  had been. Four did not work as printed: three of the --cbr
	  examples were
	  written without the file names, and the fifth was the synopsis
	  rather than a
	  command. What each of those three says the encoder will do was
	  measured, and
	  is right.
	  
	  Generating the text also made two gaps in the HTML visible:
	  --id3v2-utf8 and
	  --ignorelength were described nowhere, and three links from the
	  index table
	  pointed at anchors the body does not define.

2026-08-26 14:37  aleidinger

	* doc/html/history.html, frontend/parse.c:
	  Say so when a tag value cannot be converted
	  
	  Tag text given on the command line is converted from the locale's
	  character
	  set into the one the tag needs. That conversion stops at the
	  first character
	  it cannot produce, and its result was discarded, so a value the
	  locale
	  cannot express was written cut off at that point and nothing was
	  said. On a
	  machine whose locale is plain ASCII - which is what a process
	  gets when the
	  requested locale is not installed - that is every accented
	  character.
	  
	  What could be converted is still written, since a transliterating
	  conversion
	  is what was asked for, and the value that lost part of itself is
	  now named
	  along with the encoding it could not be written in. A run that
	  converts
	  cleanly says nothing, and --silent covers these messages as it
	  covers the
	  other tag ones.
	  
	  The three converters had the same body apart from two encoding
	  names and, in
	  one of them, a byte-order mark to leave room for, so they are now
	  one. Three
	  copies were three places for the answer to go unread.

2026-08-26 14:36  aleidinger

	* doc/html/history.html, dshow/Mpegac.cpp,
	  test/clients/dshow_test.cpp:
	  Query the DirectShow output pin's own parent class
	  
	  The encoder's output pin is built on CTransformOutputPin, and its
	  interface
	  query handed anything it did not recognise to CBaseOutputPin
	  instead - the
	  class below that. What CTransformOutputPin answers for itself is
	  seeking:
	  it passes IMediaSeeking and IMediaPosition back to whatever feeds
	  the
	  filter's input, so that an application can ask the graph how long
	  the stream
	  is and where it is. Skipping it, the pin answered E_NOINTERFACE
	  and a graph
	  containing the filter could not be seeked.
	  
	  The filter object's own query already goes to its direct parent;
	  this makes
	  the pin agree with it.
	  
	  The filter-graph test gains the arm that distinguishes the two:
	  with the
	  input pin connected, the output pin is asked for both interfaces
	  and for the
	  duration of the stream. Asking for the duration as well is what
	  separates a
	  pin that answers the question from one that also knows the
	  answer.

2026-08-26 14:36  aleidinger

	* doc/html/history.html, frontend/timestatus.c:
	  Fit the progress heading to the frame counter under it
	  
	  The heading above the encoding progress display reserved six
	  digits for the
	  frame counter. A file long enough to need more than that pushed
	  every column
	  of the row out from under the heading it belongs to, which is
	  what the
	  counter is read against.
	  
	  The width is taken from the number of frames the file will
	  produce, decided
	  once when the heading is printed and used for every row after it,
	  so the two
	  cannot drift apart. It is never narrower than the six digits it
	  always was.
	  
	  The decoding display is left as it is: its line has no heading
	  above it, so
	  there is nothing for it to line up with.

2026-08-26 14:36  aleidinger

	* doc/html/Makefile.am, doc/html/Makefile.in,
	  doc/html/history.html:
	  Install the HTML documentation where configure was told to
	  
	  doc/html assigned to docdir, which is one of the standard
	  directory
	  variables configure already sets and --docdir already controls.
	  Assigning
	  to it discarded whatever the user had asked for, so the pages
	  always landed
	  under the same place no matter what was requested.
	  
	  The pages keep their own subdirectory below the documentation
	  directory,
	  and it is now named for this package rather than reusing htmldir,
	  which is
	  another variable configure owns and which already means the
	  directory the
	  HTML documentation goes in.
	  
	  The default location does not move.

2026-08-26 14:36  aleidinger

	* config.h.in, configMS.h, configure, configure.ac,
	  doc/html/history.html, libmp3lame/util.c:
	  Arm the debug build's floating point exceptions where the
	  platform can
	  
	  --enable-debug arranges for floating point exceptions to raise a
	  signal
	  rather than pass silently. On Unix that was compiled wherever the
	  operating
	  system is Linux, which is two guesses at once: it reaches for a
	  header that
	  belongs to GNU libc, and it uses the mask names of the x87 unit.
	  Android and
	  musl are Linux and have neither, so the build stopped there; a
	  Linux machine
	  with no x87 stops on the second half for its own reason.
	  
	  Configure now asks for what the code uses. The first choice is
	  feenableexcept(), which needs no architecture names and covers
	  the systems
	  the x87 path never reached - the BSDs, and Linux on anything
	  without an x87.
	  Where it is absent the x87 path remains, and where the header
	  arrives without
	  the mask names, neither is compiled and a debug build does
	  without the aid.
	  
	  Whether an armed exception then traps at all is the hardware's
	  business: an
	  AArch64 implementation need not offer trapping exceptions, and
	  there the call
	  succeeds and execution carries on.
	  
	  Nothing here is compiled into a normal build.

2026-08-26 14:36  aleidinger

	* doc/html/history.html, libmp3lame/mpglib_interface.c,
	  libmp3lame/util.h, test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_decode_scale.c:
	  Restore clipping detection and accurate ReplayGain
	  
	  Both --clipdetect and --replaygain-accurate work by decoding the
	  stream as
	  it is encoded and measuring the result. That measurement reads
	  the decoded
	  samples in the encoder's own scale, where full scale is 32768;
	  the decoding
	  library hands them back normalised to one. The two have disagreed
	  since the
	  decoder was replaced, by a factor of 32768, and neither option
	  has worked
	  since: --clipdetect could not report clipping under any
	  circumstances, and
	  --replaygain-accurate returned the same figure for every input -
	  a figure
	  so far outside the range the tag can express that LAME warned
	  about it.
	  --replaygain-fast, the default, analyses the input rather than
	  the decoded
	  output and was never affected.
	  
	  The samples are brought into the encoder's scale where the
	  decoder's output
	  crosses into it, rather than at each of the places that read
	  them, so the
	  function's declaration and its callers agree again and a future
	  caller gets
	  what the declaration promises. It now says what that scale is.
	  
	  The encoded audio does not move. What changes is the ReplayGain
	  information
	  in the LAME tag, and only when one of the two options is given.
	  
	  A unit test covers the scale, which nothing did before - it is
	  not visible
	  in any signature, which is why the change went unnoticed. The
	  test asks the
	  question without naming the factor: the same stream decoded
	  through the
	  clipped path, whose scale its type fixes, must give the same peak
	  as the
	  unclipped one.

2026-08-26 14:35  aleidinger

	* doc/html/detailed.html, doc/html/history.html, doc/man/lame.1,
	  frontend/lame_main.c, frontend/main.h, frontend/parse.c:
	  Optionally write ReplayGain to the ID3v2 tag as well
	  
	  LAME computes the ReplayGain figures and stores them in its own
	  tag. The
	  new --replaygain-id3v2 writes them additionally as the two
	  user-defined
	  ID3v2 text frames REPLAYGAIN_TRACK_GAIN and
	  REPLAYGAIN_TRACK_PEAK. It is
	  off by default.
	  
	  The figures are not known until the audio has been encoded, and
	  the tag is
	  written before it, so the frames are reserved with placeholders
	  and
	  overwritten in place afterwards - the same way the LAME tag frame
	  is
	  filled in. The values take the form the ReplayGain specification
	  gives,
	  which is not a fixed width, so the reservation asks for a few
	  bytes of
	  ID3v2 padding that the finished frames give back. That the
	  rewritten tag
	  really is the length of the one already in the file is checked
	  before
	  anything is written, and a mismatch leaves the file alone rather
	  than
	  overwriting the audio behind the tag.
	  
	  The peak is a by-product of decoding while encoding, so its frame
	  is
	  written only where something measured one. Three situations
	  produce no
	  frames at all, each saying so rather than leaving a placeholder
	  behind: an
	  output that cannot be written to twice, an input whose own tag is
	  passed
	  through unchanged, and ReplayGain analysis switched off.

2026-08-26 14:32  aleidinger

	* config.h.in, configMS.h, configure, configure.ac,
	  doc/html/detailed.html, doc/html/history.html, doc/man/lame.1,
	  frontend/lame_main.c, frontend/lametime.c, frontend/lametime.h,
	  frontend/main.h, frontend/parse.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_file_times.c:
	  Optionally give the output file the date and time of the input
	  file
	  
	  --preserve-modtime copies the input file's modification and
	  access times
	  onto the output, when encoding and when decoding alike. It is off
	  by
	  default. Without it every file a re-encoded collection produces
	  carries
	  the moment of the run, and a library that was ordered by date no
	  longer
	  is.
	  
	  The times are read before the input is opened, because reading a
	  file
	  updates its access time: taken at the end, what would be
	  preserved is
	  the moment of the encode rather than anything the input had.
	  Creation
	  time is left alone - there is no portable way to set one.
	  
	  Standard input and standard output have no file to take the times
	  from
	  or to give them to, so a pipe on either side is passed over
	  silently. A
	  named file that could not be stamped is reported.

2026-08-23 08:45  aleidinger

	* doc/html/history.html, frontend/brhist.c, frontend/brhist.h,
	  frontend/lame_main.c:
	  Show the bitrate the encoded audio actually has
	  
	  While encoding, the figure under the histogram was the average of
	  the rates
	  the individual frames were coded at, weighted by how many frames
	  used each.
	  That is not the bitrate of the file: the frames differ in size
	  from their
	  nominal rate, and the difference, while small, is exactly what
	  the display is
	  asked about.
	  
	  The frontend writes every one of those bytes and can divide them
	  by their
	  playing time. The tag frame is left out of the total, since it
	  carries
	  metadata rather than audio and is not among the frames counted
	  alongside it -
	  with it in, a constant excess would be reported, largest on the
	  shortest
	  files. Constant bitrate encodes therefore still display their
	  nominal rate.
	  
	  Reported by Elio Blanca, patch ticket #65. The patch there
	  reaches the same
	  figure through a new library entry point over an internal
	  counter; the counter
	  has not yet seen the tag frame or the flush when the display is
	  drawn, and the
	  number the frontend already holds needs no addition to the
	  library's
	  interface.

2026-08-23 08:45  aleidinger

	* configure, configure.ac, doc/html/history.html,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in:
	  Build where the toolchain cannot restrict a library's exported
	  symbols
	  
	  libmp3lame asks libtool to link the shared library against an
	  export list, so
	  that only its documented interface leaves the object. libtool
	  does that with
	  whatever the host provides, and a toolchain that provides nothing
	  of the kind
	  could not build the library at all - reported for tcc, which
	  links executables
	  itself and rejects the option outright.
	  
	  configure now asks libtool whether such a link works and leaves
	  the export
	  list out where it does not, saying so on the console and in its
	  summary,
	  because a library built that way exports everything it defines.
	  The question
	  goes to libtool rather than to the compiler, since the mechanism
	  differs
	  between hosts and testing for any one of them would discard the
	  export list
	  where it works perfectly by another.
	  
	  Reported by Elio Blanca on the lame-dev list.

2026-08-23 08:44  aleidinger

	* config.h.in, configure, configure.ac, doc/html/history.html,
	  frontend/main.c, frontend/main.h, frontend/parse.c:
	  Build the command line tools where the C library has no strnlen
	  
	  The frontend calls strnlen in six places and nothing tested for
	  it, so a
	  system predating C11 and POSIX.1-2008 could not link the tools -
	  reported on
	  Mac OS X 10.5.8. configure now asks for the function, and the
	  frontend
	  supplies its own where the answer is no.
	  
	  The replacement is a bounded scan rather than a call to strlen.
	  Each of these
	  call sites reads a buffer that need not contain a terminator at
	  all, which is
	  why they were written bounded, and substituting strlen would put
	  back the
	  unbounded read they exist to avoid.
	  
	  Reported by Robert Kausch on the lame-dev list.

2026-08-23 08:43  aleidinger

	* configure, configure.ac:
	  Reject a compiler flag the toolchain cannot link with
	  
	  The optimisation and security hardening flags configure adds were
	  chosen by
	  compiling a trivial program with each one. Several of them make
	  the compiler
	  emit references to a support library, so a toolchain can accept
	  the flag,
	  pass the probe, and then fail to link anything built with it -
	  reported on
	  Haiku, where the stack protector needs a library the system does
	  not ship.
	  The probe now links, which is the question it was meant to ask.
	  
	  The flags chosen on a working toolchain do not change; the
	  linker-flag probe
	  beside it already worked this way.
	  
	  Reported by Robert Kausch on the lame-dev list.

2026-08-22 11:57  aleidinger

	* libmp3lame/VbrTag.c:
	  Correct the Xing/Info header layout comment
	  
	  The block describing the tag listed the table of contents before
	  the frame
	  count and the stream size, put the LAME tag at 20 bytes, and
	  totalled to a
	  figure matching neither of the two defines it sat above. The
	  writer emits
	  those fields the other way round: the six it counts are
	  VBRHEADERSIZE's 120
	  bytes, and the 36-byte LAME tag that follows brings
	  LAMEHEADERSIZE to 156.
	  Each define now carries its own sum.
	  
	  Reported by Vincent HO-SUNE on SourceForge (patch #100). The same
	  ticket also
	  sets a further bit in the header flags word. That part is not
	  taken here.
	  Nothing is known to read it - the LAME tag is found by its
	  version string,
	  here and everywhere else - while setting it would alter the
	  header of every
	  file the encoder writes, on Xing headers as well as on the Info
	  headers it is
	  named for.

2026-08-22 11:57  aleidinger

	* TODO, doc/html/history.html, libmp3lame/util.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_resample_decision.c:
	  Resample whenever the input sample rate differs from the output
	  rate
	  
	  Input and output sample rates that agreed to about four digits
	  were treated
	  as equal and the audio was passed through unresampled. That was a
	  guard
	  rather than a policy: a resampling ratio very close to one used
	  to be rounded
	  to an exact integer, which chose the wrong filter length and let
	  the window
	  index run past the precomputed filter table. The rounding was
	  corrected in
	  2015; sample rates are integers, so equality is the whole
	  question. The TODO
	  entry that recorded the workaround goes with it, and that list is
	  a bullet
	  list now, so retiring an entry no longer renumbers the rest.
	  
	  A unit test states the decision over rates that straddle the
	  width of the old
	  guard, including the two that sat on its edges, and encodes
	  across a
	  one-hertz mismatch end to end, so the path that used to fault is
	  exercised on
	  every run.
	  
	  Reported, with a patch, by Riley Gabriel on SourceForge (patch
	  #101).

2026-08-22 11:56  aleidinger

	* doc/html/history.html, frontend/lame_main.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_exit_status.sh:
	  Report a failed read through the exit status
	  
	  When an input file cannot be read to the end, the frontend's
	  readers say so,
	  but the encoding and decoding loops both ended on the same
	  condition as a
	  clean end of input. A truncated output file was written, nothing
	  was printed,
	  and the process reported success, so a caller in a script had no
	  way to tell a
	  finished encode from an abandoned one.
	  
	  Both loops now stop where the read failed, name the failure, and
	  exit
	  non-zero - the status a missing input file has always produced.
	  Output for
	  input that reads to the end is unchanged, byte for byte.
	  
	  A unit test drives the built frontend over a deliberately damaged
	  file on both
	  paths, with a clean input and a missing one beside it, so it can
	  pass neither
	  on a build where everything fails nor on one where nothing does.
	  
	  Reported, with a patch for the encoding case, by Van de Bugger on
	  SourceForge
	  (bug #513).

2026-08-21 12:53  aleidinger

	* libmp3lame/encoder.c, libmp3lame/fft.c, libmp3lame/util.c:
	  Document the absolute threshold of hearing and the analysis
	  windows
	  
	  Describes the threshold formula and the curve types built on it,
	  the Bark
	  conversion, and the adjustment that scales the threshold to the
	  loudness of
	  the granule - including why that adjustment responds
	  asymmetrically, and that
	  switching it off is the only way to encode against a threshold
	  making no
	  assumption about playback level.
	  
	  The long-block analysis window is recorded as a to-do rather than
	  a fact: the
	  tree says the choice makes no real difference and does not appear
	  to have
	  measured it, and the two candidates differ in exactly the
	  property the
	  tonality estimate depends on.

2026-08-21 10:04  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in, configure,
	  configure.ac, doc/Makefile.in, doc/html/Makefile.in,
	  doc/html/history.html, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/lame.pc.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in, test/Makefile.in, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_pkgconfig.sh,
	  vc_solution/Makefile.in:
	  Declare libmp3lame's private link dependencies in lame.pc
	  
	  Linking a program against a static libmp3lame through pkg-config
	  yields
	  only -lmp3lame. lame.pc names no private dependencies, so the
	  math
	  library is unresolved in every build, and libmpg123 as well
	  wherever the
	  decoder is configured in.
	  
	  Name them from the same configure decisions the library is built
	  with.
	  libmpg123 goes in Requires.private, which also brings its search
	  path and
	  its own private libraries; the math library has no pkg-config
	  file of its
	  own, so it goes in Libs.private, as whichever library configure
	  selected.
	  
	  Only what libmp3lame itself links is named. The command line
	  tools use
	  several further libraries, and none of them belongs in a file
	  that
	  describes the library.
	  
	  A unit test compares the file with libtool's record of the link
	  in both
	  directions, so it can neither fall behind what the library needs
	  nor
	  offer a consumer more than it does.
	  
	  Reported with a patch by Maarten on SourceForge (bug #526).

2026-08-21 10:03  aleidinger

	* libmp3lame/psymodel.c, libmp3lame/psymodel.h:
	  Document the psychoacoustic model
	  
	  The model carried no reference documentation, and the account in
	  its file
	  header no longer matched the encoder: it described two tonality
	  estimators
	  where there is one, an optional additive-masking mode that is not
	  optional,
	  and an unpredictability term that no longer exists.
	  
	  That header is replaced with a description of what the model
	  computes and in
	  what order, and every function and tuning constant gains one. Two
	  constants
	  are described as reachable only from disabled code, because that
	  is what they
	  are; the pre-echo control they belong to now runs after the
	  mapping to
	  scalefactor bands, where the position of an attack inside the
	  granule is
	  known.

2026-08-16 17:48  aleidinger

	* frontend/rtp.c:
	  Declare the RTP header's bit-fields unsigned
	  
	  The RTP version is 2 and the header's version field is two bits
	  wide, but the
	  field was declared signed, so the value does not fit and a
	  compiler that
	  rejects a truncation which changes a constant stops the build
	  outright. The
	  sequence number has the same shape: sixteen bits, seeded from a
	  random value
	  that uses all of them, and read back negative ever since.
	  
	  Nothing about the packets changes. The header goes on the wire by
	  casting the
	  structure to four words and putting each into network byte order,
	  so what is
	  sent is the bit pattern rather than the value, and that pattern
	  is the same
	  under either declaration - as the unit test covering the emitted
	  packets
	  asserts, version and payload type and an advancing sequence
	  number included.

2026-08-16 12:01  aleidinger

	* vc_solution/Makefile.am, vc_solution/Makefile.in:
	  Ship the shared Visual Studio property sheet in the distribution
	  
	  The property sheet holding the client components' common compiler
	  settings
	  was not listed among the files a distribution tarball is built
	  from, so a
	  release did not carry it. All five client projects import it
	  without a
	  condition - the ACM codec and the DirectShow filter as well as
	  their tests -
	  and an import naming a file that is not there stops the project
	  from loading
	  rather than falling back on a default, so the client solution
	  could not be
	  opened at all from an unpacked release. Working copies were
	  unaffected.

2026-08-16 10:36  aleidinger

	* ACM/AEncodeProperties.cpp, test/clients/acm_test.cpp:
	  Give the ACM codec a save that works without a file to start from
	  
	  Loading a configuration and saving one were not the same shape.
	  The load
	  side is a public method the driver calls when it is created. The
	  save side
	  was a method with an empty body next to a writer that required
	  the
	  configuration file to already exist and to already be well
	  formed: it read
	  the file, and returned having written nothing when that read
	  failed.
	  
	  The installer lays a configuration file down beside the codec, so
	  this
	  normally holds. When it does not - the file removed, emptied or
	  damaged -
	  the settings dialog appeared to work and nothing was kept, with
	  nothing said
	  about it, and the next load went back to the defaults.
	  
	  The writer now builds the elements it needs when they are not
	  there, so a
	  save succeeds against a missing or empty file, and the empty
	  method is the
	  counterpart its name promised: it writes the parameters the
	  object holds to
	  the configuration it was loaded from. The dialog saves through
	  it.
	  
	  Loading still writes nothing when there is no file to read. The
	  defaults
	  stand until something asks for a save, rather than every instance
	  of a driver
	  writing into the directory the codec was installed into, for a
	  change the
	  user did not make.
	  
	  Also here, in the same function: a configuration entry carrying
	  no name was
	  asked to compare its name, and the file is edited by hand often
	  enough for
	  that to matter.

2026-08-16 10:35  aleidinger

	* ACM/AEncodeProperties.cpp, test/clients/acm_test.cpp,
	  test/clients/ctest.h, test/clients/dshow_test.cpp:
	  Check the shape of the ACM's configuration file before reading it
	  
	  The codec looks up a root element and a child of it in the file
	  it keeps
	  beside itself, and then uses both without asking whether either
	  was found.
	  A file that parses as XML but is shaped some other way makes one
	  of them a
	  null pointer. Five of the six places that read the file did this;
	  the sixth
	  already checked, and is what the other five now look like.
	  
	  The file is hand-editable and is exactly what a failed write
	  truncates, so
	  this is not a malformed-input case so much as an ordinary one. It
	  matters
	  more than the same mistake elsewhere would: this class runs
	  inside a driver
	  that Windows has loaded into some other application's process, so
	  following
	  a null pointer takes that application down rather than the codec.
	  
	  A file that cannot be read at all already leaves the defaults in
	  place, and
	  that is now what a file that reads but is not ours does too.
	  
	  The tests cover three shapes of it, and check the values
	  afterwards rather
	  than only that nothing crashed - a build that stopped opening the
	  file would
	  otherwise pass. Their output is now unbuffered, because a test
	  that can take
	  its own process down loses a buffered transcript exactly when it
	  is needed.

2026-08-16 10:35  aleidinger

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h:
	  Drop the ACM configuration class's undefined save entry point
	  
	  The class declared a method for saving parameters out of the
	  config dialog
	  that has no definition anywhere. What looks like one is inside a
	  block
	  comment, because the line meant to close that comment has a space
	  in the
	  middle of it, and its only call site was commented out as well -
	  which is why
	  the codec still linked.
	  
	  The dialog already saves through the method it calls on OK, so
	  nothing was
	  waiting on this. What it cost was a reader's time: it is among
	  the first
	  things found when looking for where a configuration is written.

2026-08-16 10:34  aleidinger

	* maintainer/gen-build-matrix.ps1, vc_solution/README.vs.txt:
	  Run the client component tests from the build matrix
	  
	  The two cells that build the Windows client components stopped at
	  the smoke
	  test, which reports that each DLL loads and exports what it
	  should. They now
	  go on to run the tests that drive each component through the
	  framework it
	  exists for, so a cell answers what the component does and not
	  only that it
	  was produced.
	  
	  The filter's cell builds one project rather than the solution, so
	  the program
	  that exercises it is named there too; without that the cell would
	  look for a
	  binary nothing in it built. The codec's cell runs the filter's
	  test as well,
	  where it reports a skip, since that cell does not build the
	  filter.
	  
	  Every step but the last now carries its own status check. Only
	  the last
	  command's code reaches the caller, so a failure earlier in the
	  sequence was
	  previously reported by whatever ran after it.

2026-08-16 10:34  aleidinger

	* test/Makefile.am, test/Makefile.in, test/clients/dshow_test.cpp,
	  vc_solution/Makefile.am, vc_solution/Makefile.in,
	  vc_solution/vs_lame_clients.slnx,
	  vc_solution/vs_lame_dshow_test.vcxproj:
	  Test the DirectShow filter inside a real filter graph
	  
	  The filter had no test of what it does once loaded. This adds a
	  console
	  program that takes the built binary and obtains the filter from
	  the DLL's own
	  class factory - the only thing a registration adds is the CLSID
	  lookup - then
	  lets the graph manager do the rest: insert the parser the source
	  needs,
	  negotiate a media type across both pin connections, and stream a
	  WAV file
	  through the encoder into a file.
	  
	  What comes out is read back as MPEG frames rather than counted as
	  bytes, and
	  the filter's own property interface is exercised once its input
	  pin is
	  connected, which is where its header says settings hold rather
	  than being
	  overridden by the media type's defaults.
	  
	  The filter is built only where the DirectShow base class sources
	  are laid
	  out, so the test skips when there is nothing to load, and takes a
	  switch that
	  turns the same absence into a failure for a caller that expects
	  one.

2026-08-16 10:33  aleidinger

	* test/Makefile.am, test/Makefile.in, test/clients/acm_test.cpp,
	  test/clients/mp3frame.h, vc_solution/vs_lame_acm_test.vcxproj:
	  Drive the ACM codec through the Audio Compression Manager in its
	  test
	  
	  The rate and configuration checks compile the codec's sources in,
	  which says
	  nothing about the codec Windows actually loads. This adds a
	  second part that
	  takes the built binary and hands it to the ACM: the framework
	  registers the
	  driver for the calling process alone - no registry entry, no
	  administrator -
	  and then does the message dispatch, the format negotiation and
	  the buffer
	  handling, as an application converting audio would.
	  
	  A second of a sine wave is encoded and the result is read back as
	  MPEG frames
	  rather than counted as bytes. A byte total that comes out low
	  cannot
	  distinguish a tail the encoder never flushed from a bitrate it
	  quietly
	  substituted; the frame headers can.

2026-08-16 10:33  aleidinger

	* test/Makefile.am, test/Makefile.in, test/clients,
	  test/clients/acm_test.cpp, test/clients/ctest.h,
	  vc_solution/Makefile.am, vc_solution/Makefile.in,
	  vc_solution/vs_lame_acm_test.vcxproj,
	  vc_solution/vs_lame_clients.slnx:
	  Test the ACM codec's rate selection and its stored configuration
	  
	  Neither Windows client component has a test of what it computes.
	  The
	  autotools suite cannot reach them - both are distributed but
	  never compiled
	  there - and the test project already in the clients solution
	  exercises the
	  encoder library rather than either client.
	  
	  This adds a console program that compiles the codec's own sources
	  in and
	  asserts the sample rate smart output mode offers for a given
	  source and
	  bitrate, and that a configured ratio with a fractional part
	  survives being
	  written and read back. Each rate case is paired with the
	  arithmetic the
	  function used to have, which has to answer differently, so a
	  build where the
	  call had been dropped would not pass.
	  
	  The assertions come from a header in the tree rather than a test
	  library, so
	  the tests run for whoever unpacks the distribution with nothing
	  else
	  installed.

2026-08-16 10:32  aleidinger

	* ACM/DecodeStream.h:
	  Drop the decoding side's unused output sample rate declaration
	  
	  The decoding class declares a member that has never been defined
	  and never
	  been called. Nothing in the ACM's own interface asks for one: the
	  output
	  sample rate of a decoded stream is the one the destination format
	  already
	  carries, and the driver hands it straight to the stream object.
	  Only the
	  encoding side derives a rate, because only encoding may choose a
	  different
	  one.

2026-08-16 10:32  aleidinger

	* doc/html/history.html:
	  Note the ACM smart-output changes in the changelog
	  
	  One entry for both: what a user of the codec sees is a single
	  change to
	  how smart output mode behaves. The affected format pairs are
	  named,
	  because a conversion that resamples where it did not is worth
	  being able
	  to recognise.

2026-08-16 10:31  aleidinger

	* ACM/ACMStream.cpp:
	  Compare the ACM's compression ratio without flooring it first
	  
	  In smart-output mode the codec derives an output sample rate from
	  the
	  compression ratio and then refuses any conversion whose
	  destination does
	  not match it, so this ratio decides which conversions are
	  possible at
	  all. Every operand of the expression is an integer, so the
	  division
	  truncated before the comparison and a ratio of 13.7 was treated
	  as 13.
	  
	  Three source and destination pairs are affected, all of them at
	  that
	  ratio: 24 kHz stereo at 56 kbps, 48 kHz mono at 56 kbps and 48
	  kHz
	  stereo at 112 kbps. Each of them now downsamples as intended
	  instead of
	  keeping the source rate.

2026-08-16 10:31  aleidinger

	* ACM/AEncodeProperties.cpp:
	  Keep the ACM's smart-output ratio as written in its configuration
	  file
	  
	  The smart-output threshold is a compression ratio, and the ratios
	  it is
	  compared against are fractional - 13.78 for 44.1 kHz stereo at
	  128 kbps
	  against a default threshold of 15 - so the fractional part is
	  where the
	  setting does its work. It was stored as a whole number, which
	  left a
	  configured 2.5 working for the session and floored on the next
	  save.
	  
	  Both directions now go through the C locale rather than the
	  host's. The
	  codec is loaded into whatever application asked for it, so the
	  decimal
	  separator in force is that application's: without this, a ratio
	  one host
	  wrote would read back wrong on the next. Every other value in the
	  file
	  is a whole number, a flag or a name, so nothing written by an
	  earlier
	  build changes meaning, and a whole ratio is still written whole.

2026-08-16 10:31  aleidinger

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/AEncodeProperties.cpp,
	  dshow/Encoder.cpp, dshow/Mpegac.cpp, dshow/PropPage_adv.cpp,
	  dshow/aboutprp.cpp, test/lame_test.cpp,
	  vc_solution/vs_clients_config.props,
	  vc_solution/vs_dshow_config.props,
	  vc_solution/vs_lame_acm.vcxproj,
	  vc_solution/vs_lame_acm_adbg.vcxproj,
	  vc_solution/vs_lame_acm_tinyxml.vcxproj,
	  vc_solution/vs_lame_dshow.vcxproj:
	  Hold the Windows client components to the /W4 warning bar
	  
	  The ACM codec and the DirectShow filter build again, but neither
	  had ever
	  been measured against the warning level the rest of the tree is
	  held to.
	  At /W4 they were not clean, and neither was the small sample
	  encoder that
	  sits in the same solution.
	  
	  Most of what the level reported is not this project's code: the
	  vendored
	  XML parser and the DirectShow base classes, which the filter
	  compiles
	  directly because it derives from the classes they implement.
	  Neither can
	  be acted on without forking it, so both are now compiled at their
	  own
	  author's warning level and the bar measures what is left.
	  
	  Of the rest, the unreferenced formal parameters are the case the
	  library
	  already suppresses for this compiler - debug-output stubs and
	  window
	  procedures that must accept a signature they do not use - so the
	  clients
	  now share that policy instead of being the only place it does not
	  apply.
	  The deliberate narrowing conversions are cast at the site, with
	  what
	  bounds the value written beside them, rather than disabled
	  wholesale.
	  
	  The remainder was ordinary: an assignment used as a condition,
	  three
	  locals that were never read, and two returns after a switch whose
	  every
	  case already returns.
	  
	  Two narrowing conversions in the ACM configuration writer are
	  left alone.
	  They put a fractional ratio into an integer attribute, so what
	  they change
	  is a stored value and not only its type; that is a question about
	  the
	  configuration format and is handled on its own.

2026-08-16 10:30  aleidinger

	* dshow/UIDS.H:
	  Make the DirectShow filter's GUID comments agree with its GUIDs
	  
	  Each identifier is written twice: once as the registry spelling
	  in a comment
	  and once as the numbers the compiler reads. All four disagreed,
	  every one of
	  them in the second field, and nothing in a build compares the
	  two.
	  
	  The numbers are what the filter registers and instantiates under,
	  so the
	  comments are what changes. They are also the form a reader copies
	  into a
	  registry entry or a bug report, where the wrong value points at a
	  class that
	  does not exist.

2026-08-16 10:30  aleidinger

	* doc/html/history.html:
	  Note the Windows client components in the changelog
	  
	  One entry for the whole set, since what a user sees is a single
	  change: two
	  components that did not build now do.

2026-08-16 10:29  aleidinger

	* Makefile.am, Makefile.in, doc/maintainer-build-matrix.md,
	  maintainer/gen-build-matrix.ps1,
	  maintainer/setup-windows-deps.ps1, maintainer/smoke-clients.ps1,
	  vc_solution/.gitignore:
	  Bring the Windows client components into the build matrix
	  
	  The build matrix covered the main solution and never the clients
	  one, so
	  neither the ACM codec nor the DirectShow filter was built by
	  anything that
	  reports a result. The codec needs nothing beyond Visual Studio
	  and is therefore
	  always built; the filter is built where its base class sources
	  are laid out,
	  and the generator says which cells it did not produce and why,
	  the same way it
	  already does for the optional libraries.
	  
	  Both cells end in a smoke test whose exit status is the cell's,
	  because a clean
	  build answers a different question than the one that matters for
	  these two.
	  They are DLLs that Windows loads into another program's process:
	  the test loads
	  each one, which runs its initialisation, resolves the entry
	  points it exists to
	  provide, and reads its import table for a runtime that would have
	  to be shipped
	  beside it. The import table is read from the binary rather than
	  taken from the
	  project settings, since that is the only place the answer
	  actually lives.
	  
	  The setup script lays the base class sources out like the other
	  dependencies.
	  Because they are a folder inside a sample repository rather than
	  a release
	  archive, it accepts either the repository archive or just that
	  folder, and
	  finds the right directory by a file only the base classes have
	  instead of
	  trusting the shape of the archive.

2026-08-16 10:28  aleidinger

	* vc_solution/README.vs.txt:
	  Document what the DirectShow filter actually needs
	  
	  The build instructions asked for an SDK that Microsoft retired
	  years ago, and
	  for a solution inside it to be converted and built first. What
	  the filter needs
	  is the base class sources, which are published on their own, and
	  unpacking them
	  is the whole setup.

2026-08-16 10:28  aleidinger

	* aclocal.m4, configure, vc_solution/Makefile.am,
	  vc_solution/Makefile.in, vc_solution/vs_dshow_config.props,
	  vc_solution/vs_lame_clients.slnx,
	  vc_solution/vs_lame_dshow.vcxproj,
	  vc_solution/vs_win71sdk_config.props:
	  Build the DirectShow filter against base class sources, not an
	  SDK
	  
	  Building the filter used to mean installing the Windows SDK 7.1,
	  opening a
	  solution from its samples folder, converting it and compiling two
	  of its
	  configurations first. That SDK has been retired for years, and
	  none of it was
	  ever necessary: the base classes compile against the current
	  Windows SDK with
	  the current toolset, and what the filter actually needs from them
	  is their
	  source, because the classes it derives from are implemented in
	  the .cpp files.
	  
	  The sources are now compiled into the filter and configured the
	  way every other
	  external dependency here is configured - one switch and one path
	  in a property
	  sheet - so getting them amounts to unpacking a folder. They are
	  treated as
	  external headers so that their warnings stay theirs, and the
	  filter is not
	  selected for building unless they are present, the same
	  arrangement the
	  analyzer has with its toolkit. Building it without them says
	  which switch to
	  set instead of failing on a missing include.

2026-08-16 10:27  aleidinger

	* dshow/Mpegac.cpp, vc_solution/vs_lame_dshow.vcxproj:
	  Link the DirectShow filter, and let its initialisation run
	  
	  More than half the unresolved symbols were the DirectShow class,
	  interface and
	  media type identifiers. They live in a library the project never
	  listed, so
	  every one of them was missing.
	  
	  The rest were the C runtime, and the cause was the project's own
	  entry point
	  override. Since Visual Studio 2015 the runtime is three
	  libraries, and the
	  directives naming two of them arrive with the startup object that
	  the linker
	  selects for the entry point it chose itself. Override that and
	  the other two
	  are never searched.
	  
	  Dropping the override alone would have been a silent regression
	  rather than a
	  fix: with no DllMain of its own the filter gets the runtime's
	  do-nothing stub,
	  links cleanly, and never initialises. It now provides one that
	  forwards, so the
	  runtime comes up first and the filter second.
	  
	  The filter is also a DLL that Windows loads into another
	  program's process, and
	  it now links the static runtime for the same reason as the codec
	  next to it.
	  
	  Finally the project and the linker disagreed about the name of
	  the file being
	  produced.

2026-08-16 10:26  aleidinger

	* ACM/acm.rc, ACM/lameACM.def, ACM/readme.txt,
	  vc_solution/vs_lame_acm.vcxproj,
	  vc_solution/vs_lame_acm_adbg.vcxproj,
	  vc_solution/vs_lame_acm_tinyxml.vcxproj:
	  Build the ACM codec with the tools Visual Studio installs
	  
	  The resource script included an MFC header while the project is
	  configured
	  without MFC, and the Build Tools editions carry no MFC at all, so
	  the resource
	  compiler stopped before anything reached the linker. The non-MFC
	  equivalent
	  supplies everything the script uses.
	  
	  The codec is a DLL that Windows loads into another program's
	  process, where a
	  missing redistributable is indistinguishable from a broken codec.
	  It now links
	  the static runtime, which is what libmp3lame and lame.exe have
	  always used: the
	  codec was the one part built against the DLL runtime while
	  linking a static
	  library built against the other.
	  
	  Its module definition file also declared a library name and a
	  description.
	  Neither has an effect on this target and both only produced
	  linker warnings.
	  The export list is what the file is for.
	  
	  The readme asked for a development kit that has not been
	  available for two
	  decades. The Windows SDK supplies the one header that was needed
	  from it, and
	  the other has been in the source tree all along.

2026-08-16 10:26  aleidinger

	* frontend/parse.c, libmp3lame/id3tag.c, misc/abx.c:
	  Convert to unsigned char before calling the character class
	  functions
	  
	  The <ctype.h> functions are defined for the values of an unsigned
	  char
	  and for EOF, and nothing else. The seventeen call sites that
	  passed a
	  plain char directly could reach them with a negative value for
	  any byte
	  above 0x7F - on glibc, an index taken from before the start of
	  the
	  classification table. Ordinary input reaches them: a tag field or
	  a
	  genre name with an accented letter, and a command line with one.
	  
	  The two ID3 helpers that need this in several places now say it
	  once.

2026-08-16 10:25  aleidinger

	* libmp3lame/takehiro.c:
	  Bound the previous-noise lookup to the scale factor bands that
	  exist
	  
	  When the upper part of a granule is zero the quantization loop
	  skips the
	  remaining bands by setting the band index one past the last one.
	  The
	  previous-noise cache is still consulted at that index in the same
	  pass,
	  and for a short block that index is one past the end of the cache
	  array,
	  so the branch it decides was taken on whatever followed the array
	  in
	  memory rather than on a noise value. Restrict the lookup to
	  indices the
	  cache has.
	  
	  The assertion a few functions further down stated only the upper
	  half of
	  the range its two reads need; it now states both.

2026-08-09 17:56  aleidinger

	* Makefile.MSVC, doc/doxygen/vector-dispatch.dox,
	  libmp3lame/lame.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vector/Makefile.am, libmp3lame/vector/Makefile.in,
	  libmp3lame/vector/avx512_calc_sfb_noise.c,
	  libmp3lame/vector/avx512_choose_table.c,
	  libmp3lame/vector/lame_intrin.h, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_calc_sfb_noise_vector.c,
	  test/unit/test_choose_table_vector.c,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Drop the two AVX-512 routines the measurements did not support
	  
	  Two vector routines were carried unreachable, behind environment
	  variables and
	  an alpha-only build, so that the one thing nobody could answer
	  from a desk
	  could be answered from a machine: what they cost inside a real
	  encode. Testers
	  on two microarchitectures have now answered it, twenty-five runs
	  an arm in both
	  orderings, and neither routine earns its place.
	  
	  The scalefactor-band noise estimate costs about five percent of a
	  variable
	  bitrate encode where it runs at all, and cannot be told from zero
	  on the other
	  part. Its reduction is pinned to a fixed association so that
	  every processor of
	  an architecture produces the same bits, and that pin does not
	  widen with the
	  registers, so the extra lanes buy the cost of entering them and
	  nothing else.
	  
	  The Huffman table search splits by vendor, a small gain on one
	  and a cost on
	  the other, and it is reached from both the constant and the
	  variable bitrate
	  paths - so there is no setting at which it could be turned on for
	  everyone
	  without the mode test the dispatch ladder exists to avoid.
	  
	  The tier above AVX2 keeps the routine that did earn it, the
	  quantization loop.
	  Nothing a user can observe changes: no released version compiled
	  either of the
	  removed routines, and no build ran them unless its environment
	  asked.
	  
	  The reasoning, the numbers and what they were measured against
	  are in the
	  internal dispatch documentation, next to the older decisions of
	  the same shape,
	  because a routine that was written and then rejected is exactly
	  the work that
	  gets done twice if the result is not written down.

2026-08-09 13:46  aleidinger

	* libmp3lame/mpglib_interface.c:
	  Ask libmpg123 whether it answered before using the answer
	  
	  The encoder delay and padding are read into a local that is left
	  uninitialised and fetched by a call whose result was discarded.
	  Nothing
	  today makes that call fail, so nothing today goes wrong; but the
	  value
	  goes straight out to the caller as a count of samples to trim,
	  and there
	  is no version of that being worth leaving to whether a library
	  keeps
	  behaving as it does now. A call that does not answer now reports
	  the
	  figure as unavailable, which is what every caller already
	  handles.
	  
	  That is also the more useful reading of -1, and the documentation
	  gave
	  the other one: it said the figure did not fit. It usually means
	  the
	  stream never carried it, since these two live in the LAME tag and
	  most
	  of what a decoder meets has no such tag. A caller told only about
	  overflow would treat an ordinary file as a damaged one.

2026-08-09 13:46  aleidinger

	* libmp3lame/mpglib_interface.c, test/unit/test_decoder_api.c:
	  Report an error for a decoder handle that was never created
	  
	  hip_decode_init() hands back nothing where the library was built
	  without
	  libmpg123, and its documentation tells callers that checking the
	  result
	  is enough, because a caller who skips the check meets the same
	  absence at
	  the first decode instead, as an error. One of the five decoding
	  entry
	  points kept that promise. The other four read the handle first,
	  so what
	  a caller actually met was a crash - and only in a library that
	  does have
	  a decoder, which is the configuration where the check is easiest
	  to
	  forget.
	  
	  The guard goes where all of them converge rather than into each,
	  and the
	  existing one stays: it is the only one there is in a build with
	  no
	  decoder, where none of the shared path is compiled. The three
	  return
	  descriptions that did not mention a null handle now do, so the
	  five read
	  alike.

2026-08-09 13:45  aleidinger

	* test/unit/test_decoder_api.c:
	  Test the decoder API, including the build that has no decoder
	  
	  The decoding half of the library had three tests, all about a
	  handle
	  that was never created. What a decoder does with a stream this
	  library
	  encoded had no coverage: what it reports about the frames,
	  whether one
	  call returns what several return in pieces, and where the encoder
	  delay
	  comes from - which is the only reason to prefer one entry point
	  over
	  another.
	  
	  The tests skip rather than fail where the library was built
	  without
	  libmpg123, so one file covers both configurations and the
	  decoderless
	  build is checked rather than excluded. The obsolete entry points
	  need no
	  decoder at all. What they promise is to stay inert, and that
	  includes
	  not writing to the buffers they are handed - worth pinning
	  precisely
	  because nothing else would notice if it stopped being true.

2026-08-09 13:44  aleidinger

	* test/unit/test_rtp_socket.c:
	  Let the unit tests build on MinGW
	  
	  The RTP socket test asked for Winsock only where the compiler is
	  not
	  MinGW, so on a MinGW toolchain it took the POSIX branch and
	  stopped at a
	  header that does not exist there. The build stops at the first
	  target
	  that fails, so this cost not that one test but every unit test on
	  that
	  toolchain.
	  
	  rtp.c, which the test compiles in and shares a socket with, has
	  always
	  spelled the same question the other way, and the two have to
	  agree: the
	  shared socket is declared in terms of a type only one of the
	  branches
	  defines. The receive timeout goes with it, since Winsock wants
	  milliseconds where the POSIX call wants a struct.

2026-08-09 13:44  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_encode_api.c:
	  Test the encode entry points, the statistics and the LAME tag
	  
	  The integer encode entry points each declare a different input
	  scaling,
	  and nothing recorded what that means for the stream they produce.
	  Fed
	  the same audio at the scaling each one asks for, they agree byte
	  for
	  byte. That is now a test rather than something to be
	  rediscovered, and
	  it comes with the same audio at the wrong scaling, so the
	  agreement
	  cannot quietly come to hold for no reason.
	  
	  The post-encode statistics, the two LAME tag calls and the two
	  reporting calls had no coverage at all. The histograms are pinned
	  through their invariants - what they sum to, and whether the two
	  dimensional tables agree with the one dimensional ones - rather
	  than
	  through any particular set of counts, so a future encoder change
	  moves
	  them without breaking anything.

2026-08-09 13:44  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  test/unit/test_get_audio_wav.c:
	  Refuse a WAVE sample width the reader cannot unpack
	  
	  The WAVE reader guarded only against a width of zero, so any
	  other value the
	  header declared was taken at face value. Integer widths it has no
	  case for were
	  caught later, by the sample reader, in a message naming neither
	  the file nor the
	  field.
	  
	  Floating point data was worse off. Such a file is read at 32 bits
	  and converted;
	  declared at any other width it was unpacked as integers of that
	  width and the
	  buffer then reinterpreted as floats, which produced noise and no
	  diagnostic
	  anywhere along the way.
	  
	  The accepted widths are now settled where the header is read,
	  from what the
	  reader implements: 8, 16, 24 and 32 bit integers, and 32 bit
	  floating point.
	  The AIFF reader has enumerated its widths all along; the two now
	  agree.

2026-08-09 13:43  aleidinger

	* include/lame.h:
	  Name the calling convention on the last public function missing
	  it
	  
	  Every other function declaration in the installed header pins its
	  calling
	  convention explicitly; this one relied on whatever the consuming
	  project's
	  compiler defaults to. A caller built with a non-default
	  convention therefore had
	  no agreement with the library about this entry point alone.
	  
	  The exported set, the symbol's name and the generated code are
	  all unchanged.

2026-08-09 13:43  aleidinger

	* libmp3lame/set_get.c:
	  Document what an input sample rate does and does not survive
	  
	  The reference described the resampling but not what becomes of
	  the input rate
	  itself. It is recorded nowhere in the stream: a frame names its
	  own sampling
	  frequency from the fixed set the format defines, so a decoder
	  reports the rate
	  LAME encoded at, and the rate the material arrived at cannot be
	  recovered from
	  the file.
	  
	  The accepted range is now stated as well. There is no upper bound
	  beyond the
	  parameter's own type, and rates far above anything the output
	  formats allow are
	  resampled like any other.

2026-08-09 13:42  aleidinger

	* frontend/get_audio.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_get_audio_wav.c:
	  Reject a WAVE sample rate the encoder cannot represent
	  
	  A WAVE header states its sample rate in 32 unsigned bits, while
	  the encoder
	  takes that rate as a signed integer. A file declaring a rate in
	  the upper half
	  of the field's range was converted before anything examined it,
	  and the refusal
	  that followed reported the converted value - a header claiming
	  4294967295 Hz
	  was refused as a sample rate of -1, a number appearing nowhere in
	  the file.
	  
	  The rate is now checked while the declared value is still intact,
	  so such a
	  header is refused as the malformed field it is and the message
	  names the rate
	  the file actually carries. Nothing the encoder can use is
	  affected: the high
	  rates it resamples down are accepted as before, and the accepted
	  range now ends
	  exactly where the encoder's own does.
	  
	  The AIFF reader already checks its sample rate this way; the two
	  readers now
	  agree.

2026-08-04 16:32  aleidinger

	* doc/doxygen/vector-dispatch.dox, libmp3lame/util.c:
	  Read the AVX-512 experiment switches once per encode
	  
	  The two environment variables that select the experimental
	  AVX-512 kernels were
	  read wherever the question was asked. One of those places is the
	  dispatch of the
	  scalefactor-band noise estimate, which a variable-bitrate encode
	  enters millions
	  of times, so a machine running the AVX-512 tier walked the whole
	  environment
	  block on each of them - and did so precisely when the variable
	  was unset, which
	  is the arm anyone measuring the experiment uses as their
	  reference.
	  
	  The cost is large enough to be mistaken for the thing under test.
	  It appears
	  only on that tier and only in variable-bitrate mode, because
	  those are the
	  conditions the gate tests, so it wears the shape of a processor's
	  behaviour;
	  two testers on different vendors reported it as one, and it
	  reproduces on a
	  machine that cannot execute a single AVX-512 instruction.
	  
	  The environment is now read where the vector routines are chosen,
	  once, as the
	  encoder is initialised. Both switches are settings of a run
	  rather than
	  questions asked inside it, so setting either after initialisation
	  no longer has
	  an effect - which is what a switch this shape should have meant
	  all along.
	  
	  Nothing a released build does changes: the whole path is compiled
	  only into an
	  alpha, and the value returned for any given environment is the
	  same one as
	  before.

2026-08-04 16:32  aleidinger

	* doc/maintainer-build-matrix.md, doc/maintainer-check-dist.md,
	  maintainer/check-dist.ps1, maintainer/check-dist.sh,
	  maintainer/gen-build-matrix.sh:
	  Skip the strict cell where its automake is absent, rather than
	  failing
	  
	  The build matrix has one cell that enables maintainer mode, and
	  maintainer mode
	  switches on the Autotools rebuild rules. Those call the
	  version-suffixed tool
	  the generated Makefiles name, so on a machine carrying a
	  different automake the
	  tool is simply not there, make stops with status 127, and the
	  distribution check
	  reports a failure that is about the host rather than about the
	  source.
	  
	  The generator now reads the required version out of the tree's
	  own configure and
	  treats it like any other missing prerequisite: the cell is not
	  laid out, and
	  matrix-info.txt records why. Nothing has to be kept in step by
	  hand when the
	  tree is regenerated with a newer automake.
	  
	  A cell that is not laid out had no directory and so left no trace
	  in the
	  distribution check at all - the run merely had fewer checks in
	  it, with nothing
	  saying which were gone. Each one is now reported as a skipped
	  check naming its
	  reason, which also covers the cells dropped for a missing
	  library. A
	  configuration that quietly stops being checked is worse than one
	  that says it is
	  not being checked.

2026-08-04 16:32  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in:
	  Link the file io test against the backend it was built with
	  
	  The AIFF reader test compiles get_audio.c in, so it inherits
	  whichever file
	  io backend configure selected. Under --with-fileio=sndfile that
	  file calls
	  libsndfile directly, but the test named only cmocka and
	  libmp3lame, so it
	  failed to link there while every other configuration was fine and
	  the
	  default backend hid it entirely.
	  
	  The libraries are empty for the default backend, so nothing
	  changes for it.
	  The include path was already right, reaching every target in that
	  directory.

2026-08-04 16:31  aleidinger

	* Makefile.MSVC:
	  Name the Blade DLL source with the case it actually has
	  
	  Makefile.MSVC listed the encoder DLL source as
	  dll/BladeMP3EncDll.c, while the
	  file in the repository is Dll/BladeMP3EncDLL.c. This makefile is
	  the nmake build
	  and so only ever runs on Windows, where the lookup is
	  case-insensitive and the
	  path resolves regardless, which is why it went unnoticed since
	  the file was
	  imported; on a case-sensitive filesystem it names nothing.
	  
	  The remaining lower-case dll references are the object directory
	  the makefile
	  creates for itself and the target of the same name, neither of
	  which is a path
	  in the repository, and they are unchanged. The build produces the
	  same DLL.

2026-08-02 13:06  aleidinger

	* include/lame.h:
	  Correct the reason a LAME tag may not be written
	  
	  The comment named the output not being a regular file; the
	  library is handed
	  the stream when the call is made and never learns what kind of
	  file it is.
	  Whether a tag is written was decided earlier, by the caller or by
	  the encoder.

2026-08-02 13:06  aleidinger

	* doc/doxygen/api.dox, libmp3lame/VbrTag.c, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/tables.c:
	  Document the exported functions the API reference was missing
	  
	  Seventeen functions of the installed interface carried no Doxygen
	  block, so
	  the reference neither described them nor could link to them from
	  its own
	  prose: the interleaved and wider-type encode entry points, the
	  statistics
	  calls, the two LAME-tag calls and the decoder's analysis hooks.
	  Most of them
	  did carry an explanation in the header, written in a form the
	  generator does
	  not read.
	  
	  The blocks sit at the definitions, as the rest of the interface
	  does. The
	  statistics calls gain a section of their own, which is where the
	  properties
	  they share belong: when the counters may be read, what the
	  bitrate and
	  stereo-mode slots stand for, and that a block type is counted per
	  granule and
	  channel rather than per frame.
	  
	  Documentation only; the library behaves as before.

2026-08-02 10:35  aleidinger

	* doc/Doxyfile.in, doc/doxygen/api.dox, libmp3lame/set_get.c:
	  Let the API reference link to every function it documents
	  
	  The reference could not refer to eight of the functions it
	  documents. A
	  mention of one of them in the section text came out as plain
	  words while
	  every neighbouring mention became a link, and nothing reported
	  it: the
	  documentation generator had quietly resolved the name to a
	  second, empty
	  entry for the same function rather than to the one carrying the
	  text.
	  
	  That second entry exists because a definition and its declaration
	  are only
	  the same function to the generator if they read the same. Two
	  spellings in
	  this tree differ without differing: a parameter a build may
	  legitimately not
	  use is marked as such at the definition, and three callback
	  parameters are
	  written out longhand where the installed header names the type.
	  The marker
	  is now resolved before the comparison, and the three parameters
	  are spelled
	  the way the header spells them.
	  
	  Resolving it is scoped to the configuration that publishes the
	  reference.
	  That configuration reads the interface and nothing else, and it
	  is the one
	  where the empty entry has no page left to fall back on. The
	  internal
	  documentation covers the whole tree, where reading every macro
	  before the
	  comparison would change entries that have nothing to do with
	  this, and it
	  keeps its file pages, so the same mention links there either way.
	  
	  The interface is untouched - the marker is not part of any type
	  and the
	  callback type is the same type under both spellings - so the
	  library builds
	  to the same code and encodes the same bitstream. What changes is
	  that a
	  reader who meets a function's name in a description can now
	  follow it, for
	  all of the interface rather than most of it.

2026-08-02 09:19  aleidinger

	* doc/Doxyfile-common.in, doc/Doxyfile-internal.in,
	  doc/Doxyfile.in, doc/Makefile.am, doc/Makefile.in,
	  doc/doxygen/DoxygenLayout.xml, doc/doxygen/api.dox,
	  doc/doxygen/mainpage.md, include/lame.h, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/set_get.c, libmp3lame/version.c:
	  Present the API documentation as a reference rather than as files
	  
	  The generated API documentation described the library one
	  implementation
	  file at a time: five of libmp3lame's sources had a page of their
	  own, and
	  the file index listed them, so a reader looking for the interface
	  was
	  offered the encoder's internals first. That was never a decision
	  - it is
	  what a documentation generator does by default when the
	  descriptions are
	  written next to the code they describe, and it survived the
	  earlier round
	  that stopped publishing the headers verbatim.
	  
	  The descriptions stay exactly where they are, at each function's
	  definition, and are now collected into a reference organised by
	  what a
	  caller is doing: encoding, settings, tags, decoding and version,
	  in the
	  order a program uses them. The public documentation carries no
	  file pages
	  at all, so nothing implementation-side appears in it; the
	  internal
	  documentation is unchanged and still shows every file. Every
	  entity that
	  was documented before is documented after, on the section page
	  for its
	  part of the interface.

2026-08-02 09:13  aleidinger

	* doc/html/history.html, doc/man/lame.1, frontend/parse.c,
	  libmp3lame/lame.c, test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_free_format.c:
	  Refuse a free format bitrate too low to carry a frame
	  
	  Free format takes the bitrate from the caller rather than from
	  the
	  standard's tables, so it can name a frame smaller than the side
	  information that frame has to carry - MPEG-1 at 48 kHz and 8
	  kbit/s is 24
	  bytes against 38. Nothing rejected it, and what happened next
	  depended on
	  the build: with assertions live the encoder ended the process
	  partway
	  through the first frame, and with them compiled out it reported
	  success
	  after writing megabytes of data that was not an MP3.
	  
	  It is now refused before encoding starts, and the report names
	  the lowest
	  bitrate that will work for the sample rate and channel count in
	  use. The
	  floor is the room the tightest stream the bitrate tables allow
	  leaves for
	  audio, so nothing that encodes today is affected. The manual page
	  said any
	  bitrate above 8 kbit/s could be used with this option, which is
	  true only
	  where a frame is large enough for one.

2026-08-02 09:12  aleidinger

	* doc/html/history.html, libmp3lame/util.h, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_bitstream_headers.c:
	  Keep the frame headers of a file that begins with a long silence
	  
	  At the format's low end - MPEG-2 at 8 kbit/s, two channels with
	  CRC - a
	  frame has one byte of room beyond its own side info, so a silent
	  lead-in
	  leaves one frame header waiting to be written for every eight
	  bits of the
	  bit reservoir. Past six seconds of silence there were more of
	  them than
	  the encoder had room for, and it overwrote the oldest rather than
	  keeping
	  it: the file lost the headers of the next 255 frames, while the
	  encoder
	  reported it once and returned success. There is now room for
	  twice as
	  many headers as the reservoir can call for, and no other setting
	  comes
	  near even the old limit.

2026-08-01 17:03  aleidinger

	* configure, configure.ac, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_rtp_socket.c:
	  Add unit tests for the RTP sender's destination handling
	  
	  The RTP sender had no test of any kind, and it is the one
	  frontend
	  component whose behaviour is a protocol rather than a file
	  format: what
	  it accepts as a destination, what it refuses, and what it puts on
	  the
	  wire.
	  
	  The tests read back what the sender connected to rather than
	  inferring
	  it, so they hold it to the address and the address family that
	  were
	  asked for, and one of them receives what the sender emits and
	  checks
	  the RTP header against the payload it was given. An IPv4-only
	  build is
	  covered as well: it has to refuse an IPv6 destination, not accept
	  it
	  and do something else with it.
	  
	  Loopback delivery is not something a unit test can insist on - a
	  host
	  firewall may drop it, and some do - so the test that needs it
	  probes
	  the path first and skips when nothing can get through. The
	  distinction
	  matters in the other direction too: whether a name resolves on
	  this
	  machine is asked separately from whether the sender can resolve
	  names,
	  because only the first is a reason to skip.
	  
	  The tests are built where the socket and resolver calls exist,
	  which is
	  the same answer that decides whether mp3rtp can be built at all.

2026-08-01 17:03  aleidinger

	* vc_solution/vs_gtk_config.props:
	  Stop holding GTK's headers to our own warning level
	  
	  The frame analyzer compiles GTK's headers as part of its own
	  translation units, so warnings from a dependency's headers were
	  reported against this project and moved with whatever GTK release
	  happened to be installed.
	  
	  Those include directories are now declared as external ones,
	  which the
	  compiler treats accordingly and which leaves our own sources at
	  the
	  level they had. That is preferable to naming individual warning
	  numbers in our source, which would have to be revisited whenever
	  the
	  dependency changes.

2026-08-01 17:02  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.MSVC,
	  Makefile.in, config.h.in, configMS.h, configure, configure.ac,
	  doc/Makefile.in, doc/html/Makefile.in, doc/html/history.html,
	  doc/man/Makefile.in, doc/man/mp3rtp.1, dshow/Makefile.in,
	  frontend/Makefile.am, frontend/Makefile.in, frontend/mp3rtp.c,
	  frontend/rtp.c, include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in, test/Makefile.in, test/unit/Makefile.in,
	  vc_solution/Makefile.in, vc_solution/vs_lame_mp3rtp.vcxproj:
	  Make mp3rtp behave the same on every platform
	  
	  The RTP sender had two implementations, one for Windows and one
	  for
	  everything else, and they had drifted apart in what they do
	  rather than
	  only in how they spell it: only one of them resolved host names
	  or
	  accepted IPv6 destinations, only one enabled broadcast, and they
	  disagreed about the multicast loopback option and about how an
	  error is
	  reported. Which one a user got depended on their platform, and
	  the
	  manual page had to describe both.
	  
	  There is now one implementation. A destination may be an IPv4
	  address,
	  an IPv6 address or a host name everywhere, and the frontend no
	  longer
	  filters that argument through a character set no host name could
	  pass -
	  whether an address exists is the resolver's answer. A build for
	  IPv4
	  destinations only is still available, and is now offered on
	  Windows as
	  well, where there had been no such choice.
	  
	  Building it needed fixing too. Under MinGW mp3rtp could not be
	  built at
	  all: the configure check looked for the socket function in the C
	  library
	  and in libsocket, and on that platform it is in neither, so the
	  program
	  was refused. The check now links the calls the program really
	  makes, and
	  only mp3rtp links the library that provides them. The Visual
	  Studio
	  project and the nmake makefile name that library as well - the
	  interfaces replacing the deprecated ones are not in the one those
	  targets had been linking, and without the change the program does
	  not
	  link at all.
	  
	  Removed along the way: a block of #undef directives whose macros
	  are
	  defined nowhere this file can see. The preprocessed output is
	  unchanged
	  by their removal on both a Linux and a FreeBSD compiler, with and
	  without IPv6.
	  
	  This also clears the last deprecation warnings from the one
	  Windows
	  build configuration that compiles this file at the project's
	  warning
	  level.

2026-08-01 09:12  aleidinger

	* vc_solution/README.vs.txt:
	  Correct the dependency path notes in the Visual Studio readme
	  
	  The mpg123 section asked for a trailing backslash the property
	  sheet does not
	  use, and quoted a default that has one; both dependency sections
	  then named the
	  on/off switch where they meant the path macro whose default they
	  were quoting.
	  The path needs no separator of its own, and the case where a
	  trailing one really
	  does harm - a value passed on the MSBuild command line, where it
	  escapes the
	  closing quote - is now the case the text describes.

2026-08-01 09:12  aleidinger

	* maintainer/gen-build-matrix.ps1:
	  Say when a Windows matrix cell is left out for want of a library
	  
	  The optional-library cells are generated only where that library
	  is laid out for
	  the platform being built, while the closing summary asked the
	  probe made for the
	  32-bit nmake cells. A machine holding one platform's library and
	  not the other's
	  therefore got a run that quietly built no cell for it and
	  reported nothing
	  missing - and the run's own record named the library as found, so
	  a matrix that
	  passed read as covering a configuration nothing had attempted.
	  
	  The cells left out are now named, with the platform and the path
	  looked for, on
	  the console and in the run's record beside the cells that were
	  generated.

2026-08-01 09:11  aleidinger

	* Makefile.MSVC, maintainer/gen-build-matrix.ps1,
	  vc_solution/README.vs.txt, vc_solution/vs_libmpg123_config.props,
	  vc_solution/vs_libsndfile_config.props:
	  Build each project's mpg123 import library separately
	  
	  The Windows distribution of libmpg123 ships a module definition
	  file and no
	  import library, so the build makes one. It made it in the
	  distribution folder,
	  from a property sheet that every project using the decoder
	  shares, and those
	  projects are built in parallel - so on a folder nobody had built
	  in yet they all
	  wrote that one file at the same moment, and the build failed on
	  whichever
	  project lost. The library then stayed behind in a directory the
	  build does not
	  own, and the nmake build had quietly come to depend on finding it
	  there.
	  
	  Each project now builds its own into its own intermediate
	  directory, and the
	  nmake build into its output directory. Nothing is written into
	  the distribution
	  folder any more, which needs only the headers, the definition
	  file and the DLL.
	  
	  The runtime libraries copied next to the binaries are shared in
	  the same way, so
	  those copies now leave a file that is already in place alone.

2026-08-01 08:36  aleidinger

	* doc/html/history.html, libmp3lame/id3tag.c,
	  test/unit/test_id3tag.c:
	  Keep ID3v2 frames whose descriptions differ apart
	  
	  Frames identified by a description - TXXX, WXXX and COMM - were
	  taken to be
	  the same frame whenever the stored description merely began the
	  incoming one,
	  so the second of --tv "TXXX=foo=..." and --tv "TXXX=foobar=..."
	  replaced the
	  first and its text was lost. An empty description begins every
	  other one, so a
	  comment carrying none vanished as soon as a described comment was
	  added.
	  
	  A description now has to end where the stored one does. Repeating
	  a
	  description still replaces the frame it names, so correcting a
	  value is
	  unaffected.

2026-08-01 08:35  aleidinger

	* doc/html/history.html, frontend/get_audio.c,
	  maintainer/coverage-workload.txt:
	  Read a file as headerless PCM only when that was asked for
	  
	  A build configured to hand file input to libsndfile accepted
	  anything: when
	  libsndfile did not recognise a file, it was retried as raw PCM at
	  the
	  default rate and channel count, so a corrupt or wrong-format file
	  - eight
	  bytes of anything at all - encoded to a valid MP3 of noise and
	  reported
	  success. The same files are refused by a build using LAME's own
	  reader,
	  which decided long ago that failing beats encoding hundreds of
	  formats it
	  does not support.
	  
	  The retry now happens only for input the user marked as raw with
	  -r, which
	  is the only thing that can say so: libsndfile cannot recognise
	  headerless
	  PCM, so an unrecognised file is nothing more than unrecognised.
	  Everything
	  else falls through to LAME's reader and is refused there.

2026-08-01 08:35  aleidinger

	* config.h.in, configMS.h, configure, configure.ac,
	  frontend/parse.c, libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/set_get.c, libmp3lame/util.c,
	  maintainer/gen-coverage-matrix.sh, test/unit/test_set_get.c:
	  Use one name for the condition that the decoder is built in
	  
	  Two macros were defined together and meant the same thing:
	  HAVE_MPG123,
	  saying libmpg123 is present, and DECODE_ON_THE_FLY, naming the
	  feature that
	  needs it. The Windows configuration header derived the second
	  from the
	  first, so no build system here could set one without the other,
	  and a
	  reader had no way to tell that from two genuinely independent
	  conditions.
	  
	  DECODE_ON_THE_FLY is the one that goes. An out-of-tree build
	  system that
	  defines it has to define HAVE_MPG123 instead - otherwise the code
	  that
	  decodes while encoding is quietly compiled out rather than
	  failing to
	  build.

2026-08-01 08:34  aleidinger

	* doc/html/history.html, frontend/parse.c,
	  maintainer/coverage-workload.txt, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_help_exit.c:
	  Report success when the command line asked for help
	  
	  --preset help and -? printed what was asked for and then exited
	  with the
	  status that means the command line was wrong, while --help,
	  --longhelp,
	  --version and --license all reported success. A script driving
	  lame, or a
	  pipeline checking the exit status, could not tell a help request
	  from a
	  mistake.

2026-08-01 08:33  aleidinger

	* doc/maintainer-build-matrix.md, maintainer/gen-build-matrix.sh:
	  Add a matrix cell that builds without the decoder under fatal
	  warnings
	  
	  The matrix is a star, one cell per configuration axis, which
	  assumes that a
	  regression in an axis is visible from that axis alone. That does
	  not hold
	  where one surface is reachable only through another: the code
	  compiled when
	  the decoder is absent is compiled nowhere else, so an ordinary
	  build has
	  nothing to warn about, and the decoder-off cell does not make a
	  warning
	  fatal. Both flips are needed before there is anything to see,
	  which is why
	  the failures this covers were each found by hand instead.
	  
	  The unit tests stay out of the cell list on purpose. The existing
	  option for
	  extra configure arguments adds them to every cell, this one
	  included, which
	  is the wider combination for no extra cells.

2026-08-01 08:33  aleidinger

	* test/unit/test_get_audio_aiff.c:
	  Keep the AIFF header test's loop counter valid across its trap
	  
	  The test drives the parser under a setjmp trap and names the loop
	  counter
	  in the message it prints when the trap fires. A local that is
	  modified
	  between setjmp and longjmp without being volatile has an
	  indeterminate
	  value after the jump, so that message could report a case other
	  than the
	  one that failed.
	  
	  It is also what kept the unit tests from being built with the
	  stricter
	  warnings turned on, which make the diagnostic fatal.

2026-08-01 08:33  aleidinger

	* libmp3lame/bitstream.c, libmp3lame/tables.c, libmp3lame/tables.h,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_huffman_tables.c:
	  Resolve the codeword-less Huffman table where it is used
	  
	  The table search can choose a table the standard defines no code
	  words
	  for, and it is the side information writer that turns that choice
	  into the
	  one the standard does define - as a side effect of writing the
	  header,
	  into the shared structure the code word writer reads afterwards.
	  So the
	  data half was safe only because the header half runs first, in
	  every path
	  there happens to be, and nothing said so.
	  
	  Both halves now ask for the codebook they are about to use, and
	  the search
	  result is left alone. The bitstream is unchanged: the same
	  substitution
	  happens, at the point where it is needed rather than before it.

2026-08-01 08:32  aleidinger

	* doc/html/history.html, frontend/lame_main.c, libmp3lame/lame.c,
	  libmp3lame/set_get.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_encode_clipgain.c:
	  Report no headroom figure for input that is entirely silent
	  
	  The clipping figures come from the loudest sample the encoder saw
	  in its
	  own decoded output. Material whose samples are all zero leaves
	  that at
	  zero, and the headroom is derived from its logarithm, so the
	  figure was
	  computed from an infinity and converted to an integer, which is
	  undefined.
	  What a user saw was a claim that the waveform is some two hundred
	  million
	  decibels away from full scale.
	  
	  Silence is ordinary input for this measurement - a muted track,
	  an empty
	  capture, a long leading gap - so it gets an answer rather than an
	  exception: no clipping, no scaling, and a message that says there
	  is no
	  headroom to report. The encoded audio is not affected; nor is any
	  input
	  that contains a signal.

2026-08-01 08:31  aleidinger

	* libmp3lame/util.c, libmp3lame/util.h:
	  Remove the unused aligned allocation helpers

2026-08-01 08:30  aleidinger

	* frontend/get_audio.c:
	  Check that the decoder was created before reading from it
	  
	  The frontend's MP3 reader asked libmp3lame for a decoder and then
	  read a
	  field out of it without testing the handle. The library returns
	  nothing
	  when the allocation fails and when the decoder refuses to open
	  the feed,
	  so this is reachable in an ordinary build rather than only under
	  memory
	  exhaustion, and the check that was written - for a member of the
	  handle -
	  reads as though the handle itself had already been established.

2026-08-01 08:30  aleidinger

	* include/lame.h, libmp3lame/mpglib_interface.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_decoder_api.c:
	  Check the decoder handle before using it in the analysis hooks
	  
	  The null check in the call that finishes a frame's analysis data
	  was
	  written after the dereference it guards. Every caller in the tree
	  has
	  already used the handle by the time it calls, so nothing reaches
	  it that
	  way today - but the check reads as though the case were handled,
	  which is
	  what a new caller would rely on. The header now states that both
	  hooks
	  accept a decoder that was never created, which is the ordinary
	  case for a
	  library built without the mpg123 decoder.

2026-08-01 08:30  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_vector_ladder.c:
	  Test the automatic vector-rung choice against synthetic
	  capabilities
	  
	  Which set of vector routines a build picks for itself could until
	  now only
	  be observed on a machine that has the instruction sets in
	  question, so the
	  cases that matter most were exercised nowhere: a processor wider
	  than the
	  binary, which is the ordinary case rather than an exotic one, and
	  a machine
	  offering nothing at all. The choice is made from the recorded
	  capability
	  bits and nothing else, so setting those by hand covers every
	  combination on
	  any host, including the ones that host does not have.

2026-07-31 10:24  aleidinger

	* maintainer/abi/libmp3lame.abi:
	  Refresh the ABI baseline after the vector-routines work
	  
	  Adding the vector-routines interface and retiring the MMX and
	  3DNow! enables
	  changed both the set of exported entry points and the layout of
	  the state
	  behind them. Both were intended; what was missed is that the
	  baseline the ABI
	  check compares against still describes the state before them, so
	  the check
	  reports a change that nobody has recorded a decision about.
	  
	  Record it. The interface only gained entry points, and every
	  layout change
	  reported is inside structures a caller reaches solely through an
	  opaque
	  handle, so nothing that links today stops linking and the soname
	  stands.

2026-07-31 10:24  aleidinger

	* include/lame.def:
	  Add the forgotten new API functions to the Windows symbol file

2026-07-30 19:05  aleidinger

	* Makefile.MSVC, doc/doxygen/vector-dispatch.dox,
	  libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c, libmp3lame/vector/Makefile.am,
	  libmp3lame/vector/Makefile.in,
	  libmp3lame/vector/avx512_calc_sfb_noise.c,
	  libmp3lame/vector/lame_intrin.h, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_calc_sfb_noise_vector.c,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Bring back the AVX-512 band-noise routine, behind a second switch
	  
	  The variable-bitrate noise estimate had an AVX-512 form. It was
	  written,
	  measured at 6.6% slower than the narrower code on a -V 2 encode,
	  and dropped on
	  that number. The number was honest, and it is now known to have
	  been taken in a
	  configuration nobody intended: the AVX-512 table-search routines
	  could not be
	  reached from an encode at all, so this routine was the only
	  512-bit code a -V
	  run executed, and it paid the frequency cost of entering the wide
	  registers by
	  itself. Alongside a table search that is already in them, the
	  same comparison
	  need not come out the same way.
	  
	  That is a question, not a claim, and it can only be answered on
	  hardware the
	  project does not have. So the routine returns compiled but
	  unreachable, under
	  LAME_AVX512_SFB_NOISE and the same alpha-only condition as the
	  switch beside
	  it, and the two together give a tester three arms: neither, the
	  table search
	  alone, and both. Only the last can show an interaction, and it is
	  the arm that
	  did not exist when the routine was rejected.
	  
	  Not un-rejected, then: nothing here changes what any build a user
	  can obtain
	  does, and the decision stands until there are numbers against it.
	  
	  The form selected is the bit-exact one - it quantizes four blocks
	  at a time but
	  folds their sums in the pinned association, so it computes what
	  the narrower
	  tier computes. Output identity is therefore something a tester
	  can check rather
	  than something they have to trade away, and a difference would be
	  a defect.

2026-07-30 19:04  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  config.h.in, configure, configure.ac, doc/Makefile.in,
	  doc/doxygen/vector-dispatch.dox, doc/html/Makefile.in,
	  doc/html/history.html, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/lame.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vector/Makefile.am,
	  libmp3lame/vector/Makefile.in, libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/neon_choose_table.c, mac/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, test/Makefile.in,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_choose_table_neon.c, vc_solution/Makefile.in:
	  Add a NEON tier for the ARM vector routines
	  
	  ARM builds have run the plain C code: the vector routines are x86
	  intrinsics,
	  and until the selection was by name there was no way to offer a
	  set that is not
	  an x86 one. On the current shape this is a row in the table the
	  library already
	  keeps, so the option, the help text and the installed header gain
	  the set
	  without any of them being touched.
	  
	  One routine, not seven. Each of the x86 kernels was taken in turn
	  and measured
	  against what the compiler already produces on AArch64: three of
	  them clang
	  vectorises unprompted - including a native 32-bit signed maximum,
	  which is the
	  only reason the SSE2 sibling of that routine exists at all - and
	  two more were
	  written and came out slower than the scalar code. What pays is
	  the escape-coded
	  Huffman counting loop, worth about 3.5% of a 320 kbit/s encode
	  and about 1% of
	  a -V one. The internal ladder page records which is which, so the
	  four that are
	  absent read as a result rather than as work nobody got round to.
	  
	  The set is built wherever the compiler can build it and used
	  where the
	  processor reports it, on 64- and 32-bit ARM alike. Advanced SIMD
	  is
	  architectural on AArch64, so there the run-time question answers
	  itself; it is
	  optional on ARMv7, where the answer has to come from the
	  operating system, and
	  every one of them spells it differently.

2026-07-30 19:03  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.MSVC,
	  Makefile.in, config.h.in, configMS.h, configure, configure.ac,
	  doc/Makefile.in, doc/doxygen/internal.md,
	  doc/doxygen/vector-dispatch.dox, doc/html/Makefile.in,
	  doc/html/detailed.html, doc/html/history.html,
	  doc/man/Makefile.in, doc/man/lame.1, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/lame.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vector/Makefile.am,
	  libmp3lame/vector/Makefile.in,
	  libmp3lame/vector/avx512_choose_table.c,
	  libmp3lame/vector/avx512_quantize_lines.c,
	  libmp3lame/vector/lame_intrin.h, mac/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, test/Makefile.in,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_choose_table_vector.c,
	  test/unit/test_quantize_lines_vector.c, vc_solution/Makefile.in,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Add an AVX-512 tier for the constant-bitrate vector routines
	  
	  The vector ladder gains a rung above AVX2, carrying the
	  constant-bitrate
	  quantization loop. Three members of the lame-dev list measured it
	  on hardware
	  the project does not own: on an Intel Ice Lake part it is worth
	  about 6.5% of a
	  320 kbit/s encode, and on the two AMD processors it was measured
	  on it makes no
	  difference or costs about 1%.
	  
	  It is enabled wherever the processor can execute it, with no test
	  on vendor or
	  model. A percent at one setting on one microarchitecture does not
	  pay for a
	  dispatcher whose behaviour cannot be predicted from the hardware
	  alone, so the
	  trade-off is stated in the changelog instead of being encoded in
	  the selection.
	  The gate that confined the tier to alpha builds while its worth
	  was unknown is
	  gone with the question it existed for.
	  
	  The Huffman table search is a second question and is left open
	  rather than
	  answered by assumption. AVX-512 forms of it exist and are tested,
	  but the table
	  search is reached through a function pointer that only a per-tier
	  wrapper sets,
	  and no such wrapper was ever written - so those routines have
	  never executed
	  inside an encode, on any machine, and the measurements above
	  cannot be about
	  them. The wrapper is added here and installed only in an alpha
	  build with
	  LAME_AVX512_CHOOSE_TABLE set, which is a request for
	  measurements, not a
	  default. An alpha that has taken the path says so in the
	  --verbose header, so a
	  report can be told apart from a report about a run that did not.
	  
	  The variable-bitrate noise estimate deliberately has no AVX-512
	  version. One
	  was written, and it both failed to reproduce the narrower tier's
	  output exactly
	  - widening the sum breaks the association that keeps every
	  processor of an
	  architecture bit-identical - and ran 6.6% slower at -V 2 than the
	  code it would
	  have replaced, on the one machine where the tier is worth
	  anything at all. So
	  the top rung is a partial one: the internal ladder page records
	  what was built,
	  what each machine measured, and why the parts ended differently.
	  
	  Adding a set is a row in the table that answers every question
	  about vector
	  routines, so the installed header does not change.
	  
	  Measured and reported by Tobias Damisch, Elio Blanca and Todd
	  Zullinger.

2026-07-30 19:03  aleidinger

	* doc/html/detailed.html, doc/html/history.html, doc/man/lame.1,
	  frontend/parse.c, libmp3lame/lame.c:
	  Add --vector, and deprecate --noasm in favour of it
	  
	  --noasm can only take capability away: it says which
	  instruction-set families
	  the encoder may not use, and the answer to "run this encode on
	  SSE2 although
	  the processor also has AVX2" has to be assembled out of negations
	  - which stops
	  working the moment a machine has a set the option cannot name.
	  
	  --vector names the set to run instead. auto is the default and
	  keeps today's
	  behaviour, none is the plain C code, and a name pins the encode
	  to that set.
	  The names come from the library rather than from a list in the
	  frontend, so
	  --longhelp offers exactly what this binary can honour on this
	  processor, and a
	  name it cannot honour is refused rather than quietly replaced by
	  another - a
	  measurement that reports on a set nobody asked for is worse than
	  one that does
	  not run.
	  
	  Asking for none also leaves the processor unasked about its
	  capabilities,
	  nothing depending on the answer any more. For the users the
	  manual sends to
	  this option - the ones whose processor cannot survive the
	  question - that is
	  the whole of its value, so it is a documented property of none
	  rather than an
	  undocumented consequence of some flag being cleared elsewhere.
	  
	  --noasm keeps working and says what to use instead. Its mapping
	  is exact, so no
	  existing command line changes what it encodes: sse became none,
	  avx2 became
	  sse2, and mmx and 3dnow selected nothing to begin with. Where
	  both options
	  appear, --vector decides.

2026-07-30 19:02  aleidinger

	* doc/html/history.html, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c,
	  libmp3lame/util.c, libmp3lame/util.h, test/unit/test_set_get.c:
	  Report the vector routines the encoder can actually use
	  
	  The configuration report listed processor features rather than
	  choices. It
	  named MMX, 3DNow! and SSE, none of which this library selects on
	  - the
	  vectorized routines are compiled for SSE2 - so the report could
	  not be
	  reconciled with what an encode did, and the line naming the set
	  actually used
	  was printed only when there was one, where its absence could not
	  be told apart
	  from a version that never printed it.
	  
	  The feature line now lists only the instruction sets this build
	  carries
	  routines for and this processor can run, which is exactly the
	  range a
	  selection may name, and the line reporting the outcome always
	  follows it. The
	  two together answer both halves of the question a benchmark run
	  asks: what
	  could have been used, and what was.
	  
	  MMX and 3DNow! had no code behind them and now have no state
	  either.
	  lame_set_asm_optimizations() refuses those two families and
	  lame_get_asm_optimizations() reports them as not enabled, rather
	  than storing
	  a flag whose only possible effect was to change a report; --noasm
	  mmx and
	  --noasm 3dnow are inert for the same reason. No encode changes.

2026-07-30 19:02  aleidinger

	* include/lame.h, include/libmp3lame.sym, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c,
	  libmp3lame/util.c, libmp3lame/util.h, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_vector_routines.c:
	  Add an API for selecting which vector routines the encoder runs
	  
	  The existing lame_set/get_asm_optimizations() pair names x86
	  instruction-set
	  families, and that has become the wrong shape for the question. A
	  family is
	  not what the library selects: the routines are compiled for SSE2,
	  while SSE
	  there is the group switch for all of them rather than a peer of
	  AVX2. Naming
	  each further instruction set, on this architecture or another,
	  would mean a
	  new value in a public enum in the installed header every time -
	  and none of
	  that lets a caller say "run this encode on SSE2 although the
	  processor also
	  has AVX2", which is what measuring one set against another needs.
	  
	  So the set of names becomes data rather than an enum. Four calls
	  report how
	  many sets this build compiled in and what they are called, select
	  one of them
	  by name, and report which one an instance ended up running.
	  Adding a set is a
	  table row; callers and the header are untouched.
	  
	  Names are the real instruction set, lowercase, and the displayed
	  spelling is
	  that name upper-cased, so there is one string per set and the two
	  cannot
	  drift. The same table answers the enumeration, the name lookup,
	  the selection
	  and the configuration report, because keeping those answers in
	  separate places
	  is how a build comes to advertise a set the encoder never
	  dispatches.
	  
	  A request that cannot be honoured is refused where it is made
	  rather than
	  quietly downgraded at initialisation: three distinct failures
	  separate a
	  misspelt name from a set this build lacks from one the processor
	  cannot
	  execute, since those ask for three different things from whoever
	  hit them.
	  
	  The default selects the widest set the processor offers, which is
	  what the
	  library did before, so an encode that asks for nothing is
	  unchanged.
	  
	  The older pair stays exported and keeps working; it is documented
	  as
	  deprecated. MMX and 3DNow! are called out further, because this
	  library has
	  had no code for either in years and they now select nothing at
	  all.

2026-07-30 19:01  aleidinger

	* Makefile.am, Makefile.in:
	  Make the bitstream test encode with the library it just built
	  
	  The test target encodes a reference file with the frontend from
	  this build and
	  counts how many bytes the result differs from the committed
	  reference. It runs
	  the frontend through libtool's wrapper, and on the systems where
	  that wrapper
	  is an executable rather than a shell script it does nothing to
	  the library
	  search path - so the frontend starts against whatever libmp3lame
	  the host has
	  installed.
	  
	  For this target that is not a cosmetic difference. The number it
	  prints exists
	  to tell a developer whether their change moved the encoder's
	  output; produced
	  by a different library, it describes that library instead, and a
	  build with no
	  change at all can report thousands of differing bytes. Where the
	  installed
	  library is old enough not to export everything the frontend
	  needs, the encode
	  does not run at all and the comparison is made against a file
	  that was never
	  written.
	  
	  Both rules that run a binary from this build now name the build's
	  own library
	  directory, which is what the unit tests were given for the same
	  reason. On the
	  systems where libtool already handled it nothing changes.

2026-07-28 18:40  aleidinger

	* test/unit/test_id3tag.c:
	  Decide a test's skip at compile time, not at run time
	  
	  One id3tag test declares a sample count large enough to drive the
	  play-length
	  field past a 32-bit millisecond value. That count does not fit an
	  unsigned
	  long everywhere, and the test knew it: it asked at run time and
	  skipped itself
	  where the type is narrower.
	  
	  Asking at run time is too late. The constant is still compiled,
	  so on a
	  platform with a 32-bit unsigned long it is truncated on the way
	  in and the
	  compiler says so - a warning on every build of the suite there,
	  about a value
	  the guard guarantees is never used. Moving the guard into the
	  preprocessor
	  keeps the behaviour identical on both kinds of platform and
	  leaves nothing for
	  the narrow one to complain about.
	  
	  No library code changes, and the test's result is unchanged
	  everywhere: it runs
	  where the value is representable and skips where it is not,
	  exactly as before.

2026-07-28 18:40  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in:
	  Make the unit tests load the library they just built
	  
	  Every test in test/unit links libmp3lame, and on a machine that
	  already has
	  libmp3lame installed some of them were loading the installed copy
	  instead of
	  the one the build had just produced. A test suite that reports on
	  a library
	  nobody under test asked for is worse than one that fails: where
	  the installed
	  version happens to be close enough, the run is green and says
	  nothing.
	  
	  This was not visible on Unix, and the reason is worth stating
	  because it is
	  why nobody had noticed. Libtool's wrapper there is a shell script
	  and exports
	  a library search path pointing into the build tree, so the right
	  thing happens
	  by accident of the platform. On Windows libtool produces a
	  wrapper executable
	  instead, which does nothing equivalent, and the loader takes the
	  first match on
	  the ordinary search path - the installed one. Seven of the
	  fourteen programs
	  were running against a library several releases old; six failed
	  to start at
	  all, and one produced assertion failures that looked like defects
	  in this tree.
	  
	  Naming the build's own directory in the test environment makes
	  the answer the
	  same everywhere rather than a property of which wrapper libtool
	  chose. Both
	  search-path variables are set: the two families of systems
	  consult different
	  ones, and neither is harmful where it is not used.
	  
	  No library or test code changes.

2026-07-28 18:39  aleidinger

	* test/unit/test_id3tag.c:
	  Test the id3tag calls that select, reset and enumerate
	  
	  Second of the passes over the exported symbols no unit test
	  reached.
	  The existing id3tag tests cover the field setters across the
	  three text
	  encodings; what they never touch is the other half of the API -
	  the
	  calls that choose a tag version or a text encoding, reset the tag
	  state,
	  pad it, enumerate the genre list, or set the track number.
	  
	  The three deprecated UCS-2 setters are documented as aliases for
	  their
	  UTF-16 counterparts, so each is checked by building the same tag
	  twice,
	  once through the alias and once through its target, and comparing
	  the
	  two byte for byte. Looking for the frame instead would pass on an
	  alias
	  wired to the wrong function, which is the only way an alias can
	  be
	  wrong.
	  
	  Where a call has a documented consequence beyond its own field,
	  that
	  consequence is what is asserted: selecting UTF-8 is checked by
	  reading
	  the version out of the tag header, the space-padding option by
	  counting
	  the filler bytes of a known-width ID3v1 field against a build
	  without
	  it, and the padding shortcut by requiring it to produce the same
	  tag as
	  the general call with the size it documents.
	  
	  The genre enumeration is asserted to be a permutation - every
	  genre
	  reported exactly once - rather than merely the right length,
	  which a
	  list that repeated one entry and skipped another would also
	  satisfy.
	  
	  No library code changes.

2026-07-28 18:38  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_lib_info.c, test/unit/test_set_get.c:
	  Test the version, MPEG-table and alias entry points
	  
	  Comparing the library's export list against the names the
	  unit-test
	  suite mentions leaves 63 of 242 exported symbols reached by no
	  test at
	  all. This covers the cheapest part of that gap: everything that
	  reports
	  a property of the build rather than of an encoder, and needs no
	  encoder
	  instance to call.
	  
	  The version tests are written against the API rather than against
	  the
	  version macros. Each string form is built at its own site, so
	  checking
	  the strings against the numerical form asks whether they agree -
	  which a
	  version bump that updates one site and not another can answer
	  wrongly,
	  where a test that rebuilt the macro expressions could not. Beyond
	  the
	  version number itself the strings are documented as unspecified,
	  and
	  nothing here asserts more than that; the one exception is the
	  compact
	  form, whose layout is fixed because callers copy it into
	  fixed-size
	  fields.
	  
	  The bitrate and sample-rate accessors exist so that a caller of
	  the
	  shared library can read tables that are not themselves exported.
	  Their
	  values come from the MPEG specification and cannot change, so
	  they are
	  asserted exactly, along with the range checking the accessors
	  add.
	  
	  Three entry points of the parameter API were missed by the
	  earlier pass
	  over it because they are not plain pairs: the announced input
	  length,
	  whose "length not known" default a caller has to be able to
	  recognize,
	  and the two deprecated aliases that combine or forward to the
	  decode-while-encoding settings. Whether those can be switched on
	  at all
	  depends on the build, so the tests ask the library instead of
	  assuming.
	  
	  No library code changes.

2026-07-28 18:38  aleidinger

	* doc/maintainer-abi.md, maintainer/abi/libmp3lame.abi:
	  Keep the build machine out of the committed ABI baseline
	  
	  The baseline ships in the release tarball so that a build from it
	  can run the
	  ABI check, but it recorded a good deal about the machine that
	  produced it: the
	  dynamic dependency list, declarations of libc functions the
	  library only calls,
	  and types that reached the dump through some other library's
	  headers. On a
	  glibc host built with the project's own settings that dependency
	  list names
	  libmvec, which does not exist on FreeBSD, so the shipped file
	  described one
	  operating system rather than the interface.
	  
	  It is now generated with the interface alone in it. Nothing about
	  what the
	  check can detect changes: a build missing an exported function is
	  still
	  reported as missing it, on either compiler tried.
	  
	  Two consequences worth having in the guide, both of which had
	  been learned the
	  hard way and written down nowhere. The compilation flags no
	  longer affect the
	  result, so the recipe no longer depends on a detail that quietly
	  changed
	  meaning when a user's CFLAGS began adding to the project's own
	  rather than
	  replacing them. And the file is architecture-specific while being
	  deliberately
	  compiler-independent, so the guide now says which of the three
	  checks to
	  disregard when running on an architecture other than the one the
	  baseline was
	  generated for.
	  
	  The guide also now states the rule that made all of this
	  confusing to work out:
	  a regeneration is read with abidiff, never with diff. The two
	  dumps of one
	  unchanged interface can differ by thousands of lines of XML and
	  mean nothing at
	  all.

2026-07-28 18:37  aleidinger

	* doc/html/history.html, include/lame.def, include/lame.h,
	  include/libmp3lame.sym, libmp3lame/set_get.c,
	  maintainer/abi/libmp3lame.abi, test/unit/test_set_get.c:
	  Add lame_get_asm_optimizations()
	  
	  lame_set_asm_optimizations() allows or forbids an instruction-set
	  family, and
	  nothing could read that setting back: neither what a preset or
	  another part of
	  a program had left behind, nor whether a family a caller named
	  was one this
	  library knows. The setter cannot answer the second question at
	  all - it accepts
	  an unrecognized family, sets nothing, and returns the value it
	  was given, which
	  is indistinguishable from success.
	  
	  The new getter answers for one family at a time, the same shape
	  the setter
	  takes, and distinguishes four answers: the family is allowed, it
	  has been
	  forbidden, the instance is not usable, or the family is not
	  available in this
	  instance. It reports the setting rather than the instruction set
	  an encode will
	  really use: the families are not independent, so forbidding SSE
	  also rules out
	  AVX2 while the AVX2 setting stays as it was.
	  
	  This adds an exported symbol and the ABI baseline records it. No
	  existing
	  behaviour changes, and no soname bump is needed for an addition.

2026-07-28 18:37  aleidinger

	* doc/html/history.html, libmp3lame/lame.c, libmp3lame/set_get.c,
	  test/unit/test_set_get.c:
	  Report the adaptive ATH scheme that is actually in use
	  
	  lame_get_athaa_type() answered "nothing has been chosen" for the
	  whole life
	  of an encoder that had not set the scheme itself. Initialization
	  resolved
	  the default into internal state without recording it on the
	  instance, so
	  the scheme an encode was actually using could not be read at all.
	  
	  The resolved scheme is now stored where the getter can see it.
	  Which scheme
	  is chosen stays LAME's own business and may differ between
	  releases, so the
	  documentation says to ask rather than to assume a number. Every
	  neighbouring
	  setting sharing this "not chosen yet" default already resolved
	  this way;
	  this brings the last one into line. The encode itself is
	  unchanged.

2026-07-28 13:39  aleidinger

	* doc/doxygen/DoxygenLayout-internal.xml:
	  doc: stop listing two internal documentation pages twice
	  
	  The navigation panel of the internal documentation showed the
	  maintainer
	  tool collection and the frame analyzer page twice each: once as
	  an entry
	  of its own with nothing beneath it, and once a few rows further
	  down
	  with its sections. The two rows carried different capitalisation,
	  which
	  is what gave it away.
	  
	  A page given its own navigation entry is also part of the general
	  page
	  list, and the panel renders both. Dropping the two dedicated
	  entries
	  leaves each page listed exactly once, still at the top level and
	  still
	  with its sections, so nothing became harder to find. An entry
	  pointing
	  at a source file rather than at a page does not double up, which
	  is why
	  the public reference was unaffected.

2026-07-28 13:20  aleidinger

	* doc/maintainer-gen-api-docs.md, maintainer/check-dist.sh,
	  maintainer/gen-api-docs.sh:
	  build: ask the documentation check what was documented, not which
	  pages exist
	  
	  A doxygen run that reads no input still writes a complete set of
	  stylesheets, scripts and images and exits successfully. The
	  result looks
	  entirely healthy and contains no documentation at all, so both
	  the
	  publishing script and the distribution check refuse to accept a
	  set that
	  shows no sign of having parsed anything.
	  
	  They recognised that case by looking for one particular generated
	  page.
	  Which pages the renderer emits follows the settings it was run
	  with
	  rather than the content, so scoping the public reference away
	  from
	  verbatim header copies removed the page they looked for, and both
	  started rejecting a perfectly good set. They now require a
	  function from
	  the installed header to be documented somewhere in each set,
	  which is
	  the question they were meant to be asking and does not depend on
	  how the
	  pages happen to be laid out.
	  
	  The distribution check also lints the frame analyzer's manual
	  page. It
	  had been checking the other two since they existed, and the new
	  one was
	  never added to the list.

2026-07-28 12:20  aleidinger

	* doc/Doxyfile.in:
	  doc: stop publishing the internal headers in the API reference
	  
	  The public reference carried a verbatim copy of every header the
	  documentation parses - the psychoacoustic model, the quantizer,
	  the bit
	  reservoir and the rest - and listed all of them in its file
	  index, as
	  though a caller could use any of it. What a caller may use is the
	  one
	  installed header, and describing that is what this document is
	  for.
	  
	  The source is unaffected and stays where it has always been, in
	  the
	  release tarball and the repository. The internal documentation
	  goes on
	  carrying the copies, which is where reading a header beside its
	  description is useful.

2026-07-28 12:16  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in, configure,
	  configure.ac, doc/Doxyfile-internal.in, doc/Doxyfile.in,
	  doc/Makefile.am, doc/Makefile.in,
	  doc/doxygen/DoxygenLayout-internal.xml, doc/doxygen/mainpage.md,
	  doc/doxygen/mp3x-internals.dox, doc/html/Makefile.in,
	  doc/html/history.html, doc/man/Makefile.am, doc/man/Makefile.in,
	  doc/man/mp3x.1, dshow/Makefile.in, frontend/Makefile.in,
	  frontend/mp3x_core.c, frontend/mp3x_core.h, frontend/mp3x_plot.c,
	  frontend/mp3x_plot.h, frontend/mp3x_session.c,
	  frontend/mp3x_session.h, frontend/mp3x_ui.c, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in,
	  test/Makefile.in, test/unit/Makefile.in, vc_solution/Makefile.in:
	  doc: document the frame analyzer
	  
	  mp3x had no user documentation at all. What each graph shows,
	  what the
	  display controls change and which keys drive the transport
	  existed only
	  as comments in the sources, and the keyboard was written down
	  nowhere.
	  It now has a manual page, installed where the program itself is
	  and
	  distributed either way, so a release carries it for someone who
	  configures the analyzer in later.
	  
	  The analyzer's own interfaces are documented for people working
	  on it,
	  in the internal documentation set. The engine, the plotting layer
	  and
	  the per-file session already carried careful comments, but not in
	  a form
	  the documentation build publishes, so none of it reached either
	  document. A companion page collects the ownership, transport and
	  end-of-input rules that span all four implementation files, along
	  with
	  the test-only instrumentation and the regression routine that had
	  lived
	  in a plain text file beside them.

2026-07-27 19:27  aleidinger

	* configure, configure.ac:
	  build: locate iconv with the flags the frontend is built with
	  
	  The frontend's iconv check ran before any of the optional
	  libraries had been
	  located, so it tested the C library's <iconv.h> without the
	  include paths the
	  frontend is actually compiled with. Where those paths lead to a
	  GNU libiconv
	  installed beside the C library's iconv, its header renames
	  iconv_open to
	  libiconv_open - so the check answered for a different header than
	  the build
	  uses, reported that no separate library was needed, and the
	  frontend then
	  failed to link on an undefined libiconv_open.
	  
	  Whether it failed was decided by the installation prefix, because
	  the library
	  search only found that libiconv when the prefix being configured
	  happened to be
	  the one it was installed under. Configuring into any other prefix
	  - which is
	  what make distcheck does - could not link the frontend at all,
	  while the
	  default prefix concealed the problem entirely.
	  
	  The check now runs last, with the include and library search
	  paths pkg-config
	  reported for GTK, the decoder and libsndfile. They are taken from
	  what
	  pkg-config said rather than assumed, since an installation prefix
	  is
	  configurable and not always the conventional one. Only the -I and
	  -L options
	  are carried over: the link options are written for libtool, which
	  links the
	  frontend, and include flags that a bare compiler rejects
	  outright. Where no
	  such paths exist nothing is added and the check behaves exactly
	  as before -
	  nothing can bring in a foreign <iconv.h> without also saying
	  where its library
	  lives.
	  
	  The configuration summary now reports iconv as well. Tag
	  transliteration is
	  compiled out when iconv is unavailable, so an unnoticed negative
	  result would
	  silently drop a feature instead of failing the build.

2026-07-27 19:26  aleidinger

	* maintainer/check-dist.sh:
	  maintainer: fix the tarball-name check on a non-GNU sed
	  
	  The distribution check compares the version in the tarball's name
	  against the
	  one configure declares and the one version.h carries. It stripped
	  the archive
	  suffix with a basic regular expression using \| alternation,
	  which is a GNU
	  extension: on a BSD sed the suffix was never stripped, the
	  comparison could
	  not match, and the check reported the tarball as misnamed. It has
	  done that on
	  every FreeBSD run it has ever made, while passing on Linux, which
	  is why the
	  mistake survived - the platform where it works is the one it was
	  written on.
	  
	  One plain expression per suffix does the same job in a basic
	  regular
	  expression on either sed.

2026-07-27 19:26  aleidinger

	* maintainer/gen-build-matrix.ps1:
	  build: stop handing the 32-bit decoder to a 64-bit matrix cell
	  
	  The Windows matrix generator looks the optional decoder up once,
	  for the nmake
	  cells, which are 32-bit. It then passed that same path to the
	  MSBuild decoder
	  cell, which is built for the selected platform - by default x64.
	  On a machine
	  holding only the 32-bit library that cell could not resolve a
	  single decoder
	  symbol, and the matrix reported a failure that was its own doing
	  rather than
	  the tree's.
	  
	  The project file already resolves the decoder per platform, so
	  the cell only
	  has to turn the option on and leave the path alone, exactly as
	  the libsndfile
	  cell beside it does. It is now emitted only when the library for
	  that platform
	  is actually present, so a missing one drops the cell instead of
	  failing it.

2026-07-27 19:25  aleidinger

	* doc/maintainer-build-matrix.md, maintainer/gen-build-matrix.sh:
	  build: cover the frame analyzer in the configuration matrix
	  
	  The matrix built every optional surface except one: no cell
	  passed
	  --enable-mp3x, so the analyzer frontend was never compiled or
	  linked by it on
	  any platform. A change to those sources could pass the entire
	  matrix on every
	  compiler and still not build.
	  
	  There is a cell for it now, gated on GTK 4 the way the sndfile
	  cell is gated
	  on libsndfile, and probed with the same version constraint
	  configure applies,
	  so a GTK 4 too old to build the analyzer counts as absent here
	  rather than
	  producing a cell that cannot configure. Where GTK 4 is missing
	  the cell is
	  skipped with that reason recorded, which is the honest report: on
	  such a host
	  the analyzer cannot be covered at all.

2026-07-27 19:25  aleidinger

	* doc/html/history.html:
	  Changelog entries for the GTK4 analyzer and the MinGW Unicode
	  fixes
	  
	  Patches by Maya R. Odinezenko, SF patch #104.

2026-07-27 19:25  aleidinger

	* INSTALL, Makefile.MSVC, Makefile.am, Makefile.in, README.WINGTK,
	  acinclude.m4, configure, configure.ac,
	  doc/maintainer-build-matrix.md, doc/maintainer-coverage.md,
	  frontend/Makefile.am, frontend/Makefile.in,
	  frontend/README.mp3x-gtk4, frontend/gpkplotting.c,
	  frontend/gpkplotting.h, frontend/gtkanal.c, frontend/gtkanal.h,
	  frontend/mp3x.c, maintainer/gen-build-matrix.ps1,
	  maintainer/gen-coverage-matrix.sh,
	  maintainer/setup-windows-deps.ps1, vc_solution/.gitignore,
	  vc_solution/README.vs.txt, vc_solution/vs_gtk_config.props,
	  vc_solution/vs_lame.slnx, vc_solution/vs_lame_mp3x.vcxproj,
	  vc_solution/vs_lame_mp3x.vcxproj.filters:
	  build: retire the GTK 1.2 frame analyzer
	  
	  The frame analyzer is a GTK 4 program now. Its GTK 1.2
	  predecessor was still
	  in the tree and still shipped in the release archive while being
	  compiled
	  nowhere, and the build system and the build notes went on
	  describing how to
	  obtain and link GTK 1.2 on three platforms. None of it is
	  reachable any more,
	  so the sources, the build support and the notes go together
	  rather than
	  documenting something that is not there.
	  
	  On Windows the analyzer is a Visual Studio target, built for x64
	  alone: GTK 4
	  for MSVC is published for that architecture and for no other, so
	  a 32-bit
	  analyzer would have nothing to link against. That also takes it
	  out of the
	  nmake makefile, whose native target is 32-bit x86; the makefile
	  names where to
	  build it instead. The Visual Studio project finds GTK 4 the same
	  way it finds
	  mpg123 and libsndfile - a per-platform folder under vc_solution -
	  rather than
	  at a fixed path on a fixed drive.
	  
	  On the autotools side the bundled macro that located GTK 1.2
	  through the
	  obsolete gtk-config script has no caller left, so it goes, and
	  with it the
	  three configure options that came with it - --with-gtk-prefix,
	  --with-gtk-exec-prefix and --disable-gtktest - which described a
	  search that
	  no longer happens and would have gone on being offered in
	  configure --help.
	  The coverage harness was probing for that same gtk-config script
	  to decide
	  whether the analyzer could be covered, so it was answering a
	  question nothing
	  asks any more; it now asks pkg-config for the GTK 4 that
	  configure requires.

2026-07-27 19:24  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in, configure,
	  configure.ac, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.am,
	  frontend/Makefile.in, frontend/README.mp3x-gtk4,
	  frontend/mp3x_core.c, frontend/mp3x_core.h, frontend/mp3x_plot.c,
	  frontend/mp3x_plot.h, frontend/mp3x_session.c,
	  frontend/mp3x_session.h, frontend/mp3x_ui.c, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in,
	  test/Makefile.in, test/unit/Makefile.in, vc_solution/Makefile.in:
	  frontend: replace the mp3x frame analyzer with a GTK4
	  implementation
	  
	  The frame analyzer was written against GTK 1.2, which has not
	  been
	  available on current systems for many years, so in practice mp3x
	  could
	  no longer be built. This is a reimplementation on GTK 4, keeping
	  the
	  program name, the command line, the analyzer data and all eight
	  plots,
	  and adding what a desktop program is expected to have: opening
	  and
	  replacing a file from within the program, a recent-files list,
	  and PNG
	  export of the graphs.
	  
	  Frame processing, session lifetime, plotting and the user
	  interface are
	  separated so that replacing the open file and shutting down have
	  clear
	  ownership boundaries. The analyzer engine itself stays toolkit
	  independent.
	  
	  mp3x now requires GTK 4.10 and a decoder-enabled build, and GTK
	  is
	  probed with pkg-config rather than the obsolete gtk-config
	  script. At
	  end of input the delayed encoder and decoder output is drained,
	  so the
	  final frames are analyzed rather than dropped.
	  
	  The GTK 1.2 sources are no longer compiled but stay in the
	  release
	  archive.
	  
	  Patch by Maya R. Odinezenko, SF patch #104.

2026-07-27 19:23  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in, configure,
	  configure.ac, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.am,
	  frontend/Makefile.in, frontend/main.c, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in,
	  test/Makefile.in, test/unit/Makefile.in, vc_solution/Makefile.in:
	  frontend: handle Unicode command lines and paths on MinGW
	  
	  MinGW builds of the command line tools lose characters that are
	  not in
	  the active code page. The narrow entry point receives an already
	  lossy command line, and filenames are handed to the narrow
	  fopen(),
	  so a path the user can see in Explorer may be unopenable.
	  
	  Use the Windows wide-character entry point for the MinGW frontend
	  programs, convert the UTF-16 argument array to UTF-8 before the
	  common
	  frontend sees it, and open files through the existing
	  wide-character
	  helpers. -municode is applied only when linking the frontend
	  programs;
	  a library-only build needs no Windows startup object, and a MinGW
	  frontend build whose toolchain cannot provide that startup is
	  refused
	  rather than silently built without it.
	  
	  The UTF-8 to UTF-16 conversion is also made strict: an invalid
	  sequence
	  is now an error instead of being silently replaced. This affects
	  the
	  Microsoft toolchain too, which already used the wide entry point.
	  
	  The wide entry point and the strict conversion are one change and
	  not
	  two: with the strict conversion in place but the narrow entry
	  point
	  still selected, MinGW would reject the very command line it is
	  given.
	  
	  MSVC builds and every non-Windows platform are unaffected - the
	  added
	  code is inside the existing _WIN32 branch and the added configure
	  logic inside a MinGW host case.
	  
	  Patch by Maya R. Odinezenko, SF patch #104.

2026-07-27 19:22  aleidinger

	* doc/html/history.html:
	  Changelog entry for the Windows pipe decoding fix
	  
	  Patch by Maya R. Odinezenko, SF patch #107.

2026-07-27 19:21  aleidinger

	* frontend/get_audio.c:
	  frontend: report non-seekable Windows handles to mpg123
	  
	  MP3 input from a genuine Windows pipe can silently produce
	  incorrect
	  decoded output. In native cmd.exe pipelines, the affected
	  frontend
	  produced 182,204 bytes instead of 176,400 and exited with status
	  0;
	  under MSYS2 it failed with a nonzero status. The cmd.exe pipeline
	  output
	  matched decoding the same audio encoded without a Xing/Info
	  frame, so
	  encoder-delay and padding trimming had been lost.
	  
	  LAME's Windows FILE * reader bridge installs its seek callback
	  with
	  mpg123_replace_reader_handle(). mpg123_open_handle() reaches
	  get_fileinfo(), whose first operation seeks to the end to
	  determine the
	  stream length and inspect ID3v1. On Windows the fseek() return
	  value is
	  undefined for a device that cannot seek. On a pipe it reported
	  success,
	  so mpg123 classified the stream as seekable, read 128 bytes for
	  the
	  ID3v1 check, and could not return to the beginning. The value
	  reported
	  by ftell() reflects bytes buffered in the pipe rather than a
	  stream
	  length, so the failure mode depends on pipe occupancy.
	  
	  Classify the native handle with GetFileType() before seeking.
	  Keep the
	  existing fseek() path only for FILE_TYPE_DISK and report other
	  handles
	  as non-seekable without reading or repositioning the stream.
	  
	  Ordinary pathnames and standard input redirected from a regular
	  file
	  are disk handles and keep the existing path. Non-Windows builds
	  are
	  unaffected because the added code remains within the existing
	  _WIN32
	  branch.
	  
	  With the patch, lame produced identical decoded PCM for pathname,
	  redirected regular-file standard input, and genuine-pipe routes
	  on
	  macOS, Debian, Windows UCRT64, and Windows MINGW64/MSVCRT. On
	  Windows,
	  the callback consumed zero bytes before recognizing the pipe as
	  non-seekable.
	  
	  Patch by Maya R. Odinezenko, SF patch #107.

2026-07-27 14:58  aleidinger

	* doc/maintainer-abi.md, doc/maintainer-gen-api-docs.md,
	  doc/maintainer-quality.md:
	  doc: correct and complete the maintainer harness documentation
	  
	  The ABI page said libabigail is unavailable on FreeBSD, which is
	  wrong -
	  it is in the ports tree. It also explained the committed baseline
	  in
	  terms of two particular releases, which reads as if the baseline
	  goes
	  stale on its own; it does not, because it records the last
	  intentional
	  ABI change and everything since is measured against that. Say so
	  once,
	  without naming versions, and say plainly that deciding a change
	  is an
	  intentional ABI change is the maintainer's call rather than
	  something a
	  red run authorises.
	  
	  The quality page described only the reference test set, which is
	  isolated instruments and test signals. That is the right material
	  for
	  finding where an encoder breaks and the wrong material for
	  judging
	  whether a listener would notice, so a quality claim is only as
	  broad as
	  the corpus behind it. Describe the real-music set used alongside
	  it,
	  what each of the two answers, and why the setup script does not
	  and
	  cannot fetch the second one.
	  
	  Also drop the list of which website pages link to which
	  documentation
	  set from the publishing page - it describes the site, not this
	  tree, and
	  goes stale there.

2026-07-27 14:58  aleidinger

	* doc/Doxyfile-internal.in, doc/Doxyfile.in, doc/Makefile.am,
	  doc/Makefile.in, doc/doxygen/DoxygenLayout-internal.xml,
	  doc/doxygen/DoxygenLayout.xml, doc/doxygen/maintainer.md,
	  include/lame.h, libmp3lame/quantize_pvt.h, libmp3lame/util.h:
	  doc: make the generated API reference navigable and public-only
	  
	  The public documentation set had no entry point. Nothing declared
	  in
	  lame.h was published at all, because the header carried no
	  file-level
	  Doxygen block, so the reference was assembled from the library
	  sources
	  that happened to have one. The API was reachable only by opening
	  those
	  files one at a time, or through an alphabetically split index of
	  global
	  symbols named after the implementation.
	  
	  lame.h now documents itself, which gives the whole interface on
	  one
	  page, and that page is named in the navigation instead of sitting
	  among
	  implementation files. The include lists shown on those pages are
	  dropped
	  from the public set: they name headers that are neither installed
	  nor
	  usable by a caller. They stay in the internal set, where they are
	  useful.
	  
	  Four internal types were also being published as if they were
	  part of
	  the interface, because they carry a documentation comment and the
	  public
	  configuration could only filter out undocumented ones. They are
	  now
	  marked internal, like the internal functions around them.
	  
	  The development harnesses get their own place in the internal set
	  rather
	  than being scattered through the page list, with a short parent
	  page
	  explaining what each is for.

2026-07-27 14:57  aleidinger

	* configure, configure.ac:
	  configure: skip the PIE frontend flags when there is no shared
	  library
	  
	  The default hardening builds the command line tools as
	  position-independent executables. That needs position-independent
	  objects to link against: it works when the tools link the shared
	  library, and cannot work when they link the static one, whose
	  objects
	  are not built PIC.
	  
	  The guard only recognised --disable-dynamic-frontends, which asks
	  for a
	  statically linked frontend explicitly. Configuring
	  --disable-shared
	  arrives at the same place implicitly - there is no shared library
	  left
	  for the tools to link - and they were still asked to be
	  position-independent. On a toolchain that does not default to PIE
	  that
	  combination fails to link. Skip the flags in both cases.

2026-07-27 12:55  aleidinger

	* maintainer/abicheck.sh, maintainer/check-dist.sh,
	  maintainer/coverage-mkinputs.py, maintainer/coverage-report.py,
	  maintainer/coverage-run.sh, maintainer/gen-api-docs.sh,
	  maintainer/gen-coverage-matrix.sh, maintainer/perf-compare.sh,
	  maintainer/quality-compare.py,
	  maintainer/setup-quality-corpus.sh:
	  set executable

2026-07-27 11:37  aleidinger

	* doc/html/history.html, libmp3lame/mpglib_interface.c:
	  fix: refuse to create a decoder when the library cannot decode
	  
	  A libmp3lame built without libmpg123 cannot decode anything, and
	  only one of
	  the two calls that create a decoder said so. The other returned a
	  decoder
	  that failed on every use, so the absence surfaced at the first
	  piece of audio
	  rather than at the point it could be acted on - and there it
	  looks like a
	  problem with the input, not with the build. Both calls now return
	  nothing,
	  which a caller can check before reading any data.
	  
	  The failure a caller sees is unchanged in kind, only in timing: a
	  program that
	  already checks what it gets back needs no change, and one that
	  does not fails
	  at the same call it failed at before. Nothing changes in a build
	  that has the
	  decoder, which is every default build.
	  
	  The path that reported the absence was also leaking the instance
	  it had just
	  allocated; it is freed now, in both calls.

2026-07-27 11:36  aleidinger

	* doc/doxygen/internal.md, doc/doxygen/vector-dispatch.dox:
	  Record why the vector dispatch ladder stops at AVX2
	  
	  The internal dispatch page explains the two rungs that are
	  missing for
	  counter-intuitive reasons. A third belongs beside them now: an
	  AVX-512
	  tier above AVX2 was written, taken to hardware that has the
	  instruction
	  set, and measured end to end - and it is not worth carrying.
	  
	  Three machines reported. On Zen 4, constant bitrate at 320 kbit/s
	  - the
	  case with the most for such a tier to win - came out 1.0% slower
	  than
	  the AVX2 tier, with the runs each way not overlapping, and the
	  same
	  result came back from a second run on a different input file. On
	  Zen 5
	  the same comparison returned equal figures to three decimals over
	  twenty-five runs each way. Variable bitrate was a wash
	  everywhere. An
	  Ice Lake laptop reported a 7.6% gain, but from three sequential
	  runs per
	  arm on a thermally constrained part, and its own data shows
	  roughly a
	  fifth of that as drift between two runs of identical code; the
	  page
	  records it as the one measurement still outstanding rather than
	  as a
	  result.
	  
	  The kernels produced bit-identical output, so this is a statement
	  about
	  the width and not about the code. It is also not the statement
	  everyone
	  expected about double-pumped datapaths: Zen 5 does not carry that
	  restriction and still measured zero. What binds is the work
	  itself -
	  loops that are already gather-bound, or bound by a fixed
	  reduction
	  order, stay that way when the registers get wider. Widening the
	  one
	  routine whose reduction order could have been changed cost it
	  exact
	  agreement with the existing tier on both machines that ran the
	  tests,
	  which is a correctness answer rather than a speed one.
	  
	  This goes in the ladder page rather than only in the commit
	  history
	  because a negative result nobody wrote down is a negative result
	  someone
	  re-derives. The section records what was tested and on which
	  hardware,
	  not just the conclusion, and states what would reopen the
	  question.
	  
	  Measured and reported by Tobias Damisch, Elio Blanca and Todd
	  Zullinger.

2026-07-27 11:36  aleidinger

	* frontend/get_audio.c, libmp3lame/bitstream.c,
	  libmp3lame/gain_analysis.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/set_get.c:
	  build: make the non-default build configurations warning-free too
	  
	  The tree's warning-free state was established in one build
	  configuration.
	  Measured across the configuration matrix it did not hold in three
	  others:
	  a build without the decoder, a build without the analyzer hooks,
	  and -
	  in every cell, including the default one - any GCC-family
	  compiler on
	  Windows. None of this is visible from an ordinary Linux build,
	  which is
	  why it survived the earlier cleanup.
	  
	  Three causes, none of them a behaviour change:
	  
	  Parameters that exist only for the decoder or the analyzer are
	  unused
	  exactly when those are configured out. They now carry the
	  project's
	  "may be unused" attribute, which is what the earlier sweep
	  applied to
	  the parameters that were unused in the default build.
	  
	  An MSVC-specific pragma was guarded by a macro that MinGW defines
	  as
	  well, so every GCC or Clang build on Windows parsed a pragma it
	  does not
	  know and said so, twice. It now keys off the MSVC compiler macro,
	  the
	  guard the rest of the tree already uses for the same purpose.
	  
	  The audio reader asked the same question about the input format
	  three
	  times over. One local answers it once, which reads better and
	  also
	  settles a possibly-uninitialized diagnostic that the repetition
	  provoked: the compiler could not know the three answers agree,
	  and in a
	  build without the decoder it saw a buffer that is filled under
	  one of
	  them and read under another.

2026-07-27 11:36  aleidinger

	* test/unit/parse_test_stubs.c:
	  test: stub the frontend helpers the option parser reaches
	  elsewhere
	  
	  Three unit tests compile a frontend source file directly, because
	  the
	  functions they exercise are not visible outside it, and they link
	  that
	  one translation unit against a small set of stand-ins instead of
	  the
	  whole frontend. Several of the functions it calls sit behind
	  conditions
	  the usual build does not meet - a build configured without the
	  decoder,
	  a debug or unoptimized build, Windows - so no stand-in was ever
	  written
	  for them and "make check" fails to link in each of those
	  configurations.
	  
	  In the usual build the suite links only because the compiler
	  discards
	  the one remaining call together with the branch it sits in.
	  Lowering
	  the optimization level is enough to break the test suite, which
	  makes
	  the green result an accident of how the tree happens to be built
	  rather
	  than a property of the tests.
	  
	  The missing stand-ins are supplied, so the suite builds wherever
	  the
	  project does. They are defined unconditionally, and the real
	  prototypes
	  are included rather than repeated: a signature that drifts then
	  fails to
	  compile instead of failing to link, and the test-side file does
	  not have
	  to track conditions that live in the frontend.

2026-07-27 11:35  aleidinger

	* doc/html/history.html, libmp3lame/quantize_pvt.c:
	  fix: make the two ATH-only masking options reach the quantizer
	  again
	  
	  The absolute threshold of hearing is the quiet-room curve, the
	  level below
	  which a tone is inaudible with nothing else playing. In an
	  ordinary encode
	  it is only the floor, and the masking the psychoacoustic model
	  derives from
	  the signal itself does the real work. Two options ask for the
	  floor alone
	  instead - one for every block, one for short blocks only - which
	  is a way
	  to hear what the model contributes, and to compare an encode made
	  without
	  it.
	  
	  Neither had any effect on the encoded audio. The place where the
	  model's
	  threshold is folded in over the floor no longer consulted them,
	  so the two
	  settings survived only in the analyzer's display and in the flag
	  that marks
	  an encode as non-standard. A caller could set either, hear no
	  difference,
	  and have nothing to point at. The fold is now skipped when the
	  corresponding option is set, which is what the analyzer's own
	  handling of
	  the same two settings had gone on doing all along.
	  
	  Both options are off by default and the guards are unreachable
	  without
	  them, so an ordinary encode is untouched: the output is byte for
	  byte
	  identical for CBR and for VBR. With either option set the output
	  changes,
	  which is the point, and was confirmed rather than assumed - a run
	  with each
	  option differs both from the same build without it and from the
	  previous
	  behaviour.

2026-07-27 11:35  aleidinger

	* doc/html/history.html, libmp3lame/set_get.c,
	  test/unit/test_set_get.c:
	  fix: let the temporal masking getter report that nothing has been
	  chosen
	  
	  Several encoder settings are stored with a negative marker
	  meaning "not
	  chosen yet", which the initialization resolves to a real default.
	  Reading
	  one of them back before then is a legal thing for a caller to do,
	  and it
	  is how a caller learns that a setting is still at whatever the
	  encoder
	  will pick.
	  
	  The getter for the temporal masking setting asserted its value
	  was zero or
	  one, which the marker is not, so that call aborted the process
	  instead of
	  answering. Assertions are deliberately left enabled in release
	  builds of
	  this library, so this was not confined to a debugging
	  configuration. The
	  neighbouring setting with the same shape already admits its own
	  marker;
	  this one now does the same.
	  
	  The unit tests gained a case for the whole class: every setting
	  that has
	  such a marker is read on a freshly created encoder, and read
	  again after
	  initialization to confirm the marker does not survive it. The
	  existing
	  tests could not have caught this, because they all set a value
	  before
	  reading it back and so never see the marker. The new case fails
	  against
	  the unfixed source, which is what makes it worth having.
	  
	  Behaviour is otherwise unchanged: the getter returns what it
	  always
	  returned once a value has been chosen, and no encoding path reads
	  it
	  through this function. The encoded bitstream is identical for CBR
	  and
	  VBR.

2026-07-27 11:33  aleidinger

	* libmp3lame/id3tag.c:
	  doc: document the metadata API for the public API reference
	  
	  Completes the public API documentation with the ID3 tagging
	  interface -
	  the last of its files the reference did not cover. That file is
	  described too, so the tagging API has a page of its own.
	  
	  The properties that apply to all of it are stated once: the tags
	  have to
	  be set before the encoder is initialized, since the ID3v2 tag
	  precedes the
	  audio; which tags get written is a choice of its own; and the
	  setters that
	  return nothing cannot report a failure, including a failure to
	  allocate.
	  
	  The recurring hazard in this interface is silent truncation, so
	  each field
	  says what it costs. ID3v1 fields are fixed-length, and text
	  longer than
	  the field is cut without a word - which means the same call can
	  be exact
	  in one tag and lossy in the other. The comment field is the
	  sharpest
	  case: it loses two characters as soon as a track number is set,
	  because
	  those bytes are what carries the track number, so a comment that
	  was
	  accepted can be shortened by a later call about something else.
	  
	  Where a function reports something, the reference says what the
	  report
	  actually means. Setting a track number returns an error for a
	  value ID3v1
	  cannot hold, although the value is written to ID3v2 and the call
	  did what
	  it could. Setting a genre accepts a name outside the standard
	  list
	  without complaint, and replaces a name that is in the list with
	  the list's
	  own spelling. Attaching cover art determines the image format
	  from the
	  data rather than trusting the caller, and refuses anything it
	  does not
	  recognize. The two functions that hand a rendered tag back are
	  documented
	  with the buffer-sizing convention they share, in which the result
	  has to
	  be compared against the size passed in to know whether anything
	  was
	  copied.
	  
	  Documentation only: the comment-stripped source is unchanged byte
	  for
	  byte, so neither the library's behaviour nor its interface is
	  affected.

2026-07-27 11:33  aleidinger

	* libmp3lame/mpglib_interface.c:
	  doc: document the decoding API for the public API reference
	  
	  The library ships a decoder alongside the encoder, because
	  encoding needs
	  one: measuring what a file will sound like means decoding it
	  again. It is
	  exposed for its own sake as a small API of its own, and until now
	  the
	  public reference did not mention it at all. The file it lives in
	  is
	  described too, so the decoder has a page of its own.
	  
	  The entry points are documented with the two properties a caller
	  has to
	  know before anything else. The decoder is a separate object from
	  the
	  encoder, created and destroyed by its own calls. And the decoding
	  itself
	  is done by an optional external library: a build without it
	  cannot decode,
	  and neither of the two calls that create a decoder hands one
	  back, so
	  checking that result is what tells a caller whether decoding is
	  available
	  before any input has been read.
	  
	  Each decode entry point states how much output space it requires,
	  which is
	  where the two families differ and where a caller can go wrong
	  quietly: the
	  one-frame calls need room for a frame, while the convenience call
	  that
	  consumes a whole buffer needs room for every frame in it and
	  cannot check.
	  A return of zero is documented as the normal state at the start
	  of a
	  stream rather than as a failure.
	  
	  The three functions that would route the decoder's diagnostics
	  accept a
	  callback and discard it, so nothing is reported however they are
	  called;
	  they say so. The obsolete entry points from the era when the
	  decoder was
	  a single global object are marked obsolete and record that they
	  return an
	  error without decoding, which matters because a caller that
	  ignores the
	  result reads output buffers that were never written.
	  
	  Documentation only: the comment-stripped source is unchanged byte
	  for
	  byte, so neither the library's behaviour nor its interface is
	  affected.

2026-07-27 11:32  aleidinger

	* libmp3lame/set_get.c:
	  doc: finish documenting the parameter interface for the public
	  API reference
	  
	  Completes the parameter interface: the block-type and
	  noise-shaping
	  controls, the attack thresholds, emphasis, the read-only figures
	  a caller
	  can ask the encoder for, the presets, the instruction-set
	  switches and the
	  last few settings that sit outside the main groups.
	  
	  The three functions that write the block-type choice share one
	  field, so
	  the last one called decides and the other two are forgotten; each
	  of them
	  now says so, and says which values do not simply overwrite what
	  was there.
	  The preset entry point is documented as applying its bundle
	  immediately -
	  so a call that precedes it is overwritten and one that follows it
	  wins -
	  and as the one function here that returns something other than
	  zero on
	  success, while silently accepting a preset it does not recognize.
	  The
	  instruction-set switches are described as a way to forbid a
	  family rather
	  than to require one, as naming x86 instruction sets and nothing
	  else, and
	  with the dependency between two of them; the newest is noted with
	  the
	  release it arrives in and as absent from older headers, along
	  with what to
	  do about that. Their return value
	  gets an example, because it reports on the encoder instance
	  rather than
	  on the request and is easy to read as a success indication - and,
	  since
	  writing that down showed there is no way to read the setting back
	  at all,
	  a to-do to provide one.
	  
	  The read-only figures say what they are worth and when. Several
	  of them
	  share a value with "not available", so each entry states what an
	  uninitialized instance returns and whether that is
	  distinguishable from a
	  real measurement; one of them has always returned a constant,
	  because the
	  figure it names cannot be computed by an encoder that sees a
	  single track.
	  
	  Seventeen functions in this file are declared only in an internal
	  header,
	  never in the installed one. They had been documented alongside
	  the rest,
	  which would have published them in the public reference as though
	  they
	  were part of the interface; they are now marked internal and
	  appear in the
	  internal documentation only.
	  
	  Documentation only: the comment-stripped source is unchanged byte
	  for
	  byte, so neither the library's behaviour nor its interface is
	  affected.

2026-07-27 11:32  aleidinger

	* libmp3lame/set_get.c:
	  doc: document the psychoacoustic settings for the public API
	  reference
	  
	  Continues the public API documentation with the section the
	  source itself
	  introduces as the one not to touch without knowing what you are
	  doing:
	  the masking offsets, the absolute threshold of hearing and its
	  adaptive
	  adjustment, the block-type coupling, temporal and inter-channel
	  masking,
	  and substep noise shaping.
	  
	  Two of these settings no longer affect the encoded audio. Asking
	  for
	  masking against the threshold of hearing alone, whether for all
	  blocks or
	  only for short ones, is not read by the quantizer any more; what
	  remains
	  is that one of the two still marks the encode as non-standard in
	  the VBR
	  header, which is a durable property of the file even though the
	  audio is
	  identical. Saying so is the point of documenting them - the
	  alternative
	  is a caller changing a setting, hearing no difference, and
	  concluding
	  something else is wrong. Dropping the threshold entirely, which
	  looks
	  like the same kind of switch, does still work, and the reference
	  now
	  distinguishes the three.
	  
	  Several settings resolve their default during initialization
	  rather than
	  at instance creation, so the same getter answers differently
	  before and
	  after it with nobody having chosen anything. Each such entry says
	  so and
	  sends the caller to the getter, rather than naming the value:
	  what an
	  unset setting resolves to is the encoder's business and not
	  something
	  this interface undertakes to keep. One of them cannot be read
	  back even
	  after initialization; documenting it made clear that this is a
	  gap rather
	  than a decision, so the entry says so and carries a to-do to
	  close it.
	  Two others can never be returned to the unresolved state, because
	  the
	  setter accepts a narrower range of values than the field holds.
	  
	  Where a setting is a packed word rather than the number or flag
	  its name
	  suggests, the layout is written out: which bits enable substep
	  shaping,
	  extend it to short blocks, or select its variant. The block-type
	  coupling
	  entry records that it shares one field with two other functions,
	  so the
	  last one called wins, and that a stereo encode overrides the
	  request
	  during initialization regardless.
	  
	  Documentation only: the comment-stripped source is unchanged byte
	  for
	  byte, so neither the library's behaviour nor its interface is
	  affected.

2026-07-27 11:31  aleidinger

	* libmp3lame/set_get.c:
	  doc: document the bitrate modes and filters for the public API
	  reference
	  
	  Continues the public API documentation with the choice between
	  constant,
	  average and variable bitrate, the quality level and the bitrate
	  bounds
	  that go with it, and the lowpass and highpass filters.
	  
	  The quality level is documented as running the opposite way to a
	  bitrate,
	  0 being the best, and both entry points that set it are recorded
	  as
	  clamping an out-of-range value and returning an error for it
	  anyway - a
	  combination a caller will otherwise read as "nothing happened".
	  
	  Several of these settings are rewritten during initialization,
	  and the
	  reference now says so, because the alternative is a caller
	  concluding the
	  library ignored them. The bitrate bounds are snapped to the rates
	  the
	  format actually offers, so asking for one number and reading back
	  a
	  nearby one is correct behaviour. The target average is clamped
	  into the
	  window the bounds leave, and it is not confined to the average
	  bitrate
	  mode: with a constant bitrate and no bitrate set, it becomes the
	  bitrate.
	  
	  The two filters look symmetrical and are not, which is the single
	  most
	  useful thing to know about them. A lowpass cutoff of zero asks
	  LAME to
	  pick one, and picking one can change the output sample rate,
	  because the
	  encoder then chooses the lowest rate that still carries the
	  cutoff. A
	  highpass cutoff of zero means no highpass at all: nothing derives
	  one,
	  so the filter exists only if a caller asks for it. The transition
	  widths
	  are preferences in both cases, since the polyphase filter can
	  only place
	  a transition on a band boundary.
	  
	  Documentation only: the comment-stripped source is unchanged byte
	  for
	  byte, so neither the library's behaviour nor its interface is
	  affected.

2026-07-27 11:30  aleidinger

	* libmp3lame/set_get.c:
	  doc: document the frame and quantization parameters for the
	  public API reference
	  
	  Continues the public API documentation with the settings that
	  shape the
	  frames themselves - the four header flag bits, the optional
	  per-frame
	  checksum and the bit reservoir - and with the quantization and
	  noise
	  shaping controls that follow them.
	  
	  The header flags are documented as what they are: declarations
	  LAME
	  writes through and never acts on, one of which defaults to set
	  rather
	  than clear. The checksum entry states its price, two bytes per
	  frame
	  taken out of the space left for audio, and warns that the header
	  bit
	  carries the opposite sense to the setting.
	  
	  Three of these are not the booleans their names suggest. The ISO
	  compliance setting is a three-way choice of how large a bit
	  reservoir the
	  bitstream may rely on, and its default is the most permissive of
	  the
	  three rather than the zero the name implies. The psychoacoustic
	  tuning
	  entry point is a single integer packing six unrelated fields,
	  whose
	  layout and quarter-decibel encoding a caller has no way to
	  discover; it
	  is now written out, together with the advice to compose the value
	  against
	  what is already there instead of assigning it. The two
	  quantization
	  comparison settings are unset by default and acquire different
	  values
	  during initialization, so reading one before and after
	  initializing gives
	  two different answers with nobody having set anything.
	  
	  The remaining experimental switches are described by what they do
	  to the
	  encode - one gives up the extra bits normally spent on the
	  topmost
	  scalefactor band, the other forces a short-block analysis the
	  model would
	  otherwise skip - and the obsolete padding control is marked
	  obsolete and
	  states that it reports success unconditionally, having nothing
	  left to
	  fail at.
	  
	  Documentation only: the comment-stripped source is unchanged byte
	  for
	  byte, so neither the library's behaviour nor its interface is
	  affected.

2026-07-27 11:29  aleidinger

	* libmp3lame/set_get.c:
	  doc: document the general control parameters for the public API
	  reference
	  
	  Continues the public API documentation with the settings that
	  steer an
	  encode as a whole: the algorithm-effort level, the channel mode,
	  the
	  bitrate and the compression ratio it can be given as instead,
	  free
	  format, the analyzer and header switches, the ReplayGain and
	  decode-while-encoding measurements, the gapless bookkeeping, and
	  the
	  three callbacks that decide where the library's own output goes.
	  
	  Several of these behave in ways a caller cannot guess from the
	  name, and
	  each is now said where it will be read. The effort level is not
	  the
	  audio quality setting and clamps out-of-range values instead of
	  refusing
	  them. Setting a bitrate above the tabulated maximum silently
	  switches
	  the bit reservoir off, because the two cannot be combined. Asking
	  to
	  decode while encoding fails outright in a library built without
	  that
	  feature, which is a build property and not a mistake by the
	  caller. And
	  the automatic mid/side function ignores its argument entirely:
	  whatever
	  it is given, it selects joint stereo.
	  
	  The obsolete entry points are marked as such and say what to call
	  instead - several are now plain aliases for the function that
	  replaced
	  them, and one has done nothing at all since the encoder it
	  belonged to
	  was removed.
	  
	  Documentation only: the comment-stripped source is unchanged byte
	  for
	  byte, so neither the library's behaviour nor its interface is
	  affected.

2026-07-27 11:29  aleidinger

	* libmp3lame/set_get.c:
	  doc: document the input-stream parameters for the public API
	  reference
	  
	  Continues filling in the public API documentation with the
	  settings that
	  describe the encoder's input: how many samples there are, at what
	  rate
	  and in how many channels, what to scale them by, and what rate to
	  write
	  into the stream.
	  
	  Two properties of this whole interface are stated once, at the
	  file, and
	  not repeated on every function. A setter validates only what it
	  can see,
	  so whether a combination of settings is usable is decided later;
	  and a
	  getter cannot report an error at all - given an unusable instance
	  it
	  returns 0, which for most of these settings is also a legitimate
	  value,
	  so a caller has to check the instance rather than the result.
	  
	  Two things worth a caller's attention are said where they will be
	  read.
	  Both settings that describe the input have defaults, so an
	  encoder that
	  is never told about its input initializes successfully and
	  encodes as
	  though it were CD audio - a wrong result rather than an error.
	  And the
	  sample count is not a limit on encoding: it feeds a frame
	  estimate, and
	  its default is the documented "length not known" value rather
	  than a
	  count.
	  
	  Documentation only: the comment-stripped source is unchanged byte
	  for
	  byte, so neither the library's behaviour nor its interface is
	  affected.

2026-07-27 11:28  aleidinger

	* Makefile.am, Makefile.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/doxygen/internal.md,
	  doc/maintainer-gen-api-docs.md, maintainer/gen-api-docs.sh:
	  Add a "gen-api-docs" script to publish the API documentation
	  
	  The generated documentation has so far only ever existed in
	  whoever's
	  build directory produced it. This adds the step that puts it
	  where
	  people can read it: build both sets - the public library
	  reference and
	  the internal developer documentation - and copy them into the
	  website
	  working tree.
	  
	  It stops at the working tree. Committing and uploading the
	  website is
	  the maintainer's step, the same as committing to the repository,
	  so the
	  script writes, says exactly what it wrote, and gets out of the
	  way.
	  
	  Both sets are published, which is a decision rather than a
	  default: the
	  internal one documents structures and functions that carry no
	  stability
	  promise at all. It is therefore linked only from the
	  developer-facing
	  pages, and never without saying what it is not.
	  
	  The copy is a clean copy, so a page that no longer exists stops
	  being
	  served rather than lingering. And nothing is copied at all unless
	  the
	  generated documentation contains something that could only have
	  come
	  from parsed source: a run that reads no input still exits
	  successfully
	  and still writes a complete set of stylesheets and images, and
	  copying
	  that over the website would quietly replace a working reference
	  with an
	  empty one.
	  
	  The generated HTML is not added to the distribution; it is
	  reproducible
	  from the tarball, and a second copy would only go stale.

2026-07-27 11:27  aleidinger

	* Makefile.am, Makefile.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/doxygen/internal.md,
	  doc/maintainer-build-matrix.md, doc/maintainer-check-dist.md,
	  maintainer/check-dist.ps1, maintainer/check-dist.sh,
	  maintainer/gen-build-matrix.sh:
	  Add a "check-dist" validator for a finished distribution tarball
	  
	  Rolling a release currently ends with a tarball nobody has tried.
	  This
	  adds a script that takes one, unpacks it, builds it in every
	  configuration the machine can build, runs the tests in each, and
	  reports
	  every step as a named check with a short description of what it
	  establishes and a verdict.
	  
	  Everything it uses comes out of the tarball rather than the tree
	  it was
	  rolled from - the build harness, the ABI check, the man pages -
	  because
	  a file missing from the tarball is exactly the defect being
	  looked for.
	  It continues through all its checks and only exits non-zero at
	  the end,
	  since a release candidate is more useful fully surveyed than
	  abandoned
	  at the first problem.
	  
	  A check that cannot be made reports one of two verdicts rather
	  than one.
	  SKIP means a prerequisite is missing on this machine and
	  installing it
	  would produce a real answer; not applicable means the check has
	  no
	  meaning on this path at all and nothing installable would change
	  that.
	  The distinction is what makes a log from somebody else's machine
	  readable, and it is what lets the PowerShell companion for the
	  native
	  Windows build be honest about the checks that only exist in the
	  autotools build.
	  
	  The build harness gained one option along the way: extra
	  configure
	  arguments to be added to every cell. The per-cell unit tests need
	  it -
	  without it no cell enables them, so those checks would have
	  reported
	  "skipped" for ever while looking like they were being run.

2026-07-27 11:26  aleidinger

	* doc/Doxyfile.in, libmp3lame/lame.c, libmp3lame/version.c:
	  doc: document the encoder lifecycle for the public API reference
	  
	  Continues filling in the public API documentation, this round the
	  calls
	  that bracket an encode: creating an instance, turning the
	  settings into a
	  usable encoder, starting and finishing a bitstream, releasing the
	  instance, and the two reporting calls that describe the
	  configuration
	  those steps arrived at. Each states what it assumes, what it
	  returns and
	  what it does not do - notably that finishing the stream is not
	  the same
	  as releasing it, and that the one call kept only for binary
	  compatibility with older releases loses the window between the
	  two.
	  
	  The encoding engine's own file is described as well, so the
	  reference
	  has a page for it rather than a set of functions with nothing
	  around
	  them. The long-standing request for friendlier output from the
	  detailed
	  configuration dump is kept with it, as a to-do on the function it
	  concerns rather than a note floating in front of it.
	  
	  The reference had also been listing internal structures. The only
	  one a
	  caller ever names is the version struct, which is documented; the
	  rest
	  are LAME's own and are now hidden, which completes the split the
	  previous
	  round set up for functions and removes the last name in the
	  public
	  reference that is not an exported symbol.
	  
	  Documentation only: the comment-stripped source of both files is
	  unchanged byte for byte, so neither the library's behaviour nor
	  its
	  interface is affected.

2026-07-27 11:25  aleidinger

	* doc/Doxyfile.in, include/lame.h, libmp3lame/version.c:
	  doc: document the public version API, and keep internals out of
	  it
	  
	  The functions reporting LAME's version, the psychoacoustic
	  model's version
	  and the build's pointer width are part of the public API but were
	  described
	  in a line each, with no statement of what a caller may do with
	  the returned
	  string or when two builds produce the same one - which is the
	  property that
	  decides whether a version belongs in an encoded stream. Document
	  them at
	  their definitions, with the numerical form's structure and an
	  example of
	  testing for a minimum version.
	  
	  One of them was documented as feeding the LAME tag. It does not:
	  the tag
	  carries its own string from a function that is not exported,
	  whose width is
	  fixed by the tag format, and which is now marked internal. The
	  public
	  function has no length guarantee at all, so a caller copying it
	  into a fixed
	  field has to bound the copy - said where that caller will read
	  it.
	  
	  The public documentation configuration now hides undocumented
	  members. The
	  public and internal APIs are defined in the same files and are
	  told apart by
	  the internal marker, not by excluding directories; without this
	  an internal
	  function's name still appeared in the public output once its
	  comment body had
	  been stripped.
	  
	  Documentation only: the comment-stripped token stream of every
	  touched file
	  is byte-identical.

2026-07-27 11:24  aleidinger

	* frontend/get_audio.c, test/unit/test_get_audio_aiff.c:
	  get_audio: range-check the AIFF sample rate before converting it
	  
	  The sample rate declared in an AIFF header reaches the encoder as
	  an int,
	  with nothing checking that the conversion is defined. It is not,
	  for any
	  value outside that type's range, and a crafted COMM chunk can
	  declare
	  three kinds of such a value: an infinity, which is what an
	  all-ones
	  exponent means; a finite rate far above the range; and a negative
	  one. In
	  practice each converts to something the encoder then refuses, so
	  the file
	  is rejected either way - but only by accident of what the
	  conversion
	  happens to produce.
	  
	  Validate the rate where the chunk's other fields are validated
	  and treat
	  one that cannot be represented as the malformed field it is.
	  Nothing a
	  real encoder writes is affected. The unit test covers each of the
	  three,
	  none of which it reached before.
	  
	  Frontend-only; the library and the bitstream are untouched.

2026-07-27 11:24  aleidinger

	* configure, configure.ac, libmp3lame/lame.c,
	  test/unit/test_encode_nonfinite.c:
	  Keep the non-finite PCM check out of the optimizer's reach
	  
	  The floating point encoding entry points screen their input for
	  NaN and
	  infinity, because such a sample spreads through the psycho
	  acoustic model and
	  trips an assertion inside it. The library is built with
	  -ffast-math, which
	  tells the compiler that no such value ever occurs and so entitles
	  it to delete
	  the screen. Compilers have begun acting on that: from clang 21
	  onwards the
	  screen is absent from the built library on every platform, and
	  one non-finite
	  sample from the caller terminates the application. gcc and
	  earlier clang keep
	  it, which is why this presented as a single platform misbehaving.
	  
	  Spelling the test in terms of the raw exponent bits does not
	  avoid it - the
	  integer comparison is recognised as a floating point class test
	  and folded just
	  as isfinite() would be. The value instead reaches the comparison
	  through a
	  volatile object, which the compiler may not reason across. The
	  cost does not
	  rise above the run-to-run spread of an encode.
	  
	  The unit test needs the same treatment for the same reason: built
	  with the
	  library's flags it could not construct a non-finite sample to
	  test with, so it
	  failed whether or not the library was sound - and would have gone
	  on failing
	  after a fix.
	  
	  The comment on the optimization flags gains the general point,
	  since the flag
	  is where this is worth knowing about.

2026-07-27 11:23  aleidinger

	* test/unit/test_set_get.c:
	  test: keep the parameter-API test comment self-contained

2026-07-27 11:23  aleidinger

	* configure, configure.ac, doc/maintainer-abi.md:
	  Let a user's CFLAGS add to the build's flags instead of replacing
	  them
	  
	  Passing CFLAGS to configure suppressed the project's own
	  optimization
	  settings entirely. One flag on the command line was enough to
	  drop
	  -ffast-math, which the DSP code is validated with and which
	  changes the
	  encoder's output, along with the rest of them - silently, and
	  with
	  nothing in the build log to say so.
	  
	  They are now always applied, with the given CFLAGS placed after
	  them, so
	  it adds to the build and overrides only what it names.
	  CFLAGS='-g' asks
	  for debug information and keeps the optimization; CFLAGS='-O2 -g'
	  asks
	  for both and still keeps the arithmetic settings. Anything can
	  still be
	  countermanded by naming its opposite, since the later flag wins.
	  
	  A default build is unchanged, flag for flag.
	  
	  The maintainer guide for the ABI check asked for CFLAGS='-g -O2'
	  to get
	  the debug information its deepest check needs, which was this
	  problem in
	  miniature; it now asks for CFLAGS='-g'.

2026-07-27 11:22  aleidinger

	* frontend/get_audio.c, include/lame.h, libmp3lame/util.h:
	  Declare the decoder analysis hooks in the public header
	  
	  Both are exported from the library and called across the library
	  boundary by the frontends, but they were declared only in an
	  internal
	  header. An analyzer frontend outside the tree therefore had to
	  write the
	  prototypes itself, which is what our own frontend did too - with
	  a
	  comment saying the internal header pulls in too much to include.
	  
	  Declaring them where the rest of the decoding interface lives
	  removes
	  that. The plotting structure stays an incomplete type: it is only
	  ever
	  passed by pointer, its layout is internal and moves with the
	  encoder, and
	  a caller that wires the hooks up never needs its fields. One that
	  does
	  need them includes the internal header and takes what comes with
	  it.
	  
	  The two now-redundant declarations are dropped.

2026-07-27 11:21  aleidinger

	* doc/doxygen/unit_tests.dox:
	  doc: document how to run the unit tests, and what "make test" is
	  not
	  
	  The unit-test page said which configure option turns the suite on
	  and
	  left the rest to be worked out. Added the operational half: that
	  the
	  option is what matters and having CMocka installed does not
	  enable
	  anything on its own, where the per-test logs and the collected
	  failure
	  log are written, and that a single test binary runs on its own.
	  
	  The distinction worth stating is between the two targets that
	  sound
	  interchangeable. Only "make check" can fail; "make test" prints a
	  count
	  of the bytes by which the encoder's output differs from the
	  committed
	  reference and always succeeds, and that count is expected to be
	  non-zero
	  on a different processor or compiler. Reaching for it to find out
	  whether
	  something is broken gives an answer that means nothing.

2026-07-27 11:19  aleidinger

	* Makefile.am, Makefile.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/doxygen/internal.md,
	  doc/maintainer-abi.md, include/lame.def, maintainer/abi,
	  maintainer/abi/libmp3lame.abi, maintainer/abicheck.sh:
	  Add a "make abicheck" target for the exported library interface
	  
	  The exported interface is written down twice, once per platform,
	  and both
	  files are maintained by hand: the linker is told to export
	  exactly what
	  they list, so they are not documentation of the ABI, they are the
	  ABI.
	  Nothing checked them against each other or against the library
	  that was
	  actually built.
	  
	  The new target does, in three steps that get deeper as the tools
	  allow.
	  The export lists are compared with one another, which needs
	  nothing but
	  the source tree. The built library's export table is read back
	  and
	  compared with the list for its platform. Where libabigail is
	  installed
	  and the build carries debug information, the interface is
	  compared with a
	  baseline committed alongside, which also catches the changes that
	  keep
	  every name intact: a parameter that changed type, a struct that
	  gained a
	  member, an enum whose values moved.
	  
	  A missing tool skips a step and never fails the run; a step that
	  could be
	  made and did not hold does. The check needs no audio input and no
	  second
	  build to compare against, so it also runs from an unpacked
	  tarball, and
	  the baseline ships in the distribution so that it can.
	  
	  The two lists have never named the same set of symbols, and
	  bringing them
	  together is part of this. The rule they follow: one list per
	  operating
	  system, not one per build, each the logical OR over every
	  configuration
	  we ship. A symbol belongs in it when the code is compiled in and
	  is part
	  of the exported interface, whatever any single build does with
	  it.
	  Deprecated code that is still built is still exported and still
	  belongs;
	  only what is not compiled at all, or is not part of the
	  interface, stays
	  out.
	  
	  Applying that, the DLL gains what the shared library alone
	  exported so
	  far: the decoder analysis hooks, so that an analyzer frontend
	  outside the
	  library can reach them on Windows too (SF bug #515, which was
	  answered on
	  POSIX alone), the Latin-1 tag setters whose ucs2, utf16 and utf8
	  counterparts the DLL has exported all along, and the deprecated
	  entry
	  points whose declarations the header compiles out while their
	  definitions
	  are still built. Adding a name to the Windows list is additive;
	  nothing
	  that links against the DLL today is affected.
	  
	  What a green run does not say is written down with it: it
	  compares
	  against the union of the configurations we ship rather than the
	  one that
	  was built, so it cannot say what this build contains; it says
	  nothing
	  about behaviour, so an entry point that links and does nothing
	  passes; it
	  reads only the library built here, so a fault in the Windows list
	  surfaces only in a Windows build; and it inspects the library's
	  exports
	  rather than the programs we ship, so a symbol the frontend needs
	  can be
	  dropped with every check still green - which is how the analysis
	  hooks
	  were nearly lost.

2026-07-27 11:19  aleidinger

	* doc/Doxyfile-common.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in:
	  doc: give the Doxygen templates a placeholder configure will not
	  consume
	  
	  The rule that expands the Doxyfile templates substituted a
	  placeholder
	  named the same as one of Autoconf's own variables. Because the
	  makefile
	  itself goes through config.status, the search pattern was
	  replaced along
	  with everything else, leaving a rule that looked for the absolute
	  source
	  path inside the template and matched nothing.
	  
	  Nothing was substituted, so both configurations were handed input
	  paths
	  that do not exist and documented an empty tree: the generated
	  output was
	  stylesheets and icons with not one page of content in it. Doxygen
	  reports
	  missing inputs as warnings among its others and still exits
	  successfully,
	  which is why the documentation appeared to build for as long as
	  it did.
	  
	  Renaming the placeholder to one Autoconf has no reason to touch
	  fixes it.
	  The neighbouring placeholders were never affected, having names
	  it does
	  not recognize. The internal configuration now produces 849 files
	  where it
	  produced 45, and the public one 152.

2026-07-25 10:36  aleidinger

	* doc/man/mp3rtp.1:
	  mp3rtp: explain the destination-argument change in the manual
	  page
	  
	  The page documented only the current invocation, so a reader
	  coming
	  from the older single-argument form had no signal from the manual
	  that
	  the syntax had changed, or why - it had to be pieced together
	  from the
	  changelog or the source. Add a short HISTORY section so the
	  manual page
	  answers that on its own.

2026-07-25 10:36  aleidinger

	* doc/html/history.html:
	  Add the 4.1 changelog entry for the vectorized quantization loops
	  
	  Record the run-time SIMD work in the user-facing history: which
	  inner
	  loops were vectorized, the measured whole-encode speedups for CBR
	  and
	  VBR on both toolchains, that AVX2 helps CBR only, and the new
	  --noasm avx2 lever that keeps SSE2 while disabling just the AVX2
	  tier.

2026-07-25 10:35  aleidinger

	* doc/Doxyfile-internal.in, doc/Makefile.am, doc/Makefile.in,
	  doc/doxygen/internal.md, doc/doxygen/vector-dispatch.dox,
	  libmp3lame/takehiro.c, libmp3lame/util.c:
	  Document the vector dispatch design decisions in the internal
	  docs
	  
	  The run-time SIMD dispatch for the hot quantization loops makes
	  two
	  choices that are counter-intuitive and easy to reopen later: the
	  ladder
	  skips SSE4.1/SSE4.2 entirely, and the variable-bitrate noise
	  estimate is
	  offered at SSE2 only. Add an internal Doxygen page recording why
	  - a
	  16-bit code path is denser than the 32-bit SSE4 instructions for
	  data
	  this small, and the noise routine is gather-bound with a fixed
	  reduction
	  order that a wider vector cannot exploit - and cross-reference it
	  from
	  the dispatch selector and the affected routines.

2026-07-25 10:35  aleidinger

	* Makefile.MSVC, libmp3lame/vbrquantize.c,
	  libmp3lame/vector/Makefile.am, libmp3lame/vector/Makefile.in,
	  libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/xmm_calc_sfb_noise.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_calc_sfb_noise_vector.c,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Vectorise the VBR scalefactor-band noise calculation
	  
	  Estimating how much quantization noise a scalefactor band would
	  carry is one
	  of the larger costs of a variable-bitrate encode: the band is
	  quantized and
	  the squared error summed, once per candidate scalefactor, many
	  times per
	  frame. As with the xr^(3/4) quantiser, every compiler was left to
	  vectorise
	  this on its own and they did so unevenly, so "the encoder" ran
	  materially
	  different code depending only on which one built it. The loop now
	  has an SSE2
	  implementation chosen at run time, the way the Huffman table
	  search and the
	  xr^(3/4) quantiser already are, so every x86-64 CPU runs it. An
	  AVX2 version
	  was written and measured and made no difference to this routine's
	  variable-bitrate work - the eight-wide gather is no faster than
	  the SSE2 one on
	  the cores tested - so it is not carried; the capability ladder
	  allows a routine
	  to exist at a single tier. The scalar routine stays as the common
	  ground and
	  the correctness reference, and narrow bands - about a third of
	  the calls,
	  carrying a fifth of the values - keep to it, below a width taken
	  from a
	  histogram of real encodes.
	  
	  The quantization part is elementwise and produces the same
	  integers as before.
	  The sum of squares is a floating-point reduction, and there the
	  compilers had
	  quietly disagreed: gcc and clang, under this library's fast-math,
	  reassociate
	  the sum into the natural horizontal one and agree with each
	  other, while MSVC
	  keeps the source order. The vector code pins the gcc/clang order
	  and holds the
	  running sum in a register so the compiler does not reassociate it
	  back, so the
	  vector code and all three toolchains now compute the estimate the
	  same
	  way where before it was compiler-dependent. Pinning that order
	  changes the
	  last bit of the estimate on the toolchains that had reassociated
	  differently.
	  On the material tested this leaves the encoded bitstream
	  unchanged on
	  everything except the hardest transient tracks, where it flips a
	  single
	  scalefactor; scored against the original with PEAQ that flip is
	  perceptually
	  identical (the objective difference grade does not move), and
	  every track of
	  the quality corpus scores an unchanged grade across the
	  variable-bitrate modes.
	  
	  Because the compilers already vectorised much of this, the wall
	  clock is
	  unchanged on gcc and clang; the measurable gain is on MSVC, which
	  had left more
	  of it scalar - a few percent of a variable-bitrate encode on a
	  Core i7-10700.
	  
	  A unit test checks the SSE2 routine against a double-precision
	  reference over
	  the tail cases and the ends of the index range, which a real
	  encode cannot be
	  relied on to reach. The scalar routine and the alternative
	  IEEE-754
	  formulation are left untouched; a build selecting the latter
	  keeps its own path.

2026-07-25 10:34  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_choose_table_vector.c,
	  test/unit/test_encode_mono_interleaved.c,
	  test/unit/test_encode_nonfinite.c, test/unit/test_merge_argv.c,
	  test/unit/test_parse_pathlen.c, test/unit/test_sanity.c,
	  test/unit/test_totalframes.c, test/unit/test_unused.h:
	  Mark the unused unit-test parameter with the attribute macro
	  
	  Each cmocka test silences its unused fixture argument with a
	  `(void) state;`
	  statement in the body. The library marks a deliberately unused
	  parameter with
	  an attribute macro instead, and one of the tests had started to
	  follow that;
	  the rest had not, so the same intent was written two ways across
	  the suite.
	  
	  This settles on the macro everywhere the parameter is
	  unconditionally unused.
	  Most of these tests compile against the public library API alone
	  and pull in no
	  internal header; making them reach into one just to borrow the
	  macro would be
	  the wrong dependency for a public-API test, so they take it from
	  a small header
	  of their own that carries the same definition. The tests that
	  already include
	  the internal header use it directly.
	  
	  Only the always-unused case is converted. Casts that exist to
	  discard a return
	  value, and ones whose parameter is unused only in a particular
	  build
	  configuration, are deliberately left as they were - the macro
	  would say the
	  wrong thing about them.

2026-07-25 10:33  aleidinger

	* Makefile.MSVC, libmp3lame/takehiro.c,
	  libmp3lame/vector/Makefile.am, libmp3lame/vector/Makefile.in,
	  libmp3lame/vector/avx2_quantize_lines.c,
	  libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/xmm_quantize_lines.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_quantize_lines_vector.c,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Vectorise the quantization of xr^(3/4)
	  
	  Turning the scaled spectral values into quantized integers is one
	  of the
	  larger costs of a constant-bitrate encode, and every compiler was
	  left to
	  vectorise its inner loop or not on its own. Some do: current gcc
	  and clang
	  already emit the four-wide sequence. MSVC does not, and left the
	  loop
	  essentially scalar, so "the encoder" ran materially different
	  code depending
	  only on which compiler built it. Writing the loop out makes every
	  toolchain
	  run the same code, the way the CPU-dispatch ladder already makes
	  every CPU of
	  one architecture do - and it recovers about eight percent of a
	  constant-bitrate
	  encode on the toolchain that had been leaving it on the table,
	  with a gather on
	  the wider tier on top.
	  
	  The loop is elementwise - scale, truncate, look up a rounding
	  term, add,
	  truncate - with no accumulation, so the vector form produces the
	  same integers
	  as the scalar one and the output does not move. A unit test
	  checks it against
	  an independent scalar reference over the tail and the ends of the
	  index range,
	  which a real encode cannot be relied on to reach.
	  
	  There is one exception, and it is a property of the old code
	  rather than the
	  new. On 32-bit x86 the compilers compute scalar floating point in
	  the x87
	  unit, which keeps a wider intermediate than single precision; the
	  scalar loop
	  therefore rounded from that wider value, while the vector form
	  rounds from a
	  true single-precision one. So a 32-bit build's output shifts
	  slightly - the
	  same shift a 64-bit build already had relative to it, now
	  reaching 32-bit too.
	  64-bit builds, and 32-bit builds that do scalar float on SSE, are
	  unaffected.
	  
	  Short regions stay on the scalar path, where the setup does not
	  pay for
	  itself; a fifth of the calls are that short and they carry under
	  a percent of
	  the values.

2026-07-25 10:32  aleidinger

	* Makefile.MSVC, libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/takehiro.c, libmp3lame/vector/Makefile.am,
	  libmp3lame/vector/Makefile.in,
	  libmp3lame/vector/avx2_choose_table.c,
	  libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/xmm_choose_table.c,
	  libmp3lame/vector/xmm_quantize_sub.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_choose_table_vector.c,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Vectorise the Huffman table search
	  
	  Choosing the table that encodes a region in the fewest bits is a
	  quarter to a
	  third of the time a constant-bitrate encode spends, and more than
	  a fifth of
	  a variable-bitrate one - the largest single cost in the encoder
	  that had no
	  vector implementation on any platform, and none of the compilers
	  in use
	  vectorise the table lookups on their own.
	  
	  The work is integer arithmetic over constant integer tables, so
	  the vector
	  form computes the same sums in a different order and the output
	  is identical
	  by construction rather than by measurement. A unit test checks
	  the routines
	  against a plain scalar reference over the boundaries an encode
	  cannot be
	  relied on to reach.
	  
	  What makes it worth doing is narrowing the values to sixteen bits
	  first.
	  They are quantized magnitudes far below that, and the narrower
	  form both puts
	  eight of them in a register instead of four and brings two useful
	  instructions into reach on the baseline: a signed maximum, and a
	  multiply-add
	  that turns a pair of values into their table index on its own. So
	  the whole
	  gain is available on any processor of the last two decades rather
	  than only
	  on recent ones; the wider tier adds a further doubling of the
	  maximum alone,
	  and nothing above the baseline helps the lookups at all.
	  
	  Narrowing saturates, which only the maximum can meet. It reports
	  its value
	  exactly while that value can matter, and something larger than
	  the rejection
	  threshold otherwise - which is all its caller asks. A
	  compile-time check
	  ties that to the threshold, so raising the threshold past the
	  saturation
	  point stops the build instead of quietly changing a result.
	  
	  Regions below a few dozen values stay on the scalar path: a third
	  of them are
	  that short, they hold about a twentieth of the values between
	  them, and the
	  vector form is slower there than the loop it replaces.

2026-07-25 10:31  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  config.h.in, configMS.h, configure, configure.ac,
	  doc/Makefile.in, doc/html/Makefile.in, doc/html/detailed.html,
	  doc/man/Makefile.in, doc/man/lame.1, dshow/Makefile.in,
	  frontend/Makefile.in, frontend/parse.c, include/Makefile.in,
	  include/lame.h, libmp3lame/Makefile.in, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c,
	  libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in, test/Makefile.in, test/unit/Makefile.in,
	  vc_solution/Makefile.in:
	  Detect AVX2 and offer it to the vector routines
	  
	  The encoder chooses its vector code from one ordered list of
	  instruction
	  sets, so adding a wider one is a new entry rather than another
	  flag tested at
	  every call site. AVX2 becomes the top entry, above SSE2: the
	  configure
	  summary reports what was built, and --verbose reports what the
	  running
	  machine actually got.
	  
	  The --noasm switch, which already turned the vector code off
	  wholesale, gains
	  a name for the new tier. It can disable AVX2 on its own and leave
	  SSE2 in
	  use, so one tier can be measured against the one below it and
	  against the
	  plain C path the way a speed test or a bug hunt wants; disabling
	  SSE still
	  removes the whole group above it, and the switch may be repeated
	  to turn off
	  more than one.
	  
	  Detecting AVX2 is not one question but three. The processor must
	  have the
	  instructions, the operating system must preserve the wider
	  registers across a
	  context switch, and only then may that state be read at all.
	  Asking the
	  first alone reports success on a system where the registers are
	  silently
	  corrupted, so the path that queries the processor directly now
	  asks all
	  three; compilers offering their own feature test already cover
	  it.
	  
	  The build-time test compiles both an integer and a floating point
	  construct,
	  because they do not answer alike: on 32-bit x86 under the
	  optimization flags
	  used here, one toolchain accepts the first and rejects the
	  second. A test
	  that asked about integers alone would promise a capability the
	  build cannot
	  deliver. Where the per-function form is rejected, the fallback
	  gives the
	  file its own baseline, as the tier below already does.
	  
	  The list is per architecture and names real instruction sets, so
	  each
	  architecture carries only the entries that exist on it and
	  reports its own.
	  None of it is exported and the library interface is unchanged.

2026-07-25 10:30  aleidinger

	* configMS.h:
	  Let the configuration header report the standard header, not
	  include it
	  
	  A configuration header states what the platform has; including a
	  system
	  header from it puts that header in front of every translation
	  unit
	  whether or not it needs one. The sources that use the fixed-width
	  integer types already ask for them through the guard the rest of
	  the
	  tree uses, so reporting the header as present is enough to reach
	  them on
	  Windows too.

2026-07-25 10:30  aleidinger

	* frontend/gpkplotting.c, frontend/gtkanal.c:
	  Fix the analyzer's title split and its clashing stringify macro
	  
	  Two defects in the analyzer frontend, both surfaced by warnings
	  the
	  Windows build had never counted.
	  
	  The plot title is split at the first newline using a byte offset
	  held in
	  an int, which is then compared against the string length as a
	  signed
	  value and used for the remainder's length. Hold the offset in the
	  type
	  the string functions actually return and convert once, at the
	  drawing
	  call that wants its own signed length.
	  
	  Separately, the version-string stringify macro redefined a macro
	  of the
	  same name that the version header already provides with a
	  different
	  meaning, so which of the two the rest of the file saw depended on
	  include
	  order. Use the existing one instead of declaring a second.

2026-07-25 10:29  aleidinger

	* Dll/BladeMP3EncDLL.c, configMS.h:
	  Apply the Windows deprecation policy where it was not reaching
	  
	  The Windows configuration header turns off MSVC's deprecation
	  warnings
	  for the standard library names this code deliberately keeps, but
	  two
	  cases slipped past it. The legacy encoder DLL includes no
	  configuration
	  header at all, so nothing that header sets ever applied to it.
	  And the
	  POSIX names MSVC deprecates in favour of an underscored spelling
	  of its
	  own answer to a different switch than the one already set.
	  
	  Give the DLL the same configuration prologue every other source
	  has, and
	  set the second switch beside the first. The deprecation reports
	  left in
	  this build are the legacy sockets calls in the RTP frontend,
	  which are
	  worth replacing rather than silencing.

2026-07-25 10:29  aleidinger

	* configMS.h:
	  Take the fixed-width integer types from the standard header on
	  Windows
	  
	  The Windows configuration header declared int8_t and its siblings
	  itself, from a time before the compilers used here shipped
	  <stdint.h>.
	  Under MSVC that made int8_t an alias of plain char, so as soon as
	  an
	  optional library's header pulled in the real <stdint.h> the type
	  was
	  redefined as signed char and the build reported it. The
	  alternative
	  spelling in the same block declared the names as object-like
	  macros,
	  which silently corrupts any later declaration using one of them
	  as a
	  type.
	  
	  Include the standard header instead and report it as present, so
	  the
	  guards elsewhere in the tree that ask for it take the same path
	  on
	  Windows as they do everywhere else. The two floating-point
	  aliases the
	  block also declared had no users.

2026-07-25 10:28  aleidinger

	* libmp3lame/mpglib_interface.c:
	  Reject a decoded frame whose sample count does not fit the result
	  
	  The decoder returns the number of samples it produced in an int,
	  and the
	  demultiplexing loop indexes the caller's output buffers with it.
	  That
	  count was narrowed from the decoder's own size_t without a check.
	  libmpg123 bounds a single frame far below the limit, so this is
	  not
	  reachable in practice, but a wrapped count would then be used as
	  a
	  length against buffers sized by the caller.
	  
	  Report the failure at the point the count is computed instead.

2026-07-25 10:28  aleidinger

	* libmp3lame/set_get.c, test/unit/test_set_get.c:
	  Bound the maximum-samples estimate to the int it is returned in
	  
	  lame_get_maximum_number_of_samples() derives a sample count from
	  a
	  size_t buffer size and returns it in an int, without bounding any
	  step
	  of the estimate. From a buffer of about 780 MB upward the frame
	  count
	  multiplied by the frame size overflows, and the function returns
	  a
	  negative sample count, which no encode call can accept. Larger
	  buffers
	  return arbitrary values.
	  
	  This is not specific to 64-bit builds. Only the division narrows
	  with
	  the pointer width, and it is the least reachable of the three
	  conversions; the overflow of the product is reached on any
	  platform
	  whose size_t can express the buffer size.
	  
	  Bound each step to the range its result is returned in, so a
	  buffer too
	  large to describe reports the largest sample count an encode call
	  can
	  take rather than a wrapped one. Answers below that limit are
	  unchanged.

2026-07-23 08:15  aleidinger

	* doc/html/history.html:
	  Add the 4.1 changelog entries for the packaging and mp3rtp fixes

2026-07-23 08:14  aleidinger

	* frontend/mp3rtp.c:
	  mp3rtp: stop truncating the destination address
	  
	  Where LAME is built without IPv6, the destination address is read
	  with a
	  field width of eleven characters into a buffer of sixteen. A
	  dotted quad can
	  be fifteen, so anything longer is silently cut short and the
	  stream is sent to
	  a different host than the one asked for, with no diagnostic -
	  including the
	  multicast address the program's own usage text suggests. Widen
	  the field to
	  what the buffer holds.

2026-07-23 08:14  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in, configure,
	  configure.ac, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.am, doc/man/Makefile.in, doc/man/mp3rtp.1,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in,
	  test/Makefile.in, test/unit/Makefile.in, vc_solution/Makefile.in:
	  Add a manual page for mp3rtp
	  
	  mp3rtp has never had one, so a distribution that builds the
	  program installs
	  a binary its users have no documentation for. The page is
	  installed only where
	  the program is, and distributed unconditionally.
	  
	  Manual page contributed by Fabian Greffrath.

2026-07-23 08:13  aleidinger

	* Makefile.am, Makefile.in:
	  Restore the ability to build a release tarball

2026-07-23 08:13  aleidinger

	* doc/html/history.html:
	  Add the 4.1 changelog entries for the vector code

2026-07-23 08:12  aleidinger

	* INSTALL, doc/html/detailed.html, doc/man/lame.1, lame.spec.in:
	  Update the documentation for the vector code
	  
	  The manual page, the HTML option reference and the install notes
	  all described
	  the vectorized routines as assembler optimizations that appear
	  only if an
	  assembler was found at build time. Neither half is true any more:
	  the routines
	  are compiled into every x86 binary and chosen per processor at
	  run time, and
	  nothing needs an assembler.
	  
	  "--noasm" keeps its three values, but only one of them still
	  selects
	  something, so say which. The install notes also told readers to
	  build with a
	  makefile that was removed along with the bundled decoder.

2026-07-23 08:11  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/BladeMP3EncDLL.c, Dll/Makefile.in,
	  Makefile.MSVC, Makefile.in, config.h.in, configMS.h, configure,
	  configure.ac, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  frontend/parse.c, include/Makefile.in, libmp3lame/Makefile.am,
	  libmp3lame/Makefile.in, libmp3lame/fft.c, libmp3lame/i386,
	  libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/takehiro.c,
	  libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vector/Makefile.am, libmp3lame/vector/Makefile.in,
	  libmp3lame/vector/xmm_quantize_sub.c, mac/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, test/Makefile.in,
	  test/unit/Makefile.in, vc_solution/Makefile.am,
	  vc_solution/Makefile.in, vc_solution/README.vs.txt,
	  vc_solution/vs_arch_nasm.props, vc_solution/vs_arch_sse2.props,
	  vc_solution/vs_arch_x87.props, vc_solution/vs_lame.slnx,
	  vc_solution/vs_lame_clients.slnx,
	  vc_solution/vs_lame_lame.vcxproj,
	  vc_solution/vs_lame_mp3rtp.vcxproj,
	  vc_solution/vs_lame_mp3x.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters,
	  vc_solution/vs_libmp3lame_dll.vcxproj, vc_solution/vs_nasm.props,
	  vc_solution/vs_nasm.targets, vc_solution/vs_nasm.xml:
	  Use the vector routines on every x86 build, and drop the assembly
	  they replace
	  
	  The vectorized routines have been part of LAME since 3.98, and
	  the two checks
	  that could switch them on only ever succeeded on 64-bit Windows:
	  one tested a
	  macro that every x86-64 Windows toolchain defines, the other was
	  defined in a
	  single Visual Studio property sheet. Even there only one of the
	  two routines
	  was reached; the other needed the property sheet. So Linux,
	  FreeBSD, macOS
	  and every 32-bit build compiled the routines into the binary and
	  then called
	  the plain C code instead.
	  
	  32-bit x86 did not even compile them. The configure check asked
	  whether the
	  compiler targets SSE2 by default, which is a stricter question
	  than the code
	  asks, since the routines carry a per-function target attribute
	  and need no
	  such baseline. That check was tightened in 3.100 to repair a real
	  build
	  failure, and the attribute that made the strictness unnecessary
	  arrived in
	  3.101, but the check was never revisited.
	  
	  Both conditions are replaced by what every toolchain already
	  publishes: a
	  compile-time test of the target baseline, and a run-time query of
	  the
	  processor where the baseline does not already guarantee the
	  instructions. The
	  routines are now built on every x86 target and chosen per CPU, so
	  one binary
	  serves machines with and without SSE2, and "lame --verbose"
	  reports what the
	  machine it is running on actually got. The configure check is
	  compiled with
	  the flags the build really uses, because one supported
	  combination rejects the
	  attribute and a check run with other flags would enable code that
	  then fails
	  to compile - which is how 32-bit x86 lost these routines in the
	  first place.
	  
	  With the intrinsics reaching every platform, the hand-written
	  assembler has
	  nothing left to offer. It required an external assembler, was
	  unreachable on
	  any 64-bit host, targeted an instruction set AMD removed from its
	  processors
	  in 2010, and half of its files had not been assembled by any
	  build system in
	  this century. It is removed, together with the build machinery,
	  the configure
	  switch and the Visual Studio configurations that existed to
	  select it or to
	  select the vector code that is now always present. No assembler
	  is needed to
	  build LAME any more.
	  
	  This changes which code path the encoder runs, so the output can
	  differ in the
	  last bits from a build that used the C path; it is
	  quality-neutral by
	  measurement rather than identical.

2026-07-22 12:51  aleidinger

	* Makefile.am, Makefile.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/doxygen/internal.md,
	  doc/maintainer-coverage.md, maintainer/coverage-mkinputs.py,
	  maintainer/coverage-report.py, maintainer/coverage-run.sh,
	  maintainer/coverage-workload.txt,
	  maintainer/gen-coverage-matrix.sh:
	  Add a code coverage harness for the maintainer
	  
	  Sanitizer runs report only on code that actually executes, and
	  they are
	  far too slow to run against every option combination the encoder
	  accepts.
	  Choosing which invocations to run under them was a matter of
	  judgement,
	  with no way to tell an invocation that reaches new code from one
	  that only
	  repeats what another already covered.
	  
	  Measure it instead. The harness builds an instrumented matrix
	  whose cells
	  are the axes of conditional compilation - a line inside a
	  preprocessor
	  conditional that no configuration enables cannot be reached by
	  any test,
	  however the test is written - runs a workload against each cell
	  while
	  capturing coverage per invocation, and reports what each
	  configuration
	  uniquely contributes together with a ranked, minimal set of
	  invocations.
	  
	  The workload deliberately includes invocations that are supposed
	  to fail.
	  The input-rejection paths are reachable only by an invocation
	  that exits
	  non-zero, so a harness that treats a non-zero exit as its own
	  failure
	  silently drops exactly the error handling a sanitizer run most
	  wants to
	  exercise. Every entry declares its expected outcome and a
	  mismatch in
	  either direction is reported.
	  
	  The option lists under test/ are read directly rather than
	  copied, so the
	  encoder-settings coverage curated there stays maintained in one
	  place.
	  
	  Maintainer tooling; none of it is needed to build or use LAME.

2026-07-22 11:27  aleidinger

	* configure:
	  regen

2026-07-22 11:26  aleidinger

	* configure.ac:
	  Report the detected libraries and features in the configure
	  summary
	  
	  The summary printed at the end of configure covered only the
	  version
	  type, the frontend linking, and whether the internal options and
	  the
	  unit tests were enabled. Everything else about the resulting
	  build - the
	  optional libraries that were found, which programs will actually
	  be
	  built, and the code generation settings - had to be read back out
	  of the
	  configure log.
	  
	  List them. The vector routines are named by the instruction sets
	  the
	  build compiles them for rather than by what the host CPU
	  supports, since
	  the two are not the same thing.
	  
	  libsndfile is reported with three outcomes rather than two:
	  asking for it
	  and not finding it falls back to the internal file io, which
	  until now
	  was visible only as a single warning in a long log.

2026-07-22 11:13  aleidinger

	* doc/html/history.html:
	  Add changelog entries for the frontend narrowing and prototype
	  cleanup
	  
	  Extends the existing warning-cleanup point with the 64 bit
	  narrowing
	  conversions and the obsolete empty parameter lists, crediting the
	  parts
	  that came from a contributed patch, and records the truncated
	  declared
	  length as an entry of its own.

2026-07-22 11:12  aleidinger

	* frontend/console.c, frontend/parse.c, frontend/parse.h:
	  Give the frontend's remaining empty parameter lists a prototype
	  
	  Four console helpers and one character-encoding helper are
	  defined with
	  an empty parameter list, the pre-standard form that declares
	  nothing
	  about what the function takes. For the console ones the header
	  already
	  declared them as taking no arguments, so each definition
	  disagreed with
	  its own prototype. Compilers that diagnose the obsolete form warn
	  at
	  every one of them.
	  
	  Declare them as taking no arguments.
	  
	  A declaration for a frontend entry point that has no definition
	  and no
	  callers anywhere in the tree is removed rather than converted.
	  The
	  header it sat in is internal to the frontend and is not
	  installed, so
	  nothing outside the tree can refer to it.
	  
	  The console conversions are from a patch by Maya R. Odinezenko
	  (SF patch
	  #106); the remaining conversion and the removal are not.

2026-07-22 11:10  aleidinger

	* frontend/get_audio.c:
	  Keep the careful-read sample counts at the width the API provides
	  
	  When told to read exactly as many samples as the input declares,
	  the
	  frontend held that declared count, and its running total of
	  samples
	  already read, in an unsigned int. The count reaches it from the
	  library
	  as an unsigned long. Where long is the wider type, a declared
	  length
	  above 2^32 samples was truncated before it was used, so reading
	  stopped
	  at the wrong point.
	  
	  Hold both at the source width instead. Only the final clamp still
	  narrows, and the test guarding it already bounds the value to
	  less than
	  one frame. Choosing the API's own type rather than a fixed-width
	  one
	  keeps this a no-op where long and int are the same width.
	  
	  The documented sentinel for an unknown length keeps its value at
	  either
	  width, so streams of unknown length are unaffected.
	  
	  SF patch #105 flagged the same conversion; it is addressed here
	  by
	  widening rather than by casting.

2026-07-22 11:09  aleidinger

	* frontend/get_audio.c, frontend/parse.c, frontend/timestatus.c:
	  Make the frontend's narrowing conversions to int explicit
	  
	  Where long is wider than int, the frontend implicitly narrows
	  long,
	  off_t and size_t values to int when storing sample counts,
	  decoder
	  delays, sample rates and argument lengths. Compilers that
	  diagnose
	  implicit narrowing warn at every one of them, so a warning-clean
	  frontend build is not possible on 64-bit targets.
	  
	  Cast at each site. Every value is bounded to the int range either
	  by
	  construction or by a range check already in place, so this states
	  the
	  intent of conversions that were already happening and does not
	  change
	  behaviour.
	  
	  Based on patches by Maya R. Odinezenko (SF patch #105, #106).

2026-07-22 11:09  aleidinger

	* libmp3lame/bitstream.c:
	  libmp3lame: avoid signed shift UB in the bitstream writer and CRC
	  
	  The MP3 writer assembles bytes by shifting a value into position
	  and OR-ing it
	  into the buffer. The value being placed can carry the sign bit,
	  so shifting it
	  left to position it is a left shift that overflows the signed int
	  - undefined,
	  though only the low bits are ever kept. The three writers that
	  share this
	  pattern (the two putbits variants and the header writer) now
	  shift through
	  unsigned, which is defined and leaves the stored bytes unchanged.
	  
	  The CRC-16 update likewise shifts its running value left each
	  step, and that
	  value grows past bit 30, making the signed shift a shift into the
	  sign bit.
	  Computing the CRC through unsigned locals makes every step
	  defined; the tested
	  bit and the polynomial XOR are unchanged, so the CRC is
	  identical.
	  
	  Output is bitstream-critical, so this was checked byte-for-byte
	  against the
	  unmodified library across steady-tone, transient/short-block, and
	  error-protected (CRC-bearing) encodes at several bitrate and
	  quality settings,
	  plus the decode paths - all identical. UBSan no longer reports
	  either line; the
	  baseline reproduced both, including the CRC shift, on the
	  error-protected
	  encode.

2026-07-22 11:08  aleidinger

	* libmp3lame/quantize.c:
	  libmp3lame: avoid signed shift UB in short-block scalefactor
	  amplification
	  
	  In the short-block subblock-gain path, a scalefactor is reduced
	  and, when it
	  goes negative, used to compute an amplification exponent: 210 +
	  (s << (scale +
	  1)). By that point s is always negative - the non-negative case
	  has already
	  continued - so the shift is a left shift of a negative value,
	  which UBSan
	  flags. The intent is s * 2^(scale + 1); s is a small negative
	  number that
	  cannot overflow, so writing it as a multiply by a power of two is
	  defined and
	  gives the identical result.
	  
	  Because this is quantization, output is bitstream-critical.
	  Verified
	  byte-identical against the unmodified library on transient,
	  short-block-heavy
	  material under -V2, -b 320, -q 0 and --preset insane (the
	  settings that reach
	  this path), as well as the steady-tone and tagged cases, and the
	  UBSan reports
	  for this line are gone (the baseline hit it six times on the
	  transient input).

2026-07-22 11:05  aleidinger

	* libmp3lame/id3tag.c:
	  libmp3lame: avoid signed shift UB in the ID3 frame-id matcher
	  
	  frame_id_matches walks a one-byte mask up through the four bytes
	  of a frame
	  id, shifting it left by 8 each step: 0xff, 0xff00, 0xff0000,
	  0xff000000. The
	  step that produces 0xff000000 overflows the signed int, and the
	  following one
	  shifts that negative value again - both undefined, and UBSan
	  flags the first.
	  The value is only ever used as a bitmask, so the sign is
	  irrelevant to the
	  result.
	  
	  Making the mask unsigned turns reaching and shifting past the top
	  byte into a
	  defined wrap and leaves every mask value, and therefore every
	  comparison and
	  the returned bits, exactly as before.
	  
	  Verified against the unmodified library: encoding with a full set
	  of ID3v2
	  tags (which is what drives this matcher) produces byte-identical
	  output, as do
	  the plain encode and decode paths, and the UBSan report for this
	  line is gone.

2026-07-22 11:04  aleidinger

	* frontend/get_audio.c:
	  frontend: avoid signed integer UB in the audio reader
	  
	  UBSan flags several integer operations in the WAV/AIFF/PCM reader
	  as
	  undefined, all value-dependent so a plain static pass cannot see
	  them. None
	  changes the output on a two's-complement machine, but each is
	  undefined by the
	  standard and worth making defined.
	  
	  unpack_read_samples assembles each PCM sample by shifting input
	  bytes into an
	  int. A byte of 0x80 or more shifted into the top position sets
	  the sign bit,
	  which is a left shift that overflows the signed int. The fix
	  shifts through
	  unsigned, which has the same width as int here (so every shift
	  amount stays in
	  range) and converts back to the same sample value. All eight
	  branches of the
	  byte-width/endianness ladder get the same treatment, not only the
	  three the
	  test corpus happened to reach, since they share the identical
	  defect.
	  
	  The MPEG decode path likewise shifts a 16-bit sample up into an
	  int; a negative
	  sample made that a left shift of a negative value, fixed the same
	  way.
	  
	  The WAV header size writer took a signed int. The streaming path
	  writes a
	  placeholder length of 0x7FFFFFFF, and the header adds the 44-byte
	  prefix to it,
	  overflowing the int. The size fields are unsigned by definition,
	  so the writer
	  now takes unsigned and the placeholder arithmetic is a defined
	  unsigned wrap -
	  the bytes emitted are unchanged.
	  
	  Verified value-identical against the unmodified reader:
	  8/16/32-bit WAV and
	  big-endian AIFF encode to byte-identical MP3, and decode to
	  byte-identical WAV
	  both to a file and to a pipe (the streaming placeholder path),
	  with the same
	  0x80000023 length still written. Frontend-only; the library and
	  the bitstream
	  are untouched.

2026-07-22 11:03  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_set_get.c:
	  test: add a unit test for the libmp3lame parameter API
	  
	  set_get.c is almost all getter/setter pairs, and the command-line
	  frontend
	  drives only a fraction of them: a coverage run of the encoder
	  leaves more
	  than half the file's lines unexecuted. Those lines are not dead -
	  they are
	  the public interface every library caller relies on - so they are
	  exactly
	  the part that most needs a test and had none.
	  
	  The new test exercises each setter three ways where it applies: a
	  valid
	  round-trip through the matching getter, a call on an invalid
	  handle (the
	  validity-guard tail every function shares), and an out-of-range
	  value (the
	  reject path the frontend never takes). It groups the functions by
	  contract
	  - plain round-trip, 0/1 boolean validation, range checks,
	  clamping, and
	  setters with side effects on other fields - and adds the
	  read-only,
	  init-populated getters and the surviving deprecated stubs. A few
	  of those
	  stubs are still exported but no longer prototyped in the public
	  header; the
	  test declares them locally so that slice of the ABI is covered
	  too.
	  
	  The internal-only tuning setters (masking adjust, ATH curve,
	  short-block
	  thresholds, substep, sfscale, subblock gain, and the like) are
	  covered
	  under the very gate the frontend uses for them: a normal build
	  compiles
	  that section out, while a build configured with --enable-internal
	  runs the
	  extra checks against the private set_get.h. Those symbols are
	  withheld from
	  the shared library, so under that option the test links the
	  static archive
	  to reach them. This half depends on the internal-options build
	  support, so
	  it must be applied after that change.
	  
	  This lifts set_get.c line coverage from roughly one half to 93%
	  on its own,
	  and gives the ABI a regression guard: a change that alters a
	  validation
	  rule, a clamp boundary, or an invalid-handle return now fails a
	  test
	  instead of silently shifting behaviour for downstream callers.

2026-07-22 11:02  aleidinger

	* configure, configure.ac, frontend/Makefile.am,
	  frontend/Makefile.in, libmp3lame/set_get.h:
	  build: repair the alpha internal-options build
	  
	  The frontend gates its developer-only tuning options (short-block
	  thresholds,
	  masking adjust, ATH curve, substep, sfscale, subblock gain, tune,
	  preset-notune) behind a switch that, until now, nothing turned
	  on: it was tied
	  to no configure option and not to the alpha version either. So
	  even in an alpha
	  release those options parsed and silently did nothing. Worse, if
	  the switch was
	  turned on by hand the frontend no longer linked at all, because
	  the setters it
	  then calls are deliberately absent from the shared library's
	  export list.
	  
	  This adds a configure option, --enable-internal, that makes the
	  feature real
	  and buildable. It is only meaningful for an alpha version, so
	  asking for it on a
	  beta or release tree is a hard configure error rather than a
	  silent no-op.
	  Because the internal setters are intentionally not exported, a
	  frontend that
	  uses them must link the static archive; --enable-internal
	  therefore forces
	  static frontend linking, and configure reports that it did so,
	  and why,
	  alongside a short summary of the resulting configuration. The
	  export list is
	  left untouched: a build with the option enabled produces a shared
	  library with
	  exactly the same public interface as one without it.
	  
	  The switch is defined per target - for the frontend and for the
	  parameter-API
	  unit test that also exercises these setters - rather than
	  globally. A global
	  definition would drag the internal code path of the argument
	  parser into the
	  other unit tests that compile it only for their argument helpers;
	  those link
	  the shared library, where the setters are not present. Keeping
	  the definition
	  local leaves them, and the library itself, exactly as they were.
	  
	  The internal header follows suit: the prototypes for these
	  setters are now
	  visible only when the option is enabled, matching the frontend.
	  The handful
	  that the library calls internally (the preset entry point, the
	  scalefactor-scale
	  setter, the M/S fix setter) stay visible in every build, since
	  intra-library
	  callers need them regardless.

2026-07-22 11:02  aleidinger

	* frontend/get_audio.c:
	  Scale IEEE-float samples by a single full-scale factor
	  
	  The float sample reader scaled positive and negative samples by
	  two
	  constants named after the largest and the smallest representable
	  integer. Neither name described its value: the negative one was
	  negated
	  into a positive magnitude, and the positive one was converted
	  from a
	  number a float cannot hold and rounded up. Both therefore stood
	  for
	  2^31, so the distinction the names implied did not exist.
	  
	  Use one constant named for what it is, the magnitude a full-scale
	  sample
	  maps to, and record why the surrounding clamps keep the
	  conversion in
	  range. The value is unchanged, so float input encodes
	  byte-identically.

2026-07-19 15:24  aleidinger

	* doc/html/history.html:
	  Add changelog entries for the warning cleanup
	  
	  Record the truncated-header and decode write-failure detection,
	  the
	  Windows short-to-long path fix, the warning-clean build across
	  GCC,
	  Clang and MSVC, and the stricter maintainer-mode warnings.

2026-07-19 15:23  aleidinger

	* configure, configure.ac:
	  Enable stricter warnings in maintainer mode
	  
	  When maintainer mode is configured, build the GCC/Clang tree with
	  -Wextra
	  and -Wshadow and treat all warnings as errors, so a lapse in
	  warning
	  cleanliness fails the build for developers. Regular and release
	  builds
	  are unaffected: maintainer mode is off by default, and each flag
	  is
	  probed for acceptance before use so an unsupported compiler skips
	  it.

2026-07-19 15:22  aleidinger

	* frontend/get_audio.c:
	  Use or drop the header fields the readers only skipped over
	  
	  Several WAV and AIFF header fields were parsed and then
	  discarded, and
	  two AIFF format identifiers were never referenced at all.
	  
	  Put the ones that carry a real invariant to work: an extensible
	  format
	  header now has to describe a large enough extension and may not
	  claim
	  more valid bits than its samples hold, and an AIFF file in 64 bit
	  floating point is named as unsupported instead of failing as if
	  it were
	  corrupt. The block alignment and byte rate only restate what the
	  other
	  fields say and are frequently written wrong, so they are reported
	  as a
	  mismatch under --verbose rather than rejected. The channel mask
	  is
	  advisory and is skipped, and the unreferenced identifier is
	  dropped.
	  
	  Well-formed files, including extensible ones, are unaffected.

2026-07-19 15:17  aleidinger

	* frontend/get_audio.c:
	  Detect and reject truncated WAV and AIFF headers
	  
	  The fixed-width byte readers used by the WAV and AIFF header
	  parsers
	  ignored the result of fread(), so a file ending in the middle of
	  a
	  header was parsed as if the missing bytes were zeros and could
	  drive the
	  encoder with bogus format parameters.
	  
	  Let the readers report a short read instead of returning a value
	  that
	  cannot be distinguished from data, and chain them at each call
	  site so
	  the first short read stops the remaining reads and rejects the
	  file
	  through the existing "corrupt or unsupported format" path.
	  Well-formed
	  files are unaffected.

2026-07-19 11:56  aleidinger

	* frontend/get_audio.c, libmp3lame/machine.h,
	  libmp3lame/mpglib_interface.c:
	  Mark intentionally-unused function parameters
	  
	  Several functions ignore some of their parameters: the
	  compatibility
	  decode entry points always fail, the analysis-hook setters are
	  not yet
	  implemented, and the MP3 reader and its file opener reach their
	  input
	  through global state or use their arguments only when the MPEG
	  decoder
	  is built in.
	  
	  Add a portable attribute macro for a deliberately-unused
	  parameter and
	  apply it at these sites. Because the attribute states that a
	  parameter
	  may go unused rather than that it does, it also covers the
	  arguments
	  that are live only in one of the two decoder configurations,
	  leaving a
	  build with extra parameter warnings quiet either way. The normal
	  build
	  is unaffected.

2026-07-19 11:50  aleidinger

	* libmp3lame/psymodel.h, libmp3lame/quantize_pvt.c:
	  Match array-parameter bounds between declaration and definition
	  
	  The energy array passed to the VBR psychoacoustic analysis holds
	  four
	  elements, but its declaration announced two, and the compiler
	  reported
	  the call it makes with that array as writing past the end. The
	  perceptual-entropy array in the bit-allocation helper was defined
	  with
	  an unspecified leading bound while its declaration gave one.
	  
	  State the bounds the callers actually pass. An array bound is not
	  part
	  of a parameter's type, so the generated code is unchanged.
	  
	  Also drop the declaration of a psychoacoustic entry point that
	  has no
	  definition and no callers; it carried the same stale bound.

2026-07-19 11:49  aleidinger

	* frontend/main.c:
	  Expand long filenames correctly on Windows
	  
	  Expanding the input path from its short 8.3 form copied the long
	  name
	  with a bound derived from the remaining buffer space rather than
	  from the
	  name, and neither branch guaranteed a terminating NUL. Where the
	  path had
	  no directory part this left the tail of the previous, longer name
	  in
	  place, and a name that did not fit overwrote the path with a
	  truncated
	  one instead of being rejected.
	  
	  Measure the name, establish where it goes, and copy it with a
	  bounded
	  conversion that always terminates, leaving the path untouched
	  when the
	  result would not fit. The search for the last separator no longer
	  subtracts pointers that may be null.

2026-07-19 11:46  aleidinger

	* frontend/get_audio.c, frontend/get_audio.h, frontend/lame_main.c:
	  Report output write failures when decoding to a file
	  
	  The decode path wrote the WAV header and the decoded samples
	  without
	  checking whether the writes succeeded, so a failure such as a
	  full
	  output device was ignored: the whole input was decoded into a
	  stream
	  that could accept nothing, and the program still exited
	  successfully
	  leaving a truncated or empty file.
	  
	  Let the sample writer report a failed write and stop the decode
	  at the
	  first one, reporting the error and exiting with a non-zero
	  status.
	  Successful decodes are unaffected.

2026-07-19 11:41  aleidinger

	* libmp3lame/VbrTag.c:
	  Avoid undefined signed left-shift when masking LAME-tag header
	  fields
	  
	  The bit-packing that assembles the LAME/Info tag frame header
	  formed its
	  low-bit field mask by complementing a left-shifted signed -1. A
	  left
	  shift of a negative value is undefined behaviour in C, and every
	  use of
	  the macro was flagged by -Wshift-negative-value.
	  
	  Form the mask from an unsigned all-ones value instead. The mask
	  value is
	  unchanged, so the emitted tag - and thus the whole encoded
	  bitstream -
	  is byte-identical; verified across VBR, CBR and ABR encodes.

2026-07-19 09:16  aleidinger

	* configMS.h, vc_solution/vs_lame_config.props:
	  Suppress the MSVC CRT deprecation warnings in the config header
	  
	  MSVC's /W4 deprecates the standard ISO C string and stdio
	  functions in
	  favour of its own non-portable _s replacements. Define
	  _CRT_SECURE_NO_WARNINGS in the MSVC config header, which every
	  translation unit includes before any C runtime header, so both
	  the nmake
	  and the project-file builds are covered from a single place, and
	  remove
	  the duplicate definition from the project-file build.

2026-07-19 09:16  aleidinger

	* frontend/lame_main.c, libmp3lame/presets.c:
	  Rename shadowing local variables
	  
	  Two inner-scope locals reused the name of a variable in an
	  enclosing
	  scope: the gain-adjustment temporary in the VBR preset code
	  shadowed the
	  interpolation fraction, and the ID3v2 tag write count shadowed
	  the
	  per-frame write count. Give each inner variable its own name so
	  no
	  declaration hides another.

2026-07-19 09:15  aleidinger

	* libmp3lame/bitstream.c:
	  Scope decode-time locals to the decode-on-the-fly build
	  
	  The ReplayGain analysis helper declared its config and gain-state
	  pointers unconditionally, but references them only inside the
	  DECODE_ON_THE_FLY block. In the default build they are unused.
	  Move
	  the declarations inside that block so they exist only where used.

2026-07-19 09:15  aleidinger

	* libmp3lame/machine.h:
	  Suppress the unreferenced-parameter warning under MSVC
	  
	  The intentional stub and compatibility entry points leave
	  parameters
	  unused, which MSVC reports at /W4. Disable that warning for the
	  MSVC
	  build alongside the existing suppressions, rather than annotating
	  each
	  site: the GCC/Clang build already flags a genuinely unused
	  parameter, so
	  that build stays the check for real cases.

2026-07-19 09:14  aleidinger

	* libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/vbrquantize.c, libmp3lame/vector/xmm_quantize_sub.c:
	  Annotate intentional switch fall-throughs
	  
	  Several switch statements handle a remainder count by
	  deliberately
	  falling from one case into the next. Mark each fall-through with
	  a
	  comment the compiler recognizes, so the intent is explicit and
	  the
	  fall-through warning stays quiet, replacing the two
	  static-analyzer
	  markers that the compiler does not accept.

2026-07-19 09:14  aleidinger

	* libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/set_get.c:
	  Compare float options against floating-point constants
	  
	  Several float encoder settings were compared against integer
	  literals
	  through the floating-point comparison macros, which applied the
	  floating-point absolute value to an integer operand. Give the
	  constants
	  floating-point type so the comparison stays in floating point;
	  the
	  compared values are unchanged.

2026-07-19 09:13  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/vbrquantize.c:
	  From 4c3af043dacc1abd00d1f5ac7db2eb70f2725372 Mon Sep 17 00:00:00
	  2001
	  From: Alexander Leidinger <Alexander@Leidinger.net>
	  Date: Sat, 18 Jul 2026 22:25:34 +0200
	  Subject: Remove or guard variables unused in the default build
	  
	  A loop counter in the LAME tag writer and a sample-rate local in
	  the
	  analysis hook are dead, and the float-conversion constants in the
	  VBR
	  quantizer are referenced only by the IEEE-754 fast path. Remove
	  the dead
	  locals and the unused rounding constant, and compile the
	  conversion
	  constants only when that fast path is built.
	  ---
	  lame/libmp3lame/VbrTag.c | 1 -
	  lame/libmp3lame/mpglib_interface.c | 1 -
	  lame/libmp3lame/vbrquantize.c | 5 ++---
	  3 files changed, 2 insertions(+), 5 deletions(-)
	  
	  diff --git a/lame/libmp3lame/VbrTag.c b/lame/libmp3lame/VbrTag.c
	  index 2ea6dda..baedcc2 100644
	  --- a/lame/libmp3lame/VbrTag.c
	  +++ b/lame/libmp3lame/VbrTag.c
	  @@ -593,7 +593,6 @@ PutLameVBR(lame_global_flags const *gfp,
	  size_t nMusicLength, uint8_t * pbtStrea
	  SessionConfig_t const *const cfg = &gfc->cfg;
	  
	  int nBytesWritten = 0;
	  - int i;
	  
	  int enc_delay = gfc->ov_enc.encoder_delay; /* encoder delay */
	  int enc_padding = gfc->ov_enc.encoder_padding; /* encoder padding
	  */
	  diff --git a/lame/libmp3lame/mpglib_interface.c
	  b/lame/libmp3lame/mpglib_interface.c
	  index b81a60d..5cb834f 100644
	  --- a/lame/libmp3lame/mpglib_interface.c
	  +++ b/lame/libmp3lame/mpglib_interface.c
	  @@ -513,7 +513,6 @@ void hip_finish_pinfo(hip_t hip)
	  #ifndef NOANALYSIS
	  #ifdef HAVE_MPG123
	  struct mpg123_frameinfo fi;
	  - long rate;
	  plotting_data *pinfo = hip->pinfo;
	  if(!hip || !hip->pinfo)
	  return;
	  diff --git a/lame/libmp3lame/vbrquantize.c
	  b/lame/libmp3lame/vbrquantize.c
	  index 773824c..f979517 100644
	  --- a/lame/libmp3lame/vbrquantize.c
	  +++ b/lame/libmp3lame/vbrquantize.c
	  @@ -105,11 +105,10 @@ typedef VOLATILE union {
	  # define XRPOW_FTOI(src,dest) ((dest) = (int)(src))
	  #endif
	  
	  +#if TAKEHIRO_IEEE754_HACK
	  static int const MAGIC_INT = MAGIC_INT_def;
	  -#if !TAKEHIRO_IEEE754_HACK
	  -static DOUBLEX const ROUNDFAC = ROUNDFAC_def;
	  -#endif
	  static DOUBLEX const MAGIC_FLOAT = MAGIC_FLOAT_def;
	  +#endif
	  
	  
	  inline static float
	  --
	  2.55.0.windows.3

2026-07-19 09:13  aleidinger

	* libmp3lame/VbrTag.c:
	  Remove always-true null checks in the VBR tag reader
	  
	  The VBR tag reader guarded its table-of-contents copy with a test
	  that
	  the toc member is non-NULL. That member is a fixed array, so its
	  address
	  can never be NULL and the test is always true. Drop the dead
	  guard at
	  both the parsing and the debug-dump sites.

2026-07-19 09:12  aleidinger

	* libmp3lame/psymodel.c:
	  Avoid an uninitialized read in the VBR block-type decision on
	  mono
	  
	  When encoding a single channel, the VBR block-type decision still
	  consults a two-channel long/short-block array whose second entry
	  is
	  never filled, so the channel-coupling test reads uninitialized
	  stack.
	  Depending on that value it could spuriously force short blocks on
	  a mono
	  frame.
	  
	  Initialize both entries to the neutral long-block value. The real
	  channels are overwritten by the attack detector, so stereo is
	  unaffected and the absent second channel no longer influences the
	  coupled decision. Encoder output is unchanged for well-formed
	  input.

2026-07-19 09:12  aleidinger

	* maintainer/gen-build-matrix.sh:
	  gen-build-matrix.sh: build each matrix cell in its own directory
	  
	  The generated per-cell build script and the driver both ran from
	  the
	  matrix master directory, so every cell configured and built there
	  rather
	  than in its own out-of-tree directory, each overwriting the
	  previous
	  cell's configuration and object files. The matrix therefore never
	  tested
	  the configurations in isolation: a cell's pass/fail and its build
	  warnings reflected whatever survived the last cell's clobbering.
	  
	  Have each generated build script change into its own directory
	  before
	  configuring, so every cell is a genuine isolated out-of-tree
	  build.

2026-07-18 10:29  aleidinger

	* doc/html/history.html:
	  Add changelog entries for the length and play-length fixes
	  
	  Record the unknown-length sentinel recognition (external
	  contribution),
	  the frame-count estimate overflow guards, and writing the full
	  audio
	  play-length in the ID3v2 tag.

2026-07-18 10:29  aleidinger

	* libmp3lame/id3tag.c, test/unit/test_id3tag.c:
	  Write the full audio play-length in the ID3v2 tag
	  
	  The TLEN duration, derived from the declared sample count, was
	  clamped
	  to a 32-bit millisecond count of about 49.7 days - far below what
	  the
	  field and the tag format allow. Store it at the field's full
	  width so
	  a longer declared length is written in full where the platform
	  can
	  represent that many samples; shorter lengths are unchanged.

2026-07-18 10:28  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_totalframes.c:
	  Add a unit test for the length-estimate edge cases
	  
	  Cover lame_get_totalframes() for the documented unknown-length
	  sentinel, which must report unknown rather than an estimate, and
	  for a
	  declared length whose frame count would exceed the int the
	  estimate is
	  returned in, which must not overflow it. Both go through the
	  public
	  API. The overflow case is only reachable where unsigned long is
	  wider
	  than int and is skipped elsewhere.

2026-07-18 10:25  aleidinger

	* frontend/get_audio.c, frontend/timestatus.c,
	  libmp3lame/set_get.c:
	  From 22bfd07453ceedb9c32b9fdd9da8341ac0152f8a Mon Sep 17 00:00:00
	  2001
	  From: Alexander Leidinger <Alexander@Leidinger.net>
	  Date: Fri, 17 Jul 2026 23:15:38 +0200
	  Subject: Bound the frame-count estimates to their int fields
	  
	  Several places derive a frame count from an unsigned long sample
	  or
	  frame count and store it in an int: the total-frames estimate for
	  the
	  encoder and for the libmpg123 decoder, and the decoder progress
	  display. Where long is wider than int a declared length large
	  enough
	  to exceed the int overflows it; the resampling branch of the
	  encoder
	  estimate already guarded against this, the others did not. Bound
	  each
	  to its field's range, yielding the unknown estimate past the
	  limit as
	  the guarded branch does.
	  ---
	  lame/frontend/get_audio.c | 2 +-
	  lame/frontend/timestatus.c | 11 ++++++++---
	  lame/libmp3lame/set_get.c | 2 ++
	  3 files changed, 11 insertions(+), 4 deletions(-)
	  
	  diff --git a/lame/frontend/get_audio.c
	  b/lame/frontend/get_audio.c
	  index 2f8df10..5ecda2e 100644
	  --- a/lame/frontend/get_audio.c
	  +++ b/lame/frontend/get_audio.c
	  @@ -2034,7 +2034,7 @@ int lame123_decode_initfile(FILE *fd,
	  mp3data_struct *mp3data, int *enc_delay, i
	  if libmpg123 got that info from Info tag or not. */
	  /* I am paranoid about off_t being larger than long or int. */
	  len = mpg123_framelength(global.hip->mh);
	  - if(len <= (unsigned long)-1)
	  + if(len <= ((unsigned int)-1)/2) /* totalframes is int, bound to
	  INT_MAX */
	  mp3data->totalframes = len;
	  else
	  return -1;
	  diff --git a/lame/frontend/timestatus.c
	  b/lame/frontend/timestatus.c
	  index e6f1a71..c68d5b3 100644
	  --- a/lame/frontend/timestatus.c
	  +++ b/lame/frontend/timestatus.c
	  @@ -382,16 +382,21 @@ decoder_progress_init(unsigned long n, int
	  framesize)
	  /* also recognize the documented unknown sentinel; the (0ul-1ul)
	  test
	  alone matches it only where long is 32-bit */
	  if (n != MAX_U_32_NUM && n != (0ul-1ul)) {
	  + unsigned long frames_total;
	  if (framesize == 576 || framesize == 1152) {
	  - dp->frames_total = calcNumBlocks(n, framesize);
	  + frames_total = calcNumBlocks(n, framesize);
	  dp->samples = 576 + calcEndPadding(n, framesize);
	  }
	  else if (framesize > 0) {
	  - dp->frames_total = n / framesize;
	  + frames_total = n / framesize;
	  }
	  else {
	  - dp->frames_total = n;
	  + frames_total = n;
	  }
	  + /* frames_total is an int; a length that would overflow it
	  leaves the
	  + total unknown, as the sentinel above does */
	  + if (frames_total <= ((unsigned int)-1)/2)
	  + dp->frames_total = frames_total;
	  }
	  return dp;
	  }
	  diff --git a/lame/libmp3lame/set_get.c
	  b/lame/libmp3lame/set_get.c
	  index 76b99c1..0a748a2 100644
	  --- a/lame/libmp3lame/set_get.c
	  +++ b/lame/libmp3lame/set_get.c
	  @@ -2163,6 +2163,8 @@ lame_get_totalframes(const
	  lame_global_flags * gfp)
	  pcm_samples_to_encode = ceil(resampled_samples_to_encode);
	  }
	  else {
	  + if (pcm_samples_to_encode / pcm_samples_per_frame >= (unsigned
	  long)(INT_MAX-2))
	  + return 0; /* overflow, happens eventually, no estimate! */
	  frames = pcm_samples_to_encode / pcm_samples_per_frame;
	  pcm_samples_to_encode -= frames * pcm_samples_per_frame;
	  }
	  --
	  2.55.0.windows.3

2026-07-18 10:24  aleidinger

	* frontend/timestatus.c, libmp3lame/set_get.c:
	  Recognize the documented unknown sample-count sentinel
	  
	  The frontend and libmp3lame test num_samples against (0ul-1ul) to
	  detect "length unknown", but the documented default and the
	  unknown
	  sentinel used elsewhere in the code is 2^32-1 (lame.h: "default =
	  2^32-1"; MAX_U_32_NUM in util.h and the frontend). The two values
	  coincide only where long is 32 bits, so on LP64 the unknown case
	  is
	  never recognized: lame_get_totalframes() returns an estimate
	  instead
	  of 0, and stdin encodes show a bogus ~3.7M-frame total with a
	  phantom
	  ETA.
	  
	  Recognize MAX_U_32_NUM in both places. The (0ul-1ul) comparison
	  is
	  kept to preserve existing LP64 behavior. Values merely larger
	  than the
	  sentinel -- legitimate on LP64, e.g. the raw-input file-size
	  fallback
	  for very large files -- keep their real estimates.
	  
	  Tested on Apple Silicon (LP64): the sentinel and the default now
	  report unknown; 53113, 0x100000000 and ULONG_MAX behave as
	  before;
	  encoded output is byte-identical. Where long is 32 bits the new
	  test
	  collapses to the old one by value equality.
	  
	  
	  Patch by Maya R. Odinezenko <maya@elevatorium.org>

2026-07-17 18:13  aleidinger

	* doc/html/history.html:
	  doc: add changelog entries for the mono-interleaved and related
	  fixes
	  
	  Record the interleaved-encode rejection on a mono session, the
	  decoder
	  rescan loop, the terminal capability-string buffer, the ID3v2
	  tag-size
	  limit and the mp3rtp error-path close.

2026-07-17 18:13  aleidinger

	* frontend/mp3rtp.c:
	  frontend: drop closes of an unopened file on the mp3rtp error
	  paths
	  
	  mp3rtp opens its output file only after the RTP socket is set up
	  and the
	  argument vector is rebuilt. Two early failure paths - the socket
	  setup
	  failing and the argument-vector allocation failing - closed that
	  file
	  first, on an uninitialized handle. Remove both closes; nothing is
	  open
	  there yet.

2026-07-17 18:12  aleidinger

	* frontend/console.c:
	  frontend: let the terminal library store the capability string
	  
	  get_termcap_string filled a fixed-size local buffer through
	  tgetstr,
	  which takes no length for the area it writes, so a terminal whose
	  capability string was longer than the buffer overran it. Pass a
	  null
	  area so the library returns the string from its own storage, and
	  copy
	  that out under the existing bound.

2026-07-17 18:12  aleidinger

	* libmp3lame/id3tag.c, test/unit/test_id3tag.c:
	  libmp3lame: refuse an ID3v2 tag too large for its size field
	  
	  The ID3v2 tag length is stored in a 28-bit synchsafe field. Album
	  art or
	  padding supplied by the caller can push the tag past that, at
	  which point
	  the four length bytes wrap and describe a shorter tag than was
	  written,
	  leaving the frames after it unparseable. Produce no tag in that
	  case
	  rather than a corrupt one.
	  
	  The unit tests drive the size over the limit through both the
	  padding and
	  the album-art paths and confirm an ordinary tag is still written.

2026-07-17 18:11  aleidinger

	* Makefile.unix, frontend/Makefile.am, frontend/Makefile.in,
	  frontend/amiga_mpega.c, frontend/get_audio.c, frontend/parse.c,
	  include/lame.h, mac/Precompile_Common.h, mpglib/mpglib.h,
	  test/unit/test_get_audio_aiff.c:
	  frontend: remove the unused bundled and Amiga decoder paths
	  
	  Decoding has gone through libmpg123 since 3.101, and the older
	  bundled
	  mpglib and Amiga MPEGA decoder paths are no longer selected by
	  any build.
	  Remove the dead code they left: the mpglib header scan and frame
	  reader
	  in the frontend, the Amiga decoder source, the stale build flags
	  and
	  header comments, and the legacy hand-written Unix makefile that
	  still
	  enabled them.
	  
	  The --mp1input, --mp2input and --mp3input options depend on the
	  decoder,
	  so offer them only when it is built in; they were previously
	  listed under
	  the removed bundled-decoder macro and so never shown at all.
	  Remove the
	  long-deprecated --ogginput option, which only printed a notice.
	  
	  The public decode interface stays; it is implemented on top of
	  libmpg123
	  and keeps working, and libmp3lame's exported symbols are
	  unchanged.

2026-07-17 18:10  aleidinger

	* libmp3lame/lame.c, test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_encode_mono_interleaved.c:
	  libmp3lame: reject an interleaved encode on a mono session
	  
	  The interleaved encoding entry points read the left and right
	  channels
	  from one buffer with a stride of two. On a session opened with a
	  single
	  input channel that buffer holds only one channel, and nothing
	  bounds the
	  strided read, so the second channel is read a channel's worth of
	  samples
	  past the end of the buffer.
	  
	  Refuse the combination with an input-data error rather than read
	  outside
	  the caller's buffer. A mono session is fed through the
	  non-interleaved
	  entry points, which is what the interleaving front ends already
	  do.
	  
	  The unit tests cover each interleaved entry point on a mono
	  session, and
	  check that an ordinary mono encode and a stereo interleaved
	  encode are
	  left working.
	  
	  Reported on SourceForge (bug #522).

2026-07-17 12:59  aleidinger

	* doc/html/history.html:
	  doc: add changelog entries for the noise shaping and termcap
	  fixes

2026-07-17 12:58  aleidinger

	* configure, configure.ac:
	  Detect the terminal library by the interface the console code
	  uses
	  
	  The probe asked each candidate library for initscr, which nothing
	  in LAME
	  calls. The console code drives the terminal through the termcap
	  interface,
	  tgetent, tgetnum and tgetstr, and selects its header from
	  whichever one is
	  present. A library can offer initscr and still fail to satisfy
	  that header:
	  where ncurses is packaged as a static library whose termcap
	  header declares
	  those entry points as DLL imports, the probe succeeded and the
	  frontend
	  then failed to link against the library the probe had picked.
	  
	  Compiling the same header and calls the console code compiles
	  asks the
	  question the link step will ask. The candidates are also tried in
	  the order
	  they are written, keeping the first that satisfies the test
	  rather than the
	  last to be tested, which is what the unguarded checks amounted
	  to.
	  
	  Where none of them satisfies it, terminal support stays disabled,
	  a case
	  the console code already accounts for, instead of breaking the
	  build.
	  
	  SourceForge bug #436.

2026-07-17 12:57  aleidinger

	* USAGE, doc/html/detailed.html, doc/man/lame.1, libmp3lame/lame.c:
	  libmp3lame: disable amplified noise shaping for CBR and ABR
	  
	  Amplified noise shaping degrades CBR and ABR encoding instead of
	  improving it, which is why the faster quality settings, the ones
	  that
	  leave it off, sound better than the slower ones. It applies to
	  the
	  default quality setting.
	  
	  The amplification weighs bands against a distortion measure
	  produced by
	  the psychoacoustic model. CBR and ABR were moved onto the newer
	  model
	  without adapting their encoding loops to it, so that measure no
	  longer
	  means to them what the amplification assumes. VBR was written
	  against
	  that model and is unaffected, so it keeps the amplification.
	  
	  Keeping the amplification out of the two modes it hurts treats
	  the
	  symptom; reconciling those encoding loops with the psychoacoustic
	  model
	  is still to be done.
	  
	  For these two modes the amplification was the only difference
	  between
	  quality settings 3 and 4, and between 1 and 2, so each of those
	  pairs
	  selects the same encoding.
	  
	  Reported and analysed by Maik Merten (SF bug #516).

2026-07-17 12:46  aleidinger

	* doc/html/history.html:
	  doc: remove a stray end tag from the changelog

2026-07-17 12:46  aleidinger

	* doc/html/history.html:
	  doc: add changelog entries for the 4.1 changes

2026-07-17 12:45  aleidinger

	* libmp3lame/quantize.c:
	  libmp3lame: record the accepted quantization's own bit count as
	  the best
	  
	  The outer search loop filled its "best bit count" from the
	  incumbent one
	  statement before the accepted candidate replaced it, so the value
	  trailed the
	  search by one accepted candidate, while the adjacent noise
	  bookkeeping recorded
	  the correct one. The bound derived from it in the no-distortion
	  branch was
	  biased tight, because scalefactor amplification trends toward
	  more bits.
	  
	  Take the count from the candidate being accepted. VBR output is
	  unchanged; CBR
	  and ABR output changes, with no measurable difference in
	  perceived quality.

2026-07-17 12:45  aleidinger

	* libmp3lame/vbrquantize.c:
	  libmp3lame: keep VBR scalefactors above their minimum when
	  reducing bits
	  
	  When a granule runs out of bits the reduction step flattens the
	  scalefactor
	  distribution toward the global gain, but it bounded the result
	  only to the
	  representable range, not to each band's minimum scalefactor.
	  Under bit
	  starvation a band could then be handed a scalefactor below the
	  minimum the
	  allocator requires - the sibling reduction path already guards
	  against this.
	  Apply the same lower bound here, so every path that feeds the
	  allocator keeps
	  each band at or above its minimum.

2026-07-17 12:42  aleidinger

	* frontend/parse.c:
	  frontend: guard setlocale() the same way as its header include
	  
	  parse.c includes <locale.h> only when both HAVE_ICONV and
	  HAVE_LANGINFO_H
	  are defined, but the setlocale(LC_CTYPE, "") call was guarded on
	  HAVE_LANGINFO_H alone. A configuration with langinfo but without
	  iconv then
	  compiled the call with neither setlocale() nor LC_CTYPE declared,
	  which
	  breaks the frontend build (seen on macOS with a C23 default). The
	  call only
	  prepares nl_langinfo() for the iconv conversions, so require
	  iconv for it as
	  well, matching the include.

2026-07-17 12:42  aleidinger

	* libmp3lame/encoder.c:
	  libmp3lame: convert analyzer spectrum data to double when copying
	  it
	  
	  The frame analyzer's per-band spectrum buffer holds double
	  values, but the
	  encoder-side copy filled it from the encoder's FLOAT array with a
	  raw byte
	  copy, which reinterprets the bit patterns instead of converting
	  them. In the
	  usual build where FLOAT is float the analyzer's MDCT display
	  therefore showed
	  garbage. Copy element by element so the values are converted,
	  correct for
	  either width of FLOAT.

2026-07-17 12:41  aleidinger

	* libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  libmp3lame: check the resampler's buffer allocations
	  
	  The resampler allocated its FIR input and filter buffers without
	  checking
	  the result and then dereferenced them while precomputing the
	  filter
	  coefficients, turning an allocation failure into a crash. Check
	  the
	  allocations, release any partial allocation, and report the
	  failure to
	  the caller as out of memory.

2026-07-17 12:40  aleidinger

	* doc/html/Makefile.am, doc/html/Makefile.in:
	  doc: install and package the HTML documentation assets
	  
	  The HTML manual references a stylesheet and several images, but
	  only the
	  pages themselves were installed and packaged - the images and the
	  stylesheet
	  were left behind, so both the installed copy and the release
	  tarball rendered
	  the documentation without them.
	  
	  List the images and the stylesheet alongside the pages so they
	  are installed
	  into the same tree and included in the distribution.

2026-07-16 14:35  aleidinger

	* vc_solution/Makefile.am, vc_solution/Makefile.in,
	  vc_solution/README.vs.txt, vc_solution/vs_lame.sln,
	  vc_solution/vs_lame.slnx, vc_solution/vs_lame_clients.sln,
	  vc_solution/vs_lame_clients.slnx:
	  Convert the Visual Studio solutions to the .slnx format
	  
	  The two solution files move from the GUID-keyed format to the XML
	  .slnx one,
	  which lists projects by path and drops the duplicated
	  identifiers. Building
	  them needs Visual Studio 17.14, or the 17.13 Build Tools; the
	  project files are
	  unchanged and still open in older versions. The mp3x analyzer
	  stays unselected,
	  as it was.
	  
	  Based on a patch from Michael Fink (SF patch #103), adapted.

2026-07-16 14:34  aleidinger

	* Makefile.am, Makefile.in, doc/maintainer-build-matrix.md,
	  maintainer/gen-build-matrix.ps1,
	  maintainer/setup-windows-deps.ps1:
	  Cover the optional Windows libraries in the build matrix
	  
	  The native-Windows build matrix built the default configuration
	  and, when
	  pointed at an mpg123, the decoder. libsndfile and the mp3x
	  analyzer went
	  unbuilt, so a change could break either without the matrix
	  noticing. Add a cell
	  for each optional library on both the nmake and MSBuild paths,
	  generated only
	  when the library is present, so the matrix still runs with
	  whatever is
	  installed. A companion setup-windows-deps.ps1 lays the libraries
	  out where both
	  the builds and the matrix look for them, from archives fetched by
	  hand.

2026-07-16 14:33  aleidinger

	* Makefile.MSVC, vc_solution/.gitignore, vc_solution/README.vs.txt,
	  vc_solution/vs_gtk_config.props:
	  Make the mp3x analyzer buildable again on Windows
	  
	  The mp3x frame analyzer needs GTK1, for which gtk1-win is the
	  only current
	  Windows port. Neither build could compile against it: the Visual
	  Studio project
	  omitted the folders holding glib's and gdk's generated config
	  headers, and the
	  nmake build pointed at a layout no download provides. Point both
	  at the
	  gtk1-win tree and gate it behind a HaveGtk switch, mirroring
	  mpg123 and
	  libsndfile, so a normal build is unaffected and mp3x stays
	  unselected by
	  default.
	  
	  Two further defects surfaced once mp3x compiled. gtk1-win ships
	  headers named
	  like LAME's own (version.h, config.h), so the nmake build has to
	  search its own
	  include directories first. And the mp3x and mp3rtp targets
	  depended on the
	  phony that only builds the objects, not the target that archives
	  the static
	  library, so they linked only when another target built it first.
	  
	  Based on a patch from Michael Fink (SF patch #103), adapted.

2026-07-16 14:32  aleidinger

	* Makefile.MSVC, vc_solution/README.vs.txt,
	  vc_solution/vs_libsndfile_config.props:
	  Update the Windows build for current libsndfile
	  
	  Current libsndfile Windows releases ship sndfile.lib and
	  sndfile.dll as a
	  plain zip, where the build still named the old libsndfile-1
	  files. Linking
	  against a current libsndfile therefore failed on both Windows
	  build paths,
	  and the nmake path had no include directory for it at all. Name
	  the current
	  files, give the nmake build the missing include path, and point
	  the
	  instructions at the current download.
	  
	  Based on a patch from Michael Fink (SF patch #103), adapted.

2026-07-16 14:31  aleidinger

	* Makefile.am, Makefile.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/doxygen/internal.md,
	  doc/maintainer-quality.md, maintainer/quality-compare.py,
	  maintainer/setup-quality-corpus.sh:
	  Add a quality regression harness
	  
	  A change to the psycho acoustic model or the quantizer is meant
	  to alter the
	  output, so the bitstream check that guards everything else says
	  nothing about
	  whether it was an improvement. Judging that by ear does not
	  survive review
	  either.
	  
	  One script lays out reference audio from the EBU SQAM CD, which
	  is not
	  shipped: its licence covers testing but not clearly
	  redistribution, so it is
	  fetched by hand and pointed at. The other encodes that audio with
	  a baseline
	  and a candidate build, scores each against the original with
	  PEAQ, and reports
	  the difference per track. Per track rather than as a total,
	  because a
	  regression in one hard case disappears into a corpus average -
	  which is what
	  the transient tracks in the default selection are there to catch.
	  
	  The scorer is not part of any distribution and has to be built;
	  where it is
	  missing the script says so and where to get it. Its scores are a
	  relative
	  signal between two builds, not a measure of absolute quality: no
	  free PEAQ
	  implementation meets the conformance requirements.

2026-07-16 14:30  aleidinger

	* test/CBRABR.op, test/Makefile.am, test/Makefile.in, test/VBR.op,
	  test/VBRold.op, test/misc.op, test/shortVBR.op:
	  Cover the untested quality options
	  
	  The option files drove the bitrate and quality axes hard but left
	  whole
	  families of the encoder untouched.
	  
	  Of the stereo modes only mono and plain stereo were exercised;
	  joint stereo
	  is the default, so it was covered by every line that says
	  nothing, but forced
	  mid/side and dual channel were not reached at all, and neither
	  were the two
	  modes that fold a single channel down through the channel
	  scalers.
	  
	  The old variable bitrate routine had no coverage whatsoever. It
	  gets a file
	  of its own, mirroring the sweep the default routine already gets,
	  as it is a
	  separate implementation rather than a setting. --vbr-new needs no
	  such file:
	  it selects the same routine as the default.
	  
	  The remaining gaps - the hearing threshold settings, block
	  switching, the
	  high pass, and the presets - are single settings rather than an
	  axis to
	  sweep, so they get one line each alongside the other one-off
	  options.

2026-07-16 14:29  aleidinger

	* test/lametest.py:
	  Port the regression script to Python 3
	  
	  The script has not run for years: it is written for Python 2,
	  whose print
	  statement and commands module Python 3 removed outright, and no
	  interpreter
	  still ships them.
	  
	  It builds its command lines as argument lists rather than shell
	  strings, so a
	  path with a space in it survives, and counts the differing bytes
	  itself
	  instead of driving cmp and wc through a shell, which the two
	  comments asking
	  for it did not have the module to do. Decoding for comparison now
	  uses the
	  encoder under test rather than whichever lame is installed.

2026-07-16 14:27  aleidinger

	* Makefile.am, Makefile.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/doxygen/internal.md,
	  doc/maintainer-build-matrix.md, doc/maintainer-perf.md,
	  maintainer/perf-compare.sh:
	  Add a script to compare the encoding speed of two builds
	  
	  Timing an encoder by hand invites two mistakes that quietly
	  produce a wrong
	  answer. With dynamic frontends the encoder in the build directory
	  is a
	  libtool wrapper, so a naive run times a shell start-up; its
	  executable, used
	  directly, binds to an installed libmp3lame where the host has one
	  and then
	  measures that library rather than the build being tested.
	  
	  The script takes a baseline and a candidate build matrix, runs
	  both encoders
	  over the same input, and reports the difference along with
	  whether the
	  measurement varied too much to mean anything. It compares the
	  bitstreams in
	  the same pass, because a speedup that changes the output is a bug
	  unless the
	  change was meant to.

2026-07-16 14:26  aleidinger

	* .clang-format, Makefile.am, Makefile.in:
	  Add the code style as a clang-format configuration

2026-07-16 14:24  aleidinger

	* doc/maintainer-build-matrix.md, maintainer/gen-build-matrix.ps1,
	  vc_solution/vs_lame.sln, vc_solution/vs_lame_config.props,
	  vc_solution/vs_lame_mp3rtp.vcxproj,
	  vc_solution/vs_libmpg123_config.props:
	  Make the native Windows build matrix run
	  
	  None of the decoder-on cells and neither 32-bit MSBuild cell
	  could build.
	  
	  The matrix pointed every project in the solution at one
	  intermediate
	  directory, so they overwrote each other's objects and fought over
	  one program
	  database. The projects already place those per configuration and
	  per project;
	  they now take a base directory the matrix can move per cell
	  instead.
	  
	  The analyzer was still built on 32-bit, where it needs a GTK that
	  this
	  project does not ship. It is now left out of the build the way it
	  already is
	  on 64-bit, and stays in the solution.
	  
	  mp3rtp linked the decoder-enabled library without the decoder,
	  and a single
	  -Arch value ended up indexed as a string, which selected a
	  platform of "W".
	  mp3rtp also picks up the mpg123 path the way the other consumers
	  of the
	  library do, so it can be quoted like them.
	  
	  Cells are only emitted for what is present: the nmake cell needs
	  the import
	  library, which the MSBuild cell generates for itself.

2026-07-16 14:23  aleidinger

	* Makefile.MSVC:
	  Fix the include path of the mpg123 decoder for the nmake build
	  
	  Building with MPG123=YES never worked: the directory it points at
	  ends in a
	  backslash, which escapes the closing quote of the include option,
	  so the
	  compiler swallowed the source file name and stopped. The paths
	  now bring
	  their own separator, which also lets the directory be quoted
	  against spaces.

2026-07-16 14:23  aleidinger

	* include/lame.h, libmp3lame/lame.c, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_encode_nonfinite.c:
	  Reject PCM input which is not a finite number
	  
	  A NaN or an infinity has no meaning as an audio sample, but the
	  floating point
	  entry points pass whatever they are given into the psycho
	  acoustic model, where
	  it spreads across the band energies and turns the surrounding
	  frames into noise.
	  A debug build then aborts deep inside the model, far away from
	  the buffer which
	  caused it, and a release build encodes the result without a word.
	  Callers
	  computing their samples rather than reading them from a file - a
	  resampler
	  dividing by zero, a decoder passing on a corrupt stream - can
	  produce such
	  values without noticing.
	  
	  Screen the floating point sample types while the data can still
	  be attributed
	  to the caller's buffer, and refuse the call with an error code of
	  its own,
	  which leaves the ABI intact. Integer sample types cannot express
	  a non-finite
	  value and are taken as given, so those entry points are untouched
	  and the
	  encoded bitstream does not change. Refusal is an ordinary
	  documented outcome,
	  so it returns rather than asserting.
	  
	  The screening inspects the exponent bits instead of calling
	  isfinite(): the
	  library is built with fast floating point maths, under which the
	  compiler may
	  assume that no such value occurs and fold the check away. It
	  costs about a
	  tenth of a percent of the encoding time.
	  
	  The unit tests also pin the input ranges the entry points
	  document, which a
	  suite built only from NaNs could not tell apart from a screen
	  which rejected
	  everything.

2026-07-16 14:22  aleidinger

	* libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/lameerror.h, vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters:
	  Remove an unused file
	  
	  lameerror.h duplicates the error code enumeration of the public
	  header, under
	  the same name, and is included by nothing.

2026-07-16 14:20  aleidinger

	* libmp3lame/lame.c:
	  Drop three unreachable bitrate index fallbacks
	  
	  Each of the three bitrate index lookups in lame_init_params()
	  runs on a value
	  FindNearestBitrate() has just taken from the bitrate table, and
	  then clamps the
	  index in case the lookup failed. It cannot fail: both functions
	  work on the same
	  table row, derive the MPEG version from the sample rate the same
	  way, and the
	  value is handed straight from one to the other. A sweep over
	  every version and
	  sample rate, for bitrates from -10000 to 10000 kbps, reaches no
	  clamp.

2026-07-14 18:31  aleidinger

	* vc_solution/Makefile.in:
	  regen

2026-07-14 18:29  aleidinger

	* ., vc_solution/.gitignore, vc_solution/Makefile.am,
	  vc_solution/README.vs.txt, vc_solution/README.vs2026.txt,
	  vc_solution/vs2026_arch_nasm.props,
	  vc_solution/vs2026_arch_sse2.props,
	  vc_solution/vs2026_arch_x87.props,
	  vc_solution/vs2026_gtk_config.props, vc_solution/vs2026_lame.sln,
	  vc_solution/vs2026_lame_acm.vcxproj,
	  vc_solution/vs2026_lame_acm.vcxproj.filters,
	  vc_solution/vs2026_lame_acm_adbg.vcxproj,
	  vc_solution/vs2026_lame_acm_adbg.vcxproj.filters,
	  vc_solution/vs2026_lame_acm_tinyxml.vcxproj,
	  vc_solution/vs2026_lame_clients.sln,
	  vc_solution/vs2026_lame_config.props,
	  vc_solution/vs2026_lame_dshow.vcxproj,
	  vc_solution/vs2026_lame_dshow.vcxproj.filters,
	  vc_solution/vs2026_lame_enc_dll.vcxproj,
	  vc_solution/vs2026_lame_enc_dll.vcxproj.filters,
	  vc_solution/vs2026_lame_lame.vcxproj,
	  vc_solution/vs2026_lame_lame.vcxproj.filters,
	  vc_solution/vs2026_lame_mp3rtp.vcxproj,
	  vc_solution/vs2026_lame_mp3rtp.vcxproj.filters,
	  vc_solution/vs2026_lame_mp3x.vcxproj,
	  vc_solution/vs2026_lame_mp3x.vcxproj.filters,
	  vc_solution/vs2026_lame_test.vcxproj,
	  vc_solution/vs2026_libmp3lame.vcxproj,
	  vc_solution/vs2026_libmp3lame.vcxproj.filters,
	  vc_solution/vs2026_libmp3lame_dll.vcxproj,
	  vc_solution/vs2026_libmpg123_config.props,
	  vc_solution/vs2026_libsndfile_config.props,
	  vc_solution/vs2026_nasm.props, vc_solution/vs2026_nasm.targets,
	  vc_solution/vs2026_nasm.xml,
	  vc_solution/vs2026_win71sdk_config.props,
	  vc_solution/vs_arch_nasm.props, vc_solution/vs_arch_sse2.props,
	  vc_solution/vs_arch_x87.props, vc_solution/vs_gtk_config.props,
	  vc_solution/vs_lame.sln, vc_solution/vs_lame_acm.vcxproj,
	  vc_solution/vs_lame_acm.vcxproj.filters,
	  vc_solution/vs_lame_acm_adbg.vcxproj,
	  vc_solution/vs_lame_acm_adbg.vcxproj.filters,
	  vc_solution/vs_lame_acm_tinyxml.vcxproj,
	  vc_solution/vs_lame_clients.sln,
	  vc_solution/vs_lame_config.props,
	  vc_solution/vs_lame_dshow.vcxproj,
	  vc_solution/vs_lame_dshow.vcxproj.filters,
	  vc_solution/vs_lame_enc_dll.vcxproj,
	  vc_solution/vs_lame_enc_dll.vcxproj.filters,
	  vc_solution/vs_lame_lame.vcxproj,
	  vc_solution/vs_lame_lame.vcxproj.filters,
	  vc_solution/vs_lame_mp3rtp.vcxproj,
	  vc_solution/vs_lame_mp3rtp.vcxproj.filters,
	  vc_solution/vs_lame_mp3x.vcxproj,
	  vc_solution/vs_lame_mp3x.vcxproj.filters,
	  vc_solution/vs_lame_test.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj,
	  vc_solution/vs_libmp3lame.vcxproj.filters,
	  vc_solution/vs_libmp3lame_dll.vcxproj,
	  vc_solution/vs_libmpg123_config.props,
	  vc_solution/vs_libsndfile_config.props,
	  vc_solution/vs_nasm.props, vc_solution/vs_nasm.targets,
	  vc_solution/vs_nasm.xml, vc_solution/vs_win71sdk_config.props:
	  Drop the version year from the Visual Studio build file names
	  
	  The vc_solution solution, project and property files were named
	  for a
	  specific Visual Studio release, although the build supports a
	  range of
	  toolsets. Name them for the build itself so the name no longer
	  implies one
	  version and need not be reissued when the toolset advances;
	  update the
	  cross-references and the packaging list to match.

2026-07-14 18:21  aleidinger

	* INSTALL, Makefile.MSVC, config.h.in, configure, configure.ac,
	  doc/html/history.html, doc/maintainer-build-matrix.md,
	  maintainer/gen-build-matrix.sh,
	  vc_solution/vs2026_lame_config.props:
	  Detect and enable optimization and hardening flags across the
	  builds
	  
	  The Unix build chose optimization flags from hard-coded,
	  per-compiler-version
	  and per-CPU lists that had decayed over time: flags dropped from
	  current GCC,
	  a Clang path that never enabled -ffast-math, and -mtune=native
	  leaking into
	  generic i586/i686 builds. Probe the compiler at configure time
	  and keep only
	  the flags it accepts, so GCC and Clang each get a suitable set
	  without upkeep
	  as they evolve. The same probing drives --enable-expopt, and
	  --enable-native
	  offers machine-specific tuning as an explicit, non-portable
	  opt-in.
	  
	  Enable security hardening by default (fortified libc calls,
	  stack-protector-strong, stack-clash and control-flow protection,
	  full RELRO
	  and a non-executable stack); --disable-hardening opts out. Link
	  the frontends
	  dynamically by default so they can be built as
	  position-independent
	  executables; --disable-dynamic-frontends links a self-contained
	  static
	  frontend. Give the Visual Studio / MSVC build the matching
	  hardening (Control
	  Flow Guard, buffer security checks, CET, DEP, ASLR), drop several
	  obsolete
	  code-generation switches, and compile the release build's
	  consistency
	  assertions out.
	  
	  Per-component encoding speed on two machines (synthetic input,
	  single core,
	  medians of 12 runs; not comparable between machines, and
	  system-dependent):
	  
	  configuration step GCC / i7-10700 Clang / Xeon X5675
	  VBR CBR VBR CBR
	  new opt + hardening, static +0.1% -0.3% -7.2% -5.9% vs old
	  baseline
	  dynamic / PIE default +1.7% +0.6% +3.3% +4.6% vs static above
	  --enable-expopt=norm +2.1% +3.1% -0.2% -3.5% vs default
	  --enable-native -8.5% -8.8% -1.1% -5.0% vs default
	  net: default vs old baseline +1.9% +0.2% -4.1% -1.6%
	  
	  (Negative is faster.) The default is faster on Clang, which picks
	  up
	  -ffast-math, and about a percent slower on GCC, the cost of
	  position-
	  independent frontends. --enable-expopt=norm is not a win for this
	  workload;
	  --enable-native helps most on recent CPUs.
	  
	  The Windows builds have no configure step, so their flags are
	  fixed. A 32-bit
	  MSVC command-line build (i7-10700, medians of 9 runs) is about
	  3.7% faster VBR
	  and 2.5% faster CBR overall: dropping the stale switches and the
	  release
	  assertions saves roughly 4.7% / 3.0%, against which the hardening
	  options cost
	  about a percent. Fast floating point is not enabled on Windows;
	  on a 32-bit
	  build it perturbs the psychoacoustic model past the tolerance its
	  assertions
	  allow.

2026-07-13 19:44  aleidinger

	* Makefile.am, Makefile.in, doc/Doxyfile-internal.in,
	  doc/Makefile.am, doc/Makefile.in, doc/maintainer-build-matrix.md,
	  maintainer, maintainer/gen-build-matrix.ps1,
	  maintainer/gen-build-matrix.sh:
	  Add a build-configuration test matrix for maintainers
	  
	  Validating a change against the configurations it can affect
	  meant
	  configuring and building each one by hand. Provide generators
	  that lay out
	  those builds and drive them together.
	  
	  - A POSIX generator for the autotools build and a Windows
	  generator for
	  the nmake and MSBuild builds create one out-of-tree build per
	  configuration under a master directory, in a star layout: a
	  full-featured base plus one build per interesting build option
	  flipped
	  away from it.
	  - Each master directory gets a driver that builds every
	  configuration
	  without stopping at the first failure, tees each build log to a
	  file and
	  reports the failed builds together with their log locations.
	  - Compilers are detected on PATH for the POSIX build and through
	  the
	  standard installer query for Windows, and optional dependencies
	  are
	  probed so a configuration that cannot be built locally is skipped
	  with a
	  warning.
	  - A how-to under doc/ describes the layout, the per-platform
	  prerequisites
	  and the workflow, and is part of the internal documentation.

2026-07-13 19:42  aleidinger

	* vc_solution/vs2026_lame_config.props,
	  vc_solution/vs2026_libmp3lame.vcxproj,
	  vc_solution/vs2026_libmp3lame.vcxproj.filters:
	  Generate the Visual Studio config.h outside the source tree
	  
	  The Visual Studio build left a generated config.h in the source
	  root after
	  every build, dirtying the checkout. Keep it with the build output
	  instead,
	  so building never modifies the source tree.

2026-07-13 19:40  aleidinger

	* Makefile.MSVC:
	  Support out-of-tree builds in the nmake Makefile
	  
	  The nmake build could only run inside the source tree, where it
	  left its
	  intermediate files scattered among the sources. Let it build from
	  a
	  separate directory, as the autotools and Visual Studio builds
	  already can,
	  so a build can be kept apart from a clean checkout. The
	  traditional in-tree
	  build is unaffected.

2026-07-12 15:48  aleidinger

	* Makefile.MSVC, configMS.h, doc/html/history.html,
	  vc_solution/.gitignore, vc_solution/Makefile.am,
	  vc_solution/Makefile.in, vc_solution/README.vs2019.txt,
	  vc_solution/README.vs2026.txt,
	  vc_solution/vs2019_arch_nasm.props,
	  vc_solution/vs2019_arch_sse2.props,
	  vc_solution/vs2019_arch_x87.props,
	  vc_solution/vs2019_gtk_config.props, vc_solution/vs2019_lame.sln,
	  vc_solution/vs2019_lame_acm.vcxproj,
	  vc_solution/vs2019_lame_acm.vcxproj.filters,
	  vc_solution/vs2019_lame_acm_adbg.vcxproj,
	  vc_solution/vs2019_lame_acm_adbg.vcxproj.filters,
	  vc_solution/vs2019_lame_acm_tinyxml.vcxproj,
	  vc_solution/vs2019_lame_clients.sln,
	  vc_solution/vs2019_lame_config.props,
	  vc_solution/vs2019_lame_dshow.vcxproj,
	  vc_solution/vs2019_lame_dshow.vcxproj.filters,
	  vc_solution/vs2019_lame_enc_dll.vcxproj,
	  vc_solution/vs2019_lame_enc_dll.vcxproj.filters,
	  vc_solution/vs2019_lame_lame.vcxproj,
	  vc_solution/vs2019_lame_lame.vcxproj.filters,
	  vc_solution/vs2019_lame_mp3rtp.vcxproj,
	  vc_solution/vs2019_lame_mp3rtp.vcxproj.filters,
	  vc_solution/vs2019_lame_mp3x.vcxproj,
	  vc_solution/vs2019_lame_mp3x.vcxproj.filters,
	  vc_solution/vs2019_lame_test.vcxproj,
	  vc_solution/vs2019_libmp3lame.vcxproj,
	  vc_solution/vs2019_libmp3lame.vcxproj.filters,
	  vc_solution/vs2019_libmp3lame_dll.vcxproj,
	  vc_solution/vs2019_libmpg123_config.props,
	  vc_solution/vs2019_libsndfile_config.props,
	  vc_solution/vs2019_nasm.props, vc_solution/vs2019_nasm.targets,
	  vc_solution/vs2019_nasm.xml,
	  vc_solution/vs2019_win71sdk_config.props,
	  vc_solution/vs2026_arch_nasm.props,
	  vc_solution/vs2026_arch_sse2.props,
	  vc_solution/vs2026_arch_x87.props,
	  vc_solution/vs2026_gtk_config.props, vc_solution/vs2026_lame.sln,
	  vc_solution/vs2026_lame_acm.vcxproj,
	  vc_solution/vs2026_lame_acm.vcxproj.filters,
	  vc_solution/vs2026_lame_acm_adbg.vcxproj,
	  vc_solution/vs2026_lame_acm_adbg.vcxproj.filters,
	  vc_solution/vs2026_lame_acm_tinyxml.vcxproj,
	  vc_solution/vs2026_lame_clients.sln,
	  vc_solution/vs2026_lame_config.props,
	  vc_solution/vs2026_lame_dshow.vcxproj,
	  vc_solution/vs2026_lame_dshow.vcxproj.filters,
	  vc_solution/vs2026_lame_enc_dll.vcxproj,
	  vc_solution/vs2026_lame_enc_dll.vcxproj.filters,
	  vc_solution/vs2026_lame_lame.vcxproj,
	  vc_solution/vs2026_lame_lame.vcxproj.filters,
	  vc_solution/vs2026_lame_mp3rtp.vcxproj,
	  vc_solution/vs2026_lame_mp3rtp.vcxproj.filters,
	  vc_solution/vs2026_lame_mp3x.vcxproj,
	  vc_solution/vs2026_lame_mp3x.vcxproj.filters,
	  vc_solution/vs2026_lame_test.vcxproj,
	  vc_solution/vs2026_libmp3lame.vcxproj,
	  vc_solution/vs2026_libmp3lame.vcxproj.filters,
	  vc_solution/vs2026_libmp3lame_dll.vcxproj,
	  vc_solution/vs2026_libmpg123_config.props,
	  vc_solution/vs2026_libsndfile_config.props,
	  vc_solution/vs2026_nasm.props, vc_solution/vs2026_nasm.targets,
	  vc_solution/vs2026_nasm.xml,
	  vc_solution/vs2026_win71sdk_config.props:
	  Modernize and synchronize the Windows / Visual Studio build
	  
	  Bring the two Windows build paths (the nmake Makefile.MSVC and
	  the Visual
	  Studio solution) up to date and into agreement with each other
	  and with
	  the autotools build.
	  
	  - Retarget the Visual Studio projects to the current platform
	  toolset and
	  rename the solution and its files to the supported Visual Studio
	  version,
	  refreshing the build instructions to match. Newer toolchains
	  build the
	  projects without a manual retarget.
	  - Move the Windows build onto the external libmpg123 decoder,
	  matching the
	  autotools build; the bundled mpglib decoder has been retired. The
	  nmake
	  build documents how to point at a libmpg123 install, and the
	  Takehiro
	  IEEE-754 rounding shortcut is now off by default on both Windows
	  build
	  paths, as it already is elsewhere.
	  - Use the same floating-point model as the autotools build by
	  default: the
	  nmake build no longer defaults to single precision, so all three
	  builds
	  agree unless single precision is requested explicitly.
	  - Fix a Visual Studio project setting where the libmpg123
	  import-library
	  pre-link step ran even when the decoder was disabled, which broke
	  the
	  build unless the library happened to be present.
	  - Ignore the locally provided external libraries so they are
	  never
	  committed, and give the Visual Studio project files consistent
	  line
	  endings.

2026-07-12 13:27  aleidinger

	* doc/html/history.html, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c:
	  Honor the --enable-ieeehack configure switch
	  
	  The Takehiro IEEE-754 rounding shortcut is opt-in and off by
	  default, but
	  the switch never took effect: the guards tested whether the
	  configure
	  macro was defined rather than its value, and the build always
	  defines it,
	  so the shortcut was always compiled in. Test the value instead,
	  so the
	  default build uses the portable rounding path as intended.
	  
	  The shortcut only speeds up float-to-int conversion on much older
	  CPUs
	  that lack a fast native conversion instruction; on current
	  hardware
	  disabling it is faster, and the default output changes slightly
	  (a
	  different rounding table). Measured on a Xeon X5675, encoding a
	  23.8 min
	  corpus (hyperfine, 1 warmup + 10 hot runs, CPU-pinned):
	  
	  CBR -b320: 40.23 +/- 0.50 s -> 37.25 +/- 0.59 s (1.08x faster)
	  VBR -V2: 30.48 +/- 0.44 s -> 27.76 +/- 0.31 s (1.10x faster)
	  
	  SF bug #521

2026-07-12 13:27  aleidinger

	* doc/html/history.html, include/libmp3lame.sym:
	  Export the decoder analysis hooks from the shared library
	  
	  The MP3 decoder's per-frame analysis hooks hip_set_pinfo and
	  hip_finish_pinfo belong to the documented decoder API and are
	  used by the
	  bundled frontends, but they were missing from the shared
	  library's export
	  list. Linking a frontend against the shared library - as
	  --enable-dynamic-frontends does - therefore failed with
	  undefined-symbol
	  errors. Add both so dynamic-frontend builds link.
	  
	  hip_set_pinfo is the symbol named in the report; hip_finish_pinfo
	  is a
	  second omission of the same kind, reached by the core file reader
	  once the
	  first is resolved.
	  
	  SF bug #515

2026-07-12 08:06  aleidinger

	* frontend/get_audio.c, frontend/main.c, frontend/mp3rtp.c,
	  frontend/rtp.c, libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/lame.c, libmp3lame/psymodel.c:
	  libmp3lame, frontend: guard unchecked allocations against OOM
	  
	  Several allocation sites in the library init path and the
	  frontend
	  dereferenced the result without checking it, turning an
	  allocation
	  failure into a crash rather than a clean error. The trigger is
	  out-of-memory, not attacker input - but the crash is avoidable.
	  
	  Check these allocations and fail gracefully: the library init
	  path
	  reports the failure to its caller; the frontend paths print an
	  error and
	  stop. Sites that already handle failure are left alone, as are
	  two whose
	  fix is more invasive (the resampler init and an input-buffer
	  realloc),
	  which are better handled on their own.

2026-07-12 08:05  aleidinger

	* frontend/parse.c, include/lame.def, include/lame.h,
	  include/libmp3lame.sym, libmp3lame/id3tag.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_id3tag.c, test/unit/test_parse_id3.c:
	  frontend, libmp3lame: fix ID3v2 UTF-8 tag path (SF #524)
	  
	  The ID3v2 tag writer routed both the UTF-8 and UTF-16
	  command-line
	  encodings through a single path typed for UTF-16. For
	  --id3v2-utf8 that
	  mismatches the actual byte-oriented data and reaches tag setters
	  current
	  toolchains reject, so LAME fails to build with GCC 16 - and
	  misencodes
	  UTF-8 tags where it does build.
	  
	  Give each encoding its own path so the data and the setters it
	  feeds
	  always agree. The UTF-8 field-value case needs a UTF-8 setter the
	  library
	  did not expose, so add one beside the existing field-value
	  setters.
	  
	  Covered by new unit tests for the id3tag API and the frontend tag
	  dispatch across all three text encodings.

2026-07-12 08:04  aleidinger

	* doc/Doxyfile-common.in, doc/Doxyfile-internal.in,
	  doc/Doxyfile.in, doc/Makefile.am, doc/Makefile.in,
	  doc/doxygen/internal.md, doc/doxygen/unit_tests.dox,
	  test/unit/test_get_audio_aiff.c, test/unit/test_sanity.c,
	  test/unit/test_stubs.c:
	  doc: split Doxygen into public-API and internal configurations
	  
	  The Doxygen setup produced a single documentation set that mixed
	  the
	  public library API with internal implementation detail. Separate
	  the two
	  so each audience gets what it needs:
	  
	  - the public configuration documents the supported API (the
	  interface
	  declared in lame.h) and nothing else;
	  - a new internal configuration additionally covers the static
	  internals
	  of the library and frontend and the unit-test suite, with its own
	  front
	  page that points library users back to the public docs.
	  
	  Both draw from a shared base configuration and override only
	  where they
	  differ, so the two cannot drift apart. The unit-test suite gains
	  a
	  documented home: a "Unit tests" group page plus per-file and
	  per-function
	  comments on the tests and their link stubs.
	  
	  The templates, front pages, theme and logo are added to the
	  distribution
	  so the documentation can be regenerated from a release tarball,
	  not only
	  from a checkout.

2026-07-12 08:00  aleidinger

	* doc/Doxyfile-common.in, doc/Doxyfile.in, frontend/parse.c,
	  test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_merge_argv.c:
	  frontend: size the merged argument vector to fit instead of a
	  fixed 512
	  
	  parse_args() merged the LAMEOPT tokens and the real argv[] into a
	  fixed
	  str_argv[512]. merge_argv() correctly skipped writes past slot
	  512, but it
	  returned the raw, unclamped argc + str_argc - 1 - so with more
	  than ~511
	  total tokens the caller looped past the initialised entries and
	  dereferenced
	  uninitialised str_argv[] slots as argument pointers (a stack
	  over-read from
	  an ordinary but long command line), and every argument past the
	  512th was
	  silently dropped.
	  
	  Allocate str_argv to actually hold every token (argc + the
	  LAMEOPT length as
	  an upper bound on its token count + 1) and free it when done;
	  argc and the
	  environment are both bounded by the OS ARG_MAX, so the size stays
	  in range.
	  This removes both the over-read and the silent truncation.
	  merge_argv() also
	  clamps its returned count to N as a defensive invariant, so the
	  helper can
	  never report more entries than the array it was given holds.

2026-07-12 06:45  aleidinger

	* frontend/parse.c, test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/parse_test_stubs.c, test/unit/test_parse_pathlen.c:
	  frontend: bound positional input/output filename length
	  
	  The positional filename arguments (input, output, and each nogap
	  input)
	  were copied into their char[PATH_MAX+1] buffers with
	  strncpy(dst, argv[i], PATH_MAX+1). strncpy only null-terminates
	  when the
	  source is shorter than the count, so a filename of PATH_MAX+1
	  bytes or
	  more filled the buffer completely with no terminator; scanPath()
	  and
	  scanBasename() then scan it as a plain C string ("for (; *s;
	  ++s)") and
	  walk off the end of the stack array.
	  
	  The long-option paths (--nogapout, --out-dir) already guard
	  against this
	  by checking strnlen(nextArg, PATH_MAX) and force-terminating.
	  Factor that
	  bounded copy into a small helper, set_path_arg(), and route all
	  three
	  positional-filename copies through it: an over-long name is
	  rejected with
	  an error instead of leaving a non-terminated buffer.

2026-07-11 16:11  aleidinger

	* test/unit/Makefile.am, test/unit/Makefile.in,
	  test/unit/test_get_audio_aiff.c, test/unit/test_stubs.c:
	  test: add an AIFF FORM-size underflow regression test
	  
	  Guards the fix that rejects AIFF/AIFC files whose FORM chunk size
	  is less
	  than 4. Such a size underflowed the chunk-size accounting and
	  drove the
	  header parser through ~1e9 near-empty iterations before
	  returning, a hang
	  triggerable by a ~16-byte file.
	  
	  parse_aiff_header() is static, so the test compiles
	  frontend/get_audio.c
	  in directly. Because a malformed and a valid-but-empty file both
	  end at the
	  same -1 return, the test wraps fread(): once armed it caps the
	  read budget
	  and longjmps out, so a build missing the guard fails fast instead
	  of
	  hanging the suite. A well-formed minimal AIFF is checked too, to
	  show the
	  guard rejects only the malformed case.

2026-07-11 13:27  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.am,
	  Makefile.in, configure, configure.ac, doc/Makefile.in,
	  doc/html/Makefile.in, doc/html/history.html, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in, test-driver, test/Makefile.am,
	  test/Makefile.in, test/unit, test/unit/Makefile.am,
	  test/unit/Makefile.in, test/unit/test_sanity.c,
	  vc_solution/Makefile.in:
	  build: add an opt-in CMocka unit-test harness
	  (--enable-unit-tests)
	  
	  Introduce a dedicated test/unit/ tree, built with CMocka and run
	  via
	  "make check", so that fixes from here on can ship with a
	  regression test
	  instead of being verified once by hand and left unguarded.
	  
	  The harness is disabled by default: a plain "configure && make"
	  is wholly
	  unaffected, and the test binaries build only under "make check".
	  When
	  --enable-unit-tests is requested explicitly, configure requires
	  CMocka
	  through pkg-config and fails with a clear message if it is
	  absent, rather
	  than silently configuring a test target that builds nothing.
	  
	  The existing hand-maintained material under test/ (lametest.py,
	  the .op
	  reference files, the ad-hoc smoke program) is now dist-tracked
	  for the
	  first time. This commit adds only the scaffolding and a minimal
	  smoke
	  test; the first real per-function test follows separately.

2026-07-11 07:55  aleidinger

	* configure, configure.ac, libmp3lame/version.h:
	  Welcome to 4.1 alpha.

2026-07-11 07:44  aleidinger

	* doc/html/history.html:
	  Add release date.

2026-07-11 07:42  aleidinger

	* ChangeLog:
	  Update ChangeLog for release.

2026-07-11 07:40  aleidinger

	* testcase.mp3:
	  Prepare for release

2026-07-11 07:39  aleidinger

	* libmp3lame/version.h:
	  Prepare for release

2026-07-11 07:36  aleidinger

	* doc/html/history.html, frontend/get_audio.c:
	  get_audio: reject AIFF files with an undersized FORM chunk size
	  
	  A crafted AIFF/AIFC whose FORM chunk size is below 4 underflowed
	  the
	  unsigned chunk-size counter when the 4-byte form type is
	  subtracted from
	  it, wrapping it to roughly 4 billion and sending the
	  chunk-scanning loop
	  into an effectively unbounded spin - a denial of service
	  triggerable with
	  a tiny (~16-byte) malicious file. Validate the size before the
	  subtraction. The remaining in-loop subtractions are already
	  bounded by
	  the surrounding "chunk size >= 8" loop guard, so the single entry
	  check
	  closes the hole.
	  
	  CVSS 5.5 (AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H).

2026-07-11 07:36  aleidinger

	* libmp3lame/util.c:
	  util.c: document why fill_buffer_resample's offset assert is a
	  sufficient guard
	  
	  The blackfilt[joff] table index in fill_buffer_resample() looks
	  like it could
	  run one slot past the precomputed window table when offset
	  exceeds +0.5,
	  guarded only by an assert that compiles out under NDEBUG. It
	  cannot: offset is
	  bounded to [-0.5, +0.5) by construction for integer sample rates,
	  which are
	  the only kind LAME ever sees, so joff always stays within
	  blackfilt[0..2*bpc].
	  Record that reasoning next to the assert so the assert-only guard
	  is not later
	  mistaken for a missing runtime bounds check.

2026-07-11 07:35  aleidinger

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h,
	  doc/html/history.html:
	  Blade DLL: bounds-check beInitStream() config copy, fix packed
	  layout
	  
	  beInitStream() copied the caller-supplied BE_CONFIG into a stack
	  local
	  using the caller-controlled dwStructSize as the length, with no
	  check
	  that it fits. A host that sets dwStructSize larger than the
	  struct --
	  whether malicious or merely built with a different compiler --
	  overflows
	  the stack. Clamp the copy to sizeof(lameConfig).
	  
	  The size mismatch was also reachable benignly: ATTRIBUTE_PACKED
	  was
	  attached to the pointer typedef, not the struct, so under
	  GCC/MinGW the
	  packed attribute was ignored and sizeof(BE_CONFIG) was 332 rather
	  than
	  the 331 the header intends and MSVC produces. Apply the existing
	  pack(1)/pop pragma unconditionally so both compilers agree on
	  331.
	  
	  CVSS 8.4 (AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H).

2026-07-11 07:34  aleidinger

	* doc/html/history.html, include/libmp3lame.sym,
	  libmp3lame/id3tag.c:
	  id3tag: export the UTF-8 tag functions and guard against OOM
	  
	  Export id3tag_set_textinfo_utf8 and id3tag_set_comment_utf8 from
	  the shared
	  library, and fix a possible crash on out-of-memory in the ID3v2
	  user-defined
	  tag setters. SF bug #518.

2026-07-11 07:30  aleidinger

	* doc/html/history.html, include/lame.h, libmp3lame/id3tag.c:
	  id3tag: retype the UTF-8 text setters to char* to fix modern
	  GCC/Clang builds
	  
	  Their text parameter was cloned from the UTF-16 siblings
	  (unsigned short*),
	  which recent GCC/Clang reject as an incompatible pointer type;
	  char* matches
	  the UTF-8 bytes the function bodies already handle, and also
	  fixes the
	  --id3v2-utf8 --tg genre corruption.
	  
	  Patch by Rudi Heitbaum, SF patch #102. Reported by lazka, SF bug
	  #523.

2026-07-11 07:26  aleidinger

	* doc/html/history.html:
	  history.html: switch entry colours to descriptive CSS classes,
	  add security category
	  
	  Use descriptive CSS classes instead of direct per-entry colour
	  references, and
	  add a "security" category (class plus legend entry) ahead of the
	  first security
	  fix that will use it.

2026-07-10 18:28  aleidinger

	* ABOUT-NLS, ACM/ADbg/Makefile.in, ACM/Makefile.in,
	  ACM/ddk/Makefile.in, ACM/tinyxml/Makefile.in, Dll/Makefile.in,
	  Makefile.in, aclocal.m4, compile, config.h.in, config.rpath,
	  configure, configure.ac, depcomp, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in, install-sh,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, ltmain.sh, m4/build-to-host.m4,
	  m4/gettext.m4, m4/host-cpu-c-abi.m4, m4/iconv.m4,
	  m4/intlmacosx.m4, m4/lib-ld.m4, m4/lib-link.m4, m4/lib-prefix.m4,
	  m4/libtool.m4, m4/ltoptions.m4, m4/ltsugar.m4, m4/ltversion.m4,
	  m4/lt~obsolete.m4, m4/nls.m4, m4/po.m4, m4/progtest.m4,
	  mac/Makefile.in, misc/Makefile.in, missing, mpglib/Makefile.in,
	  po, po/Makefile.in.in, po/Makevars.template, po/Rules-quot,
	  po/boldquot.sed, po/en@boldquot.header, po/en@quot.header,
	  po/insert-header.sed, po/quot.sed, po/remove-potcdate.sed,
	  vc_solution/Makefile.in:
	  update/regen with more recent autotools

2026-07-10 18:18  aleidinger

	* .gitignore, configure.ac, doc/Doxyfile.in, doc/Makefile.am,
	  doc/doxygen, doc/doxygen/doxygen-awesome-css,
	  doc/doxygen/doxygen-awesome-css/doxygen-awesome.css,
	  doc/doxygen/lame-theme.css, doc/doxygen/logo,
	  doc/doxygen/logo/lame_logo.svg,
	  doc/doxygen/logo/lame_logo_full.svg, doc/doxygen/mainpage.md:
	  Add Doxyfile: API documentation build configuration
	  
	  Sets up Doxygen-based API documentation for include/,
	  libmp3lame/, and
	  frontend/, generated on demand via a new "doxygen" target in
	  doc/Makefile.am - not part of all/check, since there's no
	  per-function
	  documentation yet for it to build.
	  
	  - doc/Doxyfile.in is a template rather than a static Doxyfile.
	  - Graphviz's dot is autodetected via configure, or specified
	  explicitly
	  with --with-dot=PATH; HAVE_DOT/DOT_PATH fall back cleanly to no
	  diagrams when it isn't found.
	  - doc/doxygen/mainpage.md is a purpose-written entry page (API
	  entry
	  points, license pointer) for the generated docs.
	  - Vendored doxygen-awesome-css (MIT-licensed,
	  https://github.com/jothepro/doxygen-awesome-css) with a
	  color/font
	  override (doc/doxygen/lame-theme.css) matching the lame.sf.net
	  site's own palette.
	  - Doxyfile comments mark the public (include/lame.h) vs. internal
	  (everything else in libmp3lame/) surface split.

2026-07-10 18:17  aleidinger

	* configure.ac, doc/html/history.html, libmp3lame/VbrTag.c,
	  libmp3lame/machine.h, libmp3lame/version.c, libmp3lame/version.h:
	  Bump major version to 4.0 to fix LAME-tag encoder-version field
	  overflow
	  
	  The LAME tag embedded in every encoded MP3 has a fixed 9-byte
	  field for
	  the encoder-version string ("LAME" + major + "." + minor +
	  type-marker).
	  Since 3.100, the 3-digit minor version left no room for the
	  trailing
	  alpha/beta/release marker character: strncpy(...,9) silently
	  truncated
	  it on every 3.100-3.102 build. The field's width and format can't
	  change
	  without breaking third-party decoders that parse it at a fixed
	  offset
	  and pattern-match its exact format, so the version numbers had to
	  give
	  instead.
	  
	  - libmp3lame/version.h, configure.ac: version bump
	  (LIB_MAJOR_VERSION/
	  LIB_MINOR_VERSION, the shared-library ABI version, are untouched
	  - the
	  ABI hasn't changed)
	  - libmp3lame/version.c: add a compile-time assertion tying the
	  encoder-
	  version string's actual length to the 9-byte budget, so a future
	  version bump that reintroduces the overflow fails the build
	  instead
	  of silently truncating again; Doxygen documentation of the
	  field's
	  fixed width/format and the exact byte-budget formula
	  - libmp3lame/VbrTag.c: matching documentation at the field's
	  write site
	  - libmp3lame/machine.h: fix compiletime_assert(), which the check
	  above
	  needed - ##-pasting doesn't expand __FILE__/__LINE__, so two uses
	  in
	  one translation unit collided on the same enum tag (dead code
	  before
	  this, zero call sites)
	  - doc/html/history.html: user-facing changelog entry
	  
	  Verified: rebuilt with UCRT64 GCC; the assertion compiles clean
	  at 4.0
	  and correctly fails against the original 3.102 numbers;
	  spot-checked
	  across major/minor digit-count combinations (10.0, 99.9, 9.99
	  pass;
	  10.89, 10.90, 100.0, 9.999 correctly fail) to confirm it checks
	  the
	  true combined string length rather than a fixed digit-count rule.

2026-07-09 13:13  aleidinger

	* configure.ac, libmp3lame/version.h:
	  Welcome to 3.102

2026-07-09 12:14  aleidinger

	* ChangeLog:
	  Update ChangeLog for 3.101 release.

2026-07-09 12:08  aleidinger

	* doc/html/history.html, libmp3lame/version.h, testcase.mp3:
	  Prepare 3.101 release.

2023-12-18 13:41  aleidinger

	* frontend/get_audio.c:
	  Use suitable typedef for Windows.

2023-12-18 13:40  aleidinger

	* frontend/main.h, frontend/parse.c:
	  Use better return types and provide prototype for toUtf8().

2023-12-15 12:07  aleidinger

	* frontend/parse.c, include/lame.def, include/lame.h,
	  libmp3lame/id3tag.c:
	  Fix UTF8 id3tag support.

2023-12-14 09:13  aleidinger

	* frontend/parse.c:
	  Unbreak UTF8 genre tag. Text tags still broken.
	  
	  Disable UTF8 for now in the frontend.

2023-12-14 09:12  aleidinger

	* include/lame.def, include/lame.h, libmp3lame/id3tag.c:
	  Unbreak UTF8 genre tag, text-tags still broken (wrong encoding).

2023-12-12 21:07  aleidinger

	* frontend/parse.c:
	  Try to provide a toUtf8() function for Windows.
	  Completely untested, based upon the assumption that
	  Windows already uses UTF8. This assumption is based
	  upon the function names of the Windows implementation
	  of toUtf18() and toLatin1().

2023-11-21 18:00  aleidinger

	* libmp3lame/version.h:
	  increase beta patch-level after todays changes

2023-11-21 15:12  aleidinger

	* configure:
	  regen

2023-11-21 15:12  aleidinger

	* config.h.in:
	  regen

2023-11-21 15:12  aleidinger

	* configure.ac, doc/html/history.html, frontend/mp3rtp.c,
	  frontend/rtp.c:
	  Add IPv6 support for mp3rtp.
	  This includes an incompatible change to the command line
	  arguments for existing use cases / scripts.
	  
	  Submitted by: Surabhi
	  Patch: #69

2023-11-21 15:06  aleidinger

	* doc/html/history.html:
	  pkg-config support and some tab to spaces

2023-11-21 14:34  aleidinger

	* configure:
	  regen

2023-11-21 14:33  aleidinger

	* configure.ac:
	  fix variable name

2023-11-21 14:32  aleidinger

	* config.h.in:
	  regen / modernize with newer autotools

2023-11-21 14:32  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in, vc_solution/Makefile.in:
	  regen / modernize with newer autotools

2023-11-21 14:31  aleidinger

	* configure:
	  regen / modernize with newer autotools

2023-11-21 14:31  aleidinger

	* aclocal.m4:
	  regen / modernize with newer autotools

2023-11-21 14:30  aleidinger

	* configure.ac, configure.in:
	  rename to current convention and modernize with newer autotools

2023-11-21 14:29  aleidinger

	* Makefile.am:
	  modernize with newer autotools

2023-11-21 14:29  aleidinger

	* m4, m4/libtool.m4, m4/ltoptions.m4, m4/ltsugar.m4,
	  m4/ltversion.m4, m4/lt~obsolete.m4:
	  modernize with newer autotools

2023-11-21 14:29  aleidinger

	* compile, depcomp, missing:
	  modernize with newer autotools

2023-11-21 14:28  aleidinger

	* install-sh, ltmain.sh:
	  modernize with newer autotools

2023-11-21 14:27  aleidinger

	* libmp3lame/Makefile.am, libmp3lame/lame.pc.in:
	  Add pkg-config support.
	  
	  Taken from: Nicolas Boulenguez <nicolas@debian.org>
	  https://salsa.debian.org/multimedia-team/lame/-/blob/master/debian/patches/pkg-config.patch
	  Support Request: #42

2023-11-21 13:34  aleidinger

	* doc/html/history.html, doc/man/lame.1, frontend/parse.c,
	  include/lame.def, include/lame.h, include/libmp3lame.sym,
	  libmp3lame/id3tag.c, libmp3lame/id3tag.h:
	  Add ID3v2.4 support (UTF-8) to libmp3lame.
	  
	  Submitted by: "kris"
	  Patch: #99
	  
	  The use in the frontend is by me and only slightly tested (simple
	  --tt with umlauts).

2021-06-19 12:14  aleidinger

	* frontend/get_audio.c:
	  Fix build on cygwin with configure options --disable-static
	  --enable-shared --disable-decoder.
	  
	  Submitted by: Carlo Bramini
	  Patch: #97

2021-06-19 12:05  aleidinger

	* libmp3lame/util.c:
	  Fix build on FreeBSD riscv (replace compile time arch detection
	  by feature detection).
	  
	  Patch by: arichardson of LLVM project.
	  Submitted by: Robert Clausecker
	  Patch: #96

2021-03-12 12:58  aleidinger

	* libmp3lame/version.h:
	  bump to beta 2

2021-03-12 12:48  aleidinger

	* libmp3lame/vector/xmm_quantize_sub.c:
	  Fix some issues with clang build.
	  
	  Obtained from: VLC
	  
	  Original commit log from mstorsjo:
	  ---snip---
	  …GCC target("sse")
	  
	  Clang doesn't support the pragma, but does support the attribute.
	  Update the existing patch to do things this way.
	  
	  With previous versions of lame, this hasn't been an issue,
	  because
	  if one doesn't pass any -march flag, clang defaults to pentium4
	  as
	  target for i386 targets, and there both sse and sse2 are
	  available
	  by default.
	  
	  Since lame 3.100, lame now passes -march=i686 even for clang,
	  which
	  disables SSE by default, requiring the per-function enabling to
	  actually work.
	  ---snip---

2021-03-12 12:39  aleidinger

	* INSTALL:
	  Remove the part about NDEBUG, we use existing asserts as a quick
	  and easy way to prevent bad input to cause issues.

2021-03-10 12:07  aleidinger

	* doc/html/history.html:
	  Fix version of upcomming release.

2021-03-10 11:45  aleidinger

	* COPYING:
	  Correct the address of the FSF. [bugs#420]

2021-03-10 11:37  aleidinger

	* frontend/lametime.c:
	  Cygwin documents O_BINARY instead of _O_BINARY, see
	  https://cygwin.com/cygwin-ug-net/using-textbinary.html
	  
	  This should fix [bugs#451] and [bugs#494].

2021-03-10 11:25  aleidinger

	* configure:
	  regen

2021-03-10 11:25  aleidinger

	* configure.in:
	  Simplify (remove superflous).
	  
	  Suggested by: Thomas Orgis

2021-03-10 10:13  aleidinger

	* Makefile.in:
	  regen

2021-03-10 10:12  aleidinger

	* configure:
	  regen

2021-03-10 10:05  aleidinger

	* configure.in:
	  Fix decoder/mpg123 logic.

2021-03-10 10:05  aleidinger

	* frontend/get_audio.c:
	  Fix implicit declaration.

2021-03-10 10:03  aleidinger

	* Makefile.am:
	  Remove superflous dependency.
	  
	  Submitted by: vemakereporter via [patches#84]

2021-03-09 13:12  aleidinger

	* ChangeLog:
	  Add recent commits.

2021-03-09 12:01  aleidinger

	* libmp3lame/version.h:
	  Update version to 3.101 beta 1.

2021-03-09 11:51  aleidinger

	* configure:
	  regen

2021-03-09 11:51  aleidinger

	* config.h.in:
	  regen

2021-03-09 11:50  aleidinger

	* configure.in:
	  Remove reference to internal decoder.

2021-03-09 11:29  aleidinger

	* doc/html/history.html:
	  Add recent changes.

2021-03-09 11:29  aleidinger

	* README:
	  Document dependency on mpg123.

2021-03-09 11:27  aleidinger

	* mpglib/common.c, mpglib/common.h, mpglib/dct64_i386.c,
	  mpglib/dct64_i386.h, mpglib/decode_i386.c, mpglib/decode_i386.h,
	  mpglib/huffman.h, mpglib/interface.c, mpglib/interface.h,
	  mpglib/l2tables.h, mpglib/layer1.c, mpglib/layer1.h,
	  mpglib/layer2.c, mpglib/layer2.h, mpglib/layer3.c,
	  mpglib/layer3.h, mpglib/mpg123.h, mpglib/tabinit.c,
	  mpglib/tabinit.h:
	  RIP mpglib

2021-03-09 11:26  aleidinger

	* Makefile.in, mpglib/Makefile.in:
	  regen

2021-03-09 11:26  aleidinger

	* mpglib/mpglib.h:
	  Replace mpeg decoder with external libmpg123 - step 2
	  
	  Remove the internal mpglib.
	  
	  Submitted by: Michael Fink via [patches:#95]

2021-03-09 11:26  aleidinger

	* libmp3lame/mpglib_interface.c, mpglib/Makefile.am, mpglib/README:
	  Replace mpeg decoder with external libmpg123 - step 2
	  
	  Remove the internal mpglib.
	  
	  Submitted by: Michael Fink via [patches:#95]

2021-03-09 11:25  aleidinger

	* configure:
	  regen

2021-03-09 11:24  aleidinger

	* configure.in:
	  remove macosx xcode directory

2021-03-09 11:04  aleidinger

	* Makefile.am, macosx:
	  Remove xcode project, nobody stepped up to adapt it for the
	  libmpg123 changes.

2021-03-09 10:52  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in,
	  macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, vc_solution/Makefile.in:
	  regen

2021-03-09 10:52  aleidinger

	* configure:
	  regen

2021-03-09 10:52  aleidinger

	* aclocal.m4:
	  regen

2021-03-09 10:35  aleidinger

	* configure.in:
	  - Remove cases which matches all previous cases, this was not
	  intended.
	  - In the clang case, do the same as for gcc, if someone provied
	  CFLAGS on his own,
	  do not add flags on our own.

2020-08-25 10:24  aleidinger

	* vc_solution/README.vs2019.txt, vc_solution/vs2019_lame.sln,
	  vc_solution/vs2019_lame_config.props,
	  vc_solution/vs2019_lame_lame.vcxproj,
	  vc_solution/vs2019_libmp3lame.vcxproj,
	  vc_solution/vs2019_libmp3lame_dll.vcxproj,
	  vc_solution/vs2019_libmpg123_config.props,
	  vc_solution/vs2019_libsndfile_config.props:
	  Add 64-bit support.
	  
	  Submitted by: Michael Fink via [patches:#94]

2020-08-25 10:18  aleidinger

	* include/lame.def, include/libmp3lame.sym:
	  Add hip_decode_init_gapless() to the linker exports.
	  
	  Submitted by: Michael Fink via [patches:#92]

2020-08-25 10:14  aleidinger

	* frontend/get_audio.c:
	  Fix the "Could not close audio input file" message when decoding
	  from mp3
	  or transcoding. The message only occurred on Win32 builds of
	  lame.exe.
	  
	  Submitted by: Michael Fink via [patches:#93]

2020-08-19 14:06  aleidinger

	* ACM/acm.rc, Dll/BladeMP3EncDLL.c, LICENSE, Makefile.MSVC,
	  debian/copyright, doc/man/lame.1, dshow/Property.rc,
	  libmp3lame/id3tag.c, misc/abx.c:
	  Fix the homepage of LAME. The old www.mp3dev.org is pointing to
	  the
	  sourceforge page.

2020-08-19 13:13  aleidinger

	* libmp3lame/i386/nasm.h:
	  Fix nasm include file (UTF-8 BOM which nasm does't know about).
	  
	  Submitted by: KO Myung-Hun via [patches:#81]

2020-08-19 12:56  aleidinger

	* vc_solution/Makefile.am, vc_solution/Makefile.in,
	  vc_solution/README.vs2019.txt, vc_solution/arch_nasm.vsprops,
	  vc_solution/arch_sse2.vsprops, vc_solution/arch_x87.vsprops,
	  vc_solution/vc9_lame.sln, vc_solution/vc9_lame_acm.vcproj,
	  vc_solution/vc9_lame_acm_adbg.vcproj,
	  vc_solution/vc9_lame_acm_tinyxml.vcproj,
	  vc_solution/vc9_lame_clients.sln,
	  vc_solution/vc9_lame_config.vsprops,
	  vc_solution/vc9_lame_dll.vcproj,
	  vc_solution/vc9_lame_dll_example.vcproj,
	  vc_solution/vc9_lame_dshow.vcproj,
	  vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3rtp.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj,
	  vc_solution/vc9_lame_test.vcproj,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_mpglib.vcproj, vc_solution/vc9_nasm.rules,
	  vc_solution/vs2019_arch_nasm.props,
	  vc_solution/vs2019_arch_sse2.props,
	  vc_solution/vs2019_arch_x87.props,
	  vc_solution/vs2019_gtk_config.props, vc_solution/vs2019_lame.sln,
	  vc_solution/vs2019_lame_acm.vcxproj,
	  vc_solution/vs2019_lame_acm.vcxproj.filters,
	  vc_solution/vs2019_lame_acm_adbg.vcxproj,
	  vc_solution/vs2019_lame_acm_adbg.vcxproj.filters,
	  vc_solution/vs2019_lame_acm_tinyxml.vcxproj,
	  vc_solution/vs2019_lame_clients.sln,
	  vc_solution/vs2019_lame_config.props,
	  vc_solution/vs2019_lame_dshow.vcxproj,
	  vc_solution/vs2019_lame_dshow.vcxproj.filters,
	  vc_solution/vs2019_lame_enc_dll.vcxproj,
	  vc_solution/vs2019_lame_enc_dll.vcxproj.filters,
	  vc_solution/vs2019_lame_lame.vcxproj,
	  vc_solution/vs2019_lame_lame.vcxproj.filters,
	  vc_solution/vs2019_lame_mp3rtp.vcxproj,
	  vc_solution/vs2019_lame_mp3rtp.vcxproj.filters,
	  vc_solution/vs2019_lame_mp3x.vcxproj,
	  vc_solution/vs2019_lame_mp3x.vcxproj.filters,
	  vc_solution/vs2019_lame_test.vcxproj,
	  vc_solution/vs2019_libmp3lame.vcxproj,
	  vc_solution/vs2019_libmp3lame.vcxproj.filters,
	  vc_solution/vs2019_libmp3lame_dll.vcxproj,
	  vc_solution/vs2019_libmpg123_config.props,
	  vc_solution/vs2019_libsndfile_config.props,
	  vc_solution/vs2019_nasm.props, vc_solution/vs2019_nasm.targets,
	  vc_solution/vs2019_nasm.xml,
	  vc_solution/vs2019_win71sdk_config.props:
	  - add files to build LAME with Visual Studio 2019
	  - remove the old project files for VC 9
	  - add a readme with a description about what you can do with the
	  files
	  
	  Submitted by: Michael Fink via [patches:#91]

2020-08-19 12:45  aleidinger

	* frontend/get_audio.c:
	  On Win32 compiles it can happen that lame.exe and libmp3lame.dll
	  use
	  different C++ runtimes, which maintail different FILE* lists, and
	  fileno() would produce invalid file numbers for those msvcrt
	  instances.
	  So use mpg123_replace_reader_handle() / mpg123_open_handle() here
	  instead of mpg123_open_fd().
	  
	  Submitted by: Michael Fink via [patches:#90]

2020-08-19 12:36  aleidinger

	* libmp3lame/version.h:
	  Change the URL of the LAME webpage to the canonical https
	  location.
	  
	  Submitted by: Michael Fink via [patches:#89]

2020-08-19 12:29  aleidinger

	* ACM/ACM.cpp:
	  An obsolete method TranslateURL() is called which is not strictly
	  necessary.
	  The patch is in the ACM module, which is the ancient Windows
	  "Audio Codec
	  Manager". Applying this patch is necessary before applying the
	  Visual Studio
	  2019 project files patch.
	  
	  Submitted by: Michael Fink via [patches:#88]

2020-07-11 10:24  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in,
	  macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, vc_solution/Makefile.in:
	  Replace mpeg decoder with external libmpg123 - step 1/2 part 3/3
	  
	  Step 1 is to allow to use external libmp123 (preferred) or
	  interal mpglib to decode mpeg audio.
	  
	  Patch by Thomas Orgis <thomas-forum@orgis.org>, with some
	  minor modifications to the autotools build by me.
	  
	  Part 3: Regenerate automake stuff.
	  
	  
	  
	  Step 2 would be to remove the internal mpglib.

2020-07-11 10:23  aleidinger

	* config.h.in, configure:
	  Replace mpeg decoder with external libmpg123 - step 1/2 part 2/3
	  
	  Step 1 is to allow to use external libmp123 (preferred) or
	  interal mpglib to decode mpeg audio.
	  
	  Patch by Thomas Orgis <thomas-forum@orgis.org>, with some
	  minor modifications to the autotools build by me.
	  
	  Part 2: regenerate autoconf stuff
	  
	  
	  
	  Step 2 would be to remove the internal mpglib.

2020-07-11 10:23  aleidinger

	* configure.in:
	  Replace mpeg decoder with external libmpg123 - step 1/2 part 1/3
	  
	  Step 1 is to allow to use external libmp123 (preferred) or
	  interal mpglib to decode mpeg audio.
	  
	  Patch by Thomas Orgis <thomas-forum@orgis.org>, with some
	  minor modifications to the autotools build by me.
	  
	  Part 1: commit the source files and the input files for the
	  build.
	  
	  
	  
	  Step 2 would be to remove the internal mpglib.

2020-07-11 10:22  aleidinger

	* frontend/Makefile.am, frontend/get_audio.c, frontend/parse.c,
	  include/lame.h, libmp3lame/Makefile.am,
	  libmp3lame/mpglib_interface.c, libmp3lame/util.h,
	  mpglib/Makefile.am, mpglib/common.h, mpglib/mpglib.h,
	  mpglib/tabinit.c, mpglib/tabinit.h:
	  Replace mpeg decoder with external libmpg123 - step 1/2 part 1/3
	  
	  Step 1 is to allow to use external libmp123 (preferred) or
	  interal mpglib to decode mpeg audio.
	  
	  Patch by Thomas Orgis <thomas-forum@orgis.org>, with some
	  minor modifications to the autotools build by me.
	  
	  Part 1: commit the source files and the input files for the
	  build.
	  
	  
	  
	  Step 2 would be to remove the internal mpglib.

2020-05-08 16:02  aleidinger

	* doc/html/history.html:
	  Add recent changes.

2020-05-08 15:33  aleidinger

	* configure:
	  regen

2020-05-08 15:32  aleidinger

	* configure.in:
	  Disable Takehiros IEEE hack by default, add an option for
	  not so modern CPU architectures to enable it.
	  
	  While I'm here, make some whitespace fixes in user visible
	  parts of the --help option.

2020-05-07 07:51  aleidinger

	* libmp3lame/Makefile.in:
	  regen

2020-05-07 07:51  aleidinger

	* libmp3lame/Makefile.am, libmp3lame/VbrTag.c,
	  libmp3lame/VbrTagCrc.h:
	  Faster CRC calculation by Robert Kausch from the fre:ac project
	  (https://freac.org/developer-blog-mainmenu-9/14-freac/277-fastcrc).
	  
	  The gain here is small, but measurable. If you encode multiple
	  independent streams at once and if there are not small (not just
	  a few minutes of play time), you may notice a difference.

2020-05-06 13:17  aleidinger

	* include/libmp3lame.sym:
	  Remove lame_init_old from symbols file. It is made static in
	  lame.h and as
	  such is not available as a symbold for export in libmp3lame.

2020-05-06 12:31  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in, configure,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in,
	  macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, vc_solution/Makefile.in:
	  regen

2020-05-06 12:31  aleidinger

	* config.h.in:
	  regen

2020-05-06 12:30  aleidinger

	* aclocal.m4:
	  regen

2020-05-06 12:29  aleidinger

	* compile, config.guess, config.sub, depcomp, install-sh, missing:
	  Update do recent autotools.

2020-05-06 12:08  aleidinger

	* configure.in:
	  Fix case statements with character ranges. Autoconf is eating up
	  the single '[' and ']'. Duplicating them escapes them.
	  Reported as part of [bugs:#491].
	  
	  While I'm here add gcc v10.* until v19.x to the gcc optimization
	  parts.

2020-05-06 10:06  aleidinger

	* ChangeLog:
	  Add 2 years of changes.

2019-07-23 12:20  robert

	* doc/html/history.html:
	  udating history

2019-07-23 09:34  robert

	* ACM/ddk/msacmdrv.h:
	  fix for tracker item #444 msacmdrv.h (structure packing
	  alignment)

2019-07-23 09:18  robert

	* ACM/ACM.cpp, ACM/AEncodeProperties.cpp:
	  fixed some uninitialzed variables

2019-07-19 13:27  robert

	* libmp3lame/quantize.c:
	  bug fix for tracker item 501 Encoder: Assertion
	  `eov->bitrate_index <= cfg->vbr_max_bitrate_index' failed
	  
	  When encoding a file in ABR mode, the resulting frame size could
	  exceed the targeted amount of bits.
	  This happens, when analog silence is detected and 'analog slince
	  bits' is larger than 'target bits'.

2019-06-28 13:05  robert

	* doc/html/history.html, libmp3lame/lame.c:
	  bug fix for tracker item #500 Buffer overflow in encoder

2019-01-10 23:25  robert

	* libmp3lame/vector/xmm_quantize_sub.c:
	  bug tracker item #496: A critical bug in init_xrpow_core_sse

2018-04-06 19:40  robert

	* doc/html/history.html, frontend/console.c, frontend/console.h,
	  frontend/timestatus.c, frontend/timestatus.h:
	  - patch #82: Take advantage of terminal width on printing file
	  names
	  - some code refactorings

2018-04-05 23:31  robert

	* ACM/tinyxml/tinyxml.cpp, ACM/tinyxml/tinyxmlparser.cpp,
	  Dll/Example.cpp, dshow/Encoder.cpp, dshow/Encoder.h,
	  dshow/Mpegac.cpp, dshow/Mpegac.h, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/lame_main.c, frontend/main.c,
	  frontend/main.h, frontend/parse.c, frontend/parse.h,
	  frontend/rtp.c, libmp3lame/quantize.c, libmp3lame/reservoir.h,
	  libmp3lame/set_get.c, libmp3lame/util.c, libmp3lame/util.h,
	  misc/abx.c, misc/scalartest.c, test/lame_test.cpp:
	  fixing some issues found by the cppcheck tool

2018-03-17 10:53  robert

	* configure.in, doc/html/history.html, frontend/main.c,
	  frontend/parse.c, libmp3lame/i386/nasm.h, libmp3lame/version.h:
	  patch #80 OS/2 patches, submitted by KO Myung-Hun

2018-01-23 20:05  robert

	* frontend/get_audio.c:
	  Apple AIF format 'fl32' now possible to read, even without
	  libsndfile

2017-10-25 16:12  rjamorim

	* doc/html/about.html, doc/html/abr.html, doc/html/cbr.html,
	  doc/html/contact.html, doc/html/contributors.html,
	  doc/html/detailed.html, doc/html/images/valid-html401.png,
	  doc/html/index.html, doc/html/introduction.html,
	  doc/html/list.html, doc/html/ms_stereo.html, doc/html/usage.html,
	  doc/html/vbr.html:
	  Comprehensive HTML formatting. Code now is much more readable

2017-10-25 16:02  rjamorim

	* doc/html/links.html:
	  Minor fix

2017-10-25 15:59  rjamorim

	* doc/html/history.html:
	  Fixed lots of formatting

2017-10-24 18:41  rjamorim

	* doc/html/styles/lame.css:
	  Made the style more similar to the webpage's

2017-10-24 18:30  rjamorim

	* doc/html/history.html:
	  Ooops. Missed one remote image in the history

2017-10-24 18:17  rjamorim

	* doc/html/links.html:
	  Updated links information based on web site

2017-10-24 17:20  rjamorim

	* doc/html/about.html, doc/html/abr.html, doc/html/cbr.html,
	  doc/html/contact.html, doc/html/contributors.html,
	  doc/html/detailed.html, doc/html/images/sflogo.gif,
	  doc/html/index.html, doc/html/introduction.html,
	  doc/html/links.html, doc/html/list.html, doc/html/ms_stereo.html,
	  doc/html/usage.html, doc/html/vbr.html:
	  Replaced remote SF logo with local one. Because we care about
	  privacy!

2017-10-23 19:50  robert

	* configure.in, frontend/parse.c:
	  lame patches ticket #75: Fix for completing svn-r6410
	  submitted by Elio Blanca

2017-10-23 19:17  robert

	* doc/man/lame.1, frontend/get_audio.c:
	  some changes in spelling, suggested by Fabian Greffrath

2017-10-14 15:41  robert

	* frontend/parse.c:
	  bug tracker item #487: v3.100 breaks Windows compatibility
	  
	  the change is incomplete, we need to add 'HAVE_LANGINFO' to our
	  autoconf magic

2017-10-14 14:17  aleidinger

	* ChangeLog:
	  svn2cl does not include branches anymore, this is "current-branch
	  only" now

2017-10-14 14:14  aleidinger

	* Makefile.in:
	  regen

2017-10-14 14:12  aleidinger

	* ChangeLog.header, Makefile.am, doc/html/contact.html,
	  doc/html/contributors.html, doc/html/links.html,
	  make_changelog.sh:
	  switch references from CVS to SVN

2017-10-13 20:39  aleidinger

	* doc/html/history.html:
	  release date of lame 3.100

2017-10-13 20:27  aleidinger

	* ChangeLog, vc_solution/Makefile.in:
	  update

2017-10-13 20:21  aleidinger

	* vc_solution/Makefile.am:
	  list each file explicitely, make dist does not like globs.

2017-10-13 20:16  aleidinger

	* testcase.mp3:
	  Update testcase for 3.100 release.
	  Build on FreeBSD 64bit, clang 5.0.0, no advanced compiler
	  flags/optimisations.

2017-10-10 19:31  aleidinger

	* ChangeLog:
	  update

2017-09-26 12:25  robert

	* frontend/parse.c:
	  forcing MS for j-stereo mode only

2017-09-26 12:14  robert

	* Dll/BladeMP3EncDLL.c, frontend/parse.c, libmp3lame/lame.c:
	  forcing MS for j-stereo mode only

2017-09-26 10:31  robert

	* libmp3lame/lame.c:
	  set 'use of safe joint' for joint-stereo only, (only effects a
	  bit in vbr tag, sometimes)

2017-09-21 11:00  robert

	* libmp3lame/gain_analysis.c, libmp3lame/machine.h:
	  + fix multiple_of macro

2017-09-19 16:58  robert

	* libmp3lame/version.h:
	  release candidate

2017-09-19 16:57  robert

	* frontend/parse.c:
	  patch for bug #484: Lame doesn't recognize user locale correctly

2017-09-06 19:40  aleidinger

	* ChangeLog:
	  update

2017-09-06 19:38  aleidinger

	* libmp3lame/psymodel.c:
	  Add missing (at least in FreeBSD) include for FLT_EPSILON.

2017-09-06 15:07  robert

	* frontend/get_audio.c, libmp3lame/fft.c,
	  libmp3lame/gain_analysis.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/set_get.c,
	  libmp3lame/takehiro.c, libmp3lame/util.c, libmp3lame/util.h,
	  mpglib/interface.c, mpglib/tabinit.c:
	  small fixes

2017-08-31 14:14  robert

	* frontend/lame_main.c, include/lame.def, include/lame.h,
	  include/libmp3lame.sym, libmp3lame/bitstream.c,
	  libmp3lame/lame.c, libmp3lame/set_get.c:
	  another go to fix bug #474: Buffer Overflow possible when input
	  sample rate very low
	  
	  - do_copy_buffer did not check buffer size, when size was zero!
	  0 was a bad choice for the purpose of not checking, because
	  it was possible to get there, even when checking was demeanded!

2017-08-30 20:48  robert

	* libmp3lame/lame.c:
	  fix for bug #474: Buffer Overflow possible when input sample rate
	  very low
	  
	  When the input sample rate is very low, like 1 Hz, resampling to
	  8 kHz can result in a buffer overflow.
	  Found by Gustavo Grieco.

2017-08-30 14:11  robert

	* frontend/lame_main.c, frontend/parse.c, frontend/parse.h:
	  fix for bug #473: Buffer Overflow possible in lame_main.c
	  - replaced parse_nogap_filename by generateOutPath
	  (now there is only one function building the output file names)
	  - all nogap file names are now build before any encoding starts
	  - decode and nogap are no valid combination of arguments anymore,
	  wouldn't work as expected

2017-08-28 15:39  robert

	* libmp3lame/id3tag.c:
	  fix for bug #472: NULL Pointer Dereference in
	  id3v2AddAudioDuration
	  - plus some more pointer checks

2017-08-28 12:48  robert

	* frontend/gtkanal.c:
	  fixing resource leak, hip deocder wasn't closed on quit

2017-08-26 13:20  robert

	* libmp3lame/lame.c, libmp3lame/set_get.c:
	  lame_init_params:
	  - removed some impossible returns

2017-08-26 10:54  robert

	* libmp3lame/encoder.c:
	  more of psy downgrade

2017-08-24 20:43  robert

	* frontend/mp3rtp.c:
	  - free was called twice for some argument c-string

2017-08-23 13:22  robert

	* mpglib/interface.c, mpglib/layer1.c:
	  hip layer 1 decode: testing for illegal bit allocation value

2017-08-22 23:31  robert

	* mpglib/layer2.c:
	  fix for bug #470: global-buffer-overflow mpglib/layer2.c:144
	  II_step_one

2017-08-22 19:44  robert

	* libmp3lame/id3tag.c, libmp3lame/id3tag.h, libmp3lame/util.c:
	  id3v2: set 'language'property to a valid language code (COMM
	  field)

2017-08-22 16:02  robert

	* doc/html/history.html, frontend/get_audio.c,
	  frontend/lame_main.c:
	  fixing bug i#434: --clipdetect scale suggestion unaware of
	  --scale input value

2017-08-21 17:10  robert

	* frontend/get_audio.c:
	  fix for bug #469 Unable to build without mpglib

2017-08-20 20:06  robert

	* mpglib/layer3.c:
	  It looks like 'intensity stereo' decoding is broken. This change
	  prevents a buffer overflow (read) and instead prints an
	  indication, that there is something wrong.

2017-08-20 13:56  robert

	* mpglib/layer3.c:
	  fixing map for mixed-block band infos

2017-08-19 20:17  robert

	* mpglib/layer3.c:
	  fix for #467: stack-buffer-overflow mpglib/layer3.c:945
	  III_dequantize_sample
	  
	  fix for #468 global-buffer-overflow mpglib/layer3.c:1244
	  III_i_stereo

2017-08-19 14:20  robert

	* mpglib/common.c:
	  fix for bug #465: global-buffer-overflow layer2.c:144 II_step_one
	  - mpeg-2.5 is only supported by Layer3

2017-08-19 12:51  aleidinger

	* ChangeLog:
	  update

2017-08-18 22:06  robert

	* mpglib/layer3.c:
	  fix for bug #466: "stack-buffer-overflow in
	  ./mpglib/layer3.c:1264 III_i_stereo"

2017-08-18 21:56  robert

	* frontend/get_audio.c, frontend/lame_main.c:
	  fix for bug #464: attempting double-free

2017-08-15 19:51  robert

	* libmp3lame/set_get.c:
	  estimation of total number of frames changed

2017-08-15 15:40  aleidinger

	* make_changelog.sh:
	  add comment how to generate the input file

2017-08-15 15:35  aleidinger

	* ChangeLog:
	  update ChangeLog

2017-08-15 15:20  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, configure, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, vc_solution/Makefile.in:
	  regenerate

2017-08-15 15:16  aleidinger

	* configure.in:
	  Improve compiler version detection (clang changed version
	  string).
	  Add some future GCC versions in the optimisation part.

2017-08-15 15:15  aleidinger

	* include/lame.def:
	  Add lame_encode_buffer_interleaved_int() by Michael Fink.
	  LAME Patch #27.

2017-08-15 14:51  aleidinger

	* include/lame.h, include/libmp3lame.sym, libmp3lame/lame.c:
	  patch #27: Add lame_encode_buffer_interleaved_int() by Michael
	  Fink

2017-08-15 14:46  aleidinger

	* compile, config.sub, depcomp, frontend/depcomp, install-sh,
	  libmp3lame/depcomp, misc/depcomp, missing, mpglib/depcomp:
	  update to a more recent autotools version

2017-08-13 16:18  robert

	* libmp3lame/set_get.c:
	  prevent some rare overflow in estimation of total number of
	  frames
	  in case samplerate-out is higher than samplerate-in
	  (it is a visual glitch only)

2017-08-13 13:57  robert

	* doc/html/history.html, frontend/get_audio.c,
	  libmp3lame/version.h:
	  - sanity check when reading pcm data, avoid buffer overflow
	  - version 3.100 beta 0

2017-08-12 18:56  robert

	* frontend/get_audio.c, frontend/lame_main.c, frontend/main.c,
	  mpglib/common.c, mpglib/interface.c:
	  - avoid calling exit in library
	  - free up resources after error condition in frontend

2017-08-12 18:14  robert

	* mpglib/layer3.c:
	  - workaround for heap corruption while decoding layer3
	  - some fix for mpeg2.5 decoding with mixed block feature, taken
	  from Takehiro's branch

2017-08-08 17:52  robert

	* doc/man/lame.1:
	  fix typo

2017-08-08 13:56  robert

	* doc/html/history.html, libmp3lame/vector/xmm_quantize_sub.c:
	  Fix for bug ticket [ #449 ] Make sure functions with SSE
	  instructions maintain their own properly aligned stack. Thanks to
	  Fabian Greffrath

2017-08-08 13:26  robert

	* doc/html/history.html, mpglib/interface.c:
	  Fix for bug ticket #445: dereference of a null pointer possible
	  in loop. Thanks to Renu Tyagi

2017-08-08 12:40  robert

	* doc/html/history.html, frontend/lame_main.c:
	  Fix for bug ticket #447: Fix dereference NULL and Buffer not NULL
	  terminated issues. Thanks to Surabhi Mishra

2017-08-06 18:15  robert

	* frontend/get_audio.c, libmp3lame/VbrTag.c, libmp3lame/lame.c,
	  libmp3lame/set_get.c, libmp3lame/util.h:
	  in case of illegal sample rate and number of channel, let lame
	  print an error message

2017-08-03 13:07  rjamorim

	* HACKING, INSTALL, LICENSE, README, README.WINGTK, TODO:
	  Typos, formatting, grammar...

2017-08-03 13:06  rjamorim

	* ACM/lame.ico:
	  Updated icon to the current LAME logo

2017-08-03 12:32  rjamorim

	* USAGE:
	  Updated USAGE: fixed typos, formatting, some grammar

2017-08-02 19:56  robert

	* doc/html/history.html:
	  - reverting psymodel to that of version 3.99.5
	  - changing version back to 3.99.5 to ease binary mp3 file
	  comparisons with version 3.99.5
	  
	  Committed on the Free edition of March Hare Software CVSNT
	  Client.
	  Upgrade to CVS Suite for more features and support:
	  http://march-hare.com/cvsnt/

2017-08-02 19:48  robert

	* libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.h,
	  libmp3lame/version.h:
	  - reverting psymodel to that of version 3.99.5
	  - changing version back to 3.99.5 to ease binary mp3 file
	  comparisons with version 3.99.5
	  
	  Committed on the Free edition of March Hare Software CVSNT
	  Client.
	  Upgrade to CVS Suite for more features and support:
	  http://march-hare.com/cvsnt/

2016-01-29 21:36  aleidinger

	* ChangeLog:
	  update history

2016-01-29 21:09  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, compile, config.guess, config.h.in, config.sub,
	  configure, depcomp, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  frontend/depcomp, include/Makefile.in, install-sh,
	  libmp3lame/Makefile.in, libmp3lame/depcomp,
	  libmp3lame/i386/Makefile.in, libmp3lame/vector/Makefile.in,
	  ltmain.sh, mac/Makefile.in, macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, misc/depcomp, missing, mpglib/Makefile.in,
	  mpglib/depcomp, vc_solution/Makefile.in:
	  regen

2016-01-29 21:06  aleidinger

	* Dll/Makefile.mingw32, HACKING, Makefile.MSVC, Makefile.am.global,
	  Makefile.unix, configure.in, libmp3lame/Makefile.am:
	  - better check for SSE intrinsics [1]
	  - don't define NDEBUG in release builds
	  -> asserts trigger segfaults even in release builds
	  -> quick and user-unfriendly fix until user-fiendly error checks
	  are
	  implemented
	  - fix spelling mistakes in HACKING file [2]
	  - enable cross-compiling for mingw [3]
	  - fix linker path for libmpelame for mingw [3]
	  - fix out-of-source-tree builds [4]
	  - update files to automake 1.15, autconf 2.69
	  - update compiler version detection (detect clang and add
	  optimization
	  for at least 3.7.x)
	  
	  Submitted by: dim@FreeBSD.org [1], FreeBSD PR 206620
	  tridecethtwelve@users.sf.net [2] LAME patch #68
	  bardware@users.sf.net [3] LAME patch #70
	  m-a@users.sf.net [4] LAmE patch #67

2015-05-11 07:01  rbrito

	* libmp3lame/lame.c:
	  libmp3lame: Remove redundant check for invalid sample rates.
	  
	  Thanks Fabian Greffrath for letting me know that I screwed up. :)
	  
	  Signed-off-by: Rogério Brito <rbrito@ime.usp.br>

2015-04-28 22:12  rbrito

	* libmp3lame/lame.c:
	  libmp3lame: Add check for negative number of channels.
	  
	  Thanks to Maks Naumov and Fabian Greffrath for the contribution.

2015-04-28 22:07  rbrito

	* libmp3lame/util.c:
	  libmp3lame: Fix decision if sample rate ratio is an integer value
	  or not.
	  
	  From Fabian Greffrath's original description:
	  
	  If the sample rate of the input file is sufficiently close to an
	  integer
	  multiple of the output sample rate, the value of the intratio
	  variable is
	  calculated incorrectly. This leads to further values being
	  miscalculated
	  up to the joff variable which is used as an index to dereference
	  the
	  esv->blackfilt array. This leads top an overflow and causes a
	  segmentation fault.
	  
	  See https://bugs.debian.org/778529 for a discussion.

2015-04-28 22:03  rbrito

	* frontend/get_audio.c:
	  frontend: Avoid division by zero.
	  
	  Avoid malformed wav causing floating point exception (integer
	  divide by
	  zero).
	  
	  Kindly contributed by Fabian Greffrath
	  <fabian+debian@greffrath.com>. See
	  https://bugs.debian.org/777159 for a discussion.

2015-02-21 23:03  rbrito

	* test/lametest.py:
	  test: lametest: Some more PEP-8 love.
	  
	  This is mostly regarding not having parentheses around boolean
	  expressions.

2015-02-21 22:57  rbrito

	* test/lametest.py:
	  test: lametest: Add some PEP-8 love.

2015-02-21 22:49  rbrito

	* test/lametest.py:
	  test: lametest: Fix some trailing whitespace.

2015-02-21 22:35  rbrito

	* libmp3lame/lame.c:
	  Add check for invalid input sample rate.
	  
	  Patch by Maks Naumov <maksqwe1@ukr.net>, forwared to me by Fabian
	  Greffrath
	  <fabian@greffrath.com> from the Debian packaging.
	  
	  Thanks for the contribution!

2014-10-01 17:54  jaz001

	* doc/html/links.html:
	  links update.

2014-01-28 14:48  robert

	* Makefile.MSVC, frontend/get_audio.c, frontend/main.c,
	  libmp3lame/lame.c, libmp3lame/machine.h:
	  some minor changes:
	  - do not disable replaygain calc and find peak sample, when not
	  writing LAME info header (patch tracker item #66)
	  - remove deprecated linker option from Makefile.MSVC (bug tracker
	  item #442)

2013-12-17 11:24  robert

	* frontend/get_audio.c, frontend/main.h, frontend/parse.c,
	  libmp3lame/gain_analysis.c, libmp3lame/id3tag.h,
	  libmp3lame/machine.h, libmp3lame/psymodel.c:
	  applying submitted patches
	  #64: fixes undefined references when compiling with mingw32 (send
	  by Patrick)
	  #59: changes static_assert into a more portable form (send by
	  Doug Cook)
	  #55: fixes issues with wave files >2GB (send by lvqcl)
	  
	  plus some minor changes to reduce compiler warnings with GCC
	  v4.7.1

2013-12-08 19:49  jaz001

	* LICENSE, USAGE, doc/html/detailed.html, doc/man/lame.1,
	  frontend/parse.c:
	  Documentation updates:
	  -q setting rewording in all docs
	  general updates in USAGE and man page.
	  Small updates in the --help, --longhelp , and --preset help

2013-06-12 09:16  rbrito

	* aclocal.m4, configure.in, doc/html/Makefile.am,
	  doc/man/Makefile.am, libmp3lame/i386/Makefile.am:
	  Make autotools files compatible with automake >= 1.12
	  
	  Newer autotools versions don't seem to support ansi2knr, which is
	  probably
	  not a problem, given the last decades of compilers accepting ANSI
	  C (and the
	  rest of the LAME code being written in ANSI C itself).
	  
	  This fixes ticket #60, with a patch of Brennan Shacklett, from
	  the Gentoo
	  community.

2013-06-12 09:06  rbrito

	* aclocal.m4, configure.in, doc/html/Makefile.am:
	  Kill spurious trailing whitespace from some files.

2013-06-11 08:41  rbrito

	* frontend/brhist.c:
	  frontend: brhist: Fix type of formatting string parameter.

2013-06-11 08:41  rbrito

	* frontend/get_audio.c:
	  frontend: get_audio: Add missing formatting string parameter.

2013-06-11 08:19  rbrito

	* Dll/Example.cpp:
	  Dll/Example.cpp: Remove (some) spurious whitespace.

2013-06-11 08:19  rbrito

	* Dll/Example.cpp:
	  Dll/Example.cpp: Remove resource leaks in example code.

2013-06-11 07:26  rbrito

	* config.guess, config.sub:
	  config.{sub,guess}: Update files to pick up newer arches.
	  
	  This closes ticket #406. Perhaps a better solution would be to
	  rewrite part
	  of our autotools system, but, then, perhaps not.

2013-06-11 06:46  rbrito

	* include/libmp3lame.sym:
	  libmp3lame: Regenerate libmp3lame.sym from header file.
	  
	  Part of patch submitted by Reimar Döffinger on ticket #63.
	  
	  We should review this to check if there are any symbols that we
	  don't want
	  exported.

2013-06-11 06:45  rbrito

	* libmp3lame/Makefile.am:
	  libmp3lame: Consider the symbols in libmp3lame.sym.
	  
	  This is part of the patch sent by Reimar Döffinger in ticket #63.

2013-03-20 20:41  robert

	* doc/html/history.html, frontend/timestatus.c:
	  fix for tracker item 3608583: input file name displayed with
	  wrong character encoding (windows)

2012-10-25 18:46  robert

	* doc/html/history.html, frontend/lame_main.c, frontend/main.c,
	  frontend/parse.c:
	  fix for tracker item 3580176: environment variable LAMEOPT
	  doesn't work anymore

2012-10-13 14:10  robert

	* doc/html/history.html, frontend/parse.c:
	  fix for tracker item 3544957: scale (empty) silent encode without
	  warning
	  
	  switches with missing argument values will be ignored. (Those
	  with integer or floating point values, at least)

2012-10-13 11:50  robert

	* doc/html/history.html, frontend/parse.c:
	  fix for tracker item 3565659: no progress indication with pipe
	  input :(

2012-10-13 10:36  robert

	* doc/html/history.html, frontend/parse.c:
	  fix for tracker item 3567844: problem with Tag genre

2012-10-13 09:33  robert

	* doc/html/history.html, frontend/parse.c:
	  fix for tracker item ID 3558466: Bug in path handling

2012-10-09 21:39  robert

	* libmp3lame/version.h:
	  bump version

2012-10-09 21:37  robert

	* libmp3lame/psymodel.c, libmp3lame/util.h:
	  trying to improve strong tonal samples (lead-voice.wav)

2012-10-09 21:35  robert

	* frontend/parse.c:
	  ignore developer-only switches, but parse them for command line
	  compatibility reasons

2012-10-06 15:54  robert

	* libmp3lame/psymodel.c:
	  - code refactorings
	  - changed "mask adding", get the max of 3 different ones

2012-10-06 11:44  robert

	* frontend/parse.c, libmp3lame/presets.c:
	  + new command line option "--gain x", more convenient to adjust
	  replay gain, just use whatever decibel value LAME reported
	  before.
	  + out-dir option mentioned in long help text

2012-07-19 20:57  robert

	* doc/html/history.html, frontend/get_audio.c,
	  libmp3lame/version.h, mpglib/interface.c:
	  Improved detection of MPEG audio data in RIFF WAVE files.
	  see tracker item 3545112: Invalid sampling detection

2012-07-11 18:31  jaz001

	* doc/html/detailed.html, doc/html/usage.html:
	  Updates to reflect recent changes in hydrogenaudio wiki.

2012-06-19 00:15  rbrito

	* doc/html/history.html:
	  Update the HTML changelog.

2012-06-19 00:00  rbrito

	* Makefile.am, configure.in:
	  Don't use use the debian directory as build target.
	  
	  Include debian as extra directory instead of directory part of
	  subdirectories used in build.
	  
	  Patch stolen from:
	  
	  http://anonscm.debian.org/gitweb/?p=pkg-multimedia/lame.git;a=tree;f=debian/patches

2012-06-18 20:52  rbrito

	* libmp3lame/i386/Makefile.am:
	  libmp3lame: Fix possible race condition causing build failures.
	  
	  This was discovered in automated builds by the Debian project and
	  the patch
	  provided is by Andres Mejia <amejia@debian.org>.
	  
	  Please, note that only part of the patch (the one touching
	  Makefile.am) was
	  applied, since the changes in Makefile.in should be picked up
	  when
	  regenerating files with autotools.
	  
	  Signed-off-by: Rogério Brito <rbrito@ime.usp.br>

2012-06-18 20:51  rbrito

	* acinclude.m4:
	  acinclude.m4: Include GTK-1 autoconf directives in build system.
	  
	  Resurrect knowledge of how to deal with GTK1. This was
	  transplanted back
	  from aclocal.m4 with a patch provided by Andres Mejia
	  <amejia@debian.org>,
	  from Debian's packaging of lame, even though the original code is
	  attributed
	  to Owen Taylor and dated 97-11-3.
	  
	  Anyway, with this part not lost again, we can easily regenerate
	  autotools'
	  files with a simple `autoconf -vfi` invocation.
	  
	  Signed-off-by: Rogério Brito <rbrito@ime.usp.br>

2012-06-18 20:48  rbrito

	* configure.in:
	  configure.in: Rewrap for (human) legibility.
	  
	  This also helps by being diff-friendly if we change only one of
	  the parts of
	  a long line.
	  
	  Signed-off-by: Rogério Brito <rbrito@ime.usp.br>

2012-02-28 19:07  robert

	* doc/html/history.html:
	  bug-fix release 3.99.5

2012-02-19 00:02  robert

	* libmp3lame/gain_analysis.c, libmp3lame/machine.h:
	  rearranging gain analysis kernel table and calc loops to allow
	  better code vectorization

2012-02-18 13:09  robert

	* frontend/get_audio.c, frontend/lame_main.c, frontend/main.c,
	  frontend/parse.c, libmp3lame/mpglib_interface.c, mpglib/layer2.c:
	  removing VC11 x86-64 compiler warnings

2012-02-14 14:24  robert

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/id3tag.c, libmp3lame/lame.c,
	  libmp3lame/machine.h, libmp3lame/mpglib_interface.c,
	  libmp3lame/psymodel.c, libmp3lame/util.c, libmp3lame/util.h:
	  some code refactorings

2012-02-14 11:06  robert

	* ChangeLog, doc/html/history.html:
	  updating history/changelog

2012-02-11 12:11  robert

	* frontend/get_audio.c:
	  Bug fix for tracker item ID: 3486753 Artifacts at the beginning
	  of decoded file

2012-02-11 12:10  robert

	* mpglib/layer3.c:
	  some hidden global temporary variable removed, could have been
	  some multi threading issue

2012-02-10 18:13  rbrito

	* config.h.in, configure.in:
	  autotools: Fix compilation on alpha using proper ifdef guards.
	  
	  Thanks to Andres Mejia <amejia004@gmail.com> for the patch (with
	  minor
	  whitespace adjustments).

2012-02-10 18:11  rbrito

	* doc/html/history.html:
	  doc: Correctly spell Fabian Greffrath's name to give proper
	  credit.

2012-02-07 13:52  robert

	* configMS.h:
	  reduce compiler warnings for SSE2 build with VC 11

2012-02-07 13:40  robert

	* configMS.h, libmp3lame/l3side.h, libmp3lame/machine.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/vbrquantize.c:
	  - workaround for cygwin build problem
	  - fix "fast floating point math" problem (possible bitrate bloat
	  on quiet pieces)

2012-01-25 13:01  robert

	* ChangeLog, doc/html/history.html:
	  updating Changelog and History

2012-01-19 12:15  robert

	* frontend/parse.c:
	  fix for tracker item ID: 3475581 - lame crashes at .w64 input
	  file

2012-01-08 23:50  robert

	* include/lame.h, libmp3lame/id3tag.c, libmp3lame/id3tag.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.c:
	  Addressing things brought to attention by tracker item ID:
	  3463197: 3.99.x problem WFED and PCST frames
	  
	  + WFED and PCST frames can now be added, to tag podcasts iTunes
	  recognizes
	  + USER frames are now supported
	  + COMM frames can now have a description, when passed via --tv
	  "COMM=<description>=<full text>"
	  
	  - possible divide-by-zero exception should be fixed
	  - adding malformed user-defined-frames could result in abnormal
	  program termination, fixed

2011-12-04 12:35  robert

	* libmp3lame/lame.rc:
	  re tracker ID: 3449748 Incomplete resource definition file
	  A translation value seems to be mandatory by convention.

2011-11-26 18:26  robert

	* doc/html/history.html, libmp3lame/id3tag.c:
	  bug fix for tracker item ID: 3441349: --tg does not handle genre
	  number when adding unicode tag

2011-11-18 09:51  robert

	* doc/html/history.html, libmp3lame/VbrTag.c, libmp3lame/version.c:
	  merge with 3.99 branch modifications

2011-11-10 18:46  robert

	* libmp3lame/version.c:
	  small fix

2011-11-10 13:19  robert

	* frontend/get_audio.c, frontend/get_audio.h, frontend/lame_main.c:
	  when transcoding, copy old id3v2 tag. feature requests:
	  ID: 1985551 lame --pass-ID3-tags-unscathed
	  ID: 977304 ID3 tags lost when transcoding

2011-11-09 00:15  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  aiming for a more balanced bit reservoir usage

2011-11-08 19:36  robert

	* libmp3lame/quantize_pvt.c:
	  at lower sample rates, make a cut at last usable scalefactor band

2011-11-08 18:19  robert

	* libmp3lame/quantize.c:
	  8kHz mpeg2.5 uses fewer scale factor bands.

2011-11-06 18:34  robert

	* libmp3lame/encoder.c:
	  use same meaning of PE in cbr and vbr modes

2011-11-05 17:53  robert

	* configure, configure.in, doc/html/history.html,
	  libmp3lame/version.h:
	  main-branch open for next development cycle

2011-11-05 16:32  robert

	* misc/Makefile.am, misc/Makefile.in:
	  adding mk_mp3.sh to dist files

2011-11-02 00:20  robert

	* Makefile.MSVC, libmp3lame/machine.h:
	  fixing vc6 compile problem, VC6 seems to miss exporting some
	  inline functions from math.h.

2011-11-01 19:50  robert

	* libmp3lame/id3tag.c:
	  oops, fixing writing unicode strings in id3tags

2011-11-01 16:59  robert

	* doc/html/history.html, frontend/parse.c, libmp3lame/id3tag.c:
	  Fixes for several issues with ID3v2 unicode tags, using
	  Big-Endian text
	  encodings. Because of several other software (like Windows Media
	  Player),
	  LAME writes Little-Endian unicode tags only.
	  Thanks to Taihei Monma, for reporting these issues.
	  Tracker items: [ 3431203, 3431222, 3431241 ]

2011-10-18 21:54  robert

	* Dll/BladeMP3EncDLL.h:
	  Fixing typedef HBE_STREAM, has to be a pointer type.

2011-10-18 21:51  robert

	* libmp3lame/lame.c, libmp3lame/util.c:
	  Fixing a small display problem for x64 platform

2011-10-17 23:15  robert

	* libmp3lame/id3tag.c:
	  bug fix: ID3 tags are always added

2011-10-16 13:09  robert

	* frontend/Makefile.am, frontend/Makefile.in:
	  file rtp.h was missing in distribution list

2011-10-15 15:46  robert

	* doc/html/history.html:
	  Updating history

2011-10-15 14:37  robert

	* ChangeLog:
	  updating Changelog

2011-10-15 14:31  robert

	* debian/Makefile.am, debian/Makefile.in, doc/html/detailed.html,
	  libmp3lame/version.h, testcase.mp3:
	  preparing release

2011-10-15 12:38  robert

	* misc/mk_mp3.sh:
	  updating script for building mp3 collection from lossless archive

2011-10-06 23:12  robert

	* ChangeLog:
	  updating changelog

2011-10-06 23:07  robert

	* frontend/get_audio.c, frontend/parse.c:
	  let iconv use transliteration

2011-10-04 15:37  robert

	* frontend/parse.c:
	  fixing compiler warning

2011-10-04 11:42  robert

	* frontend/get_audio.c, frontend/lame_main.c, frontend/main.h,
	  frontend/parse.c:
	  using our input file reading functions when reading from stdin
	  allow -s to overwrite input samplerate for non RAW input files
	  too

2011-10-03 13:28  robert

	* doc/html/history.html, frontend/lame_main.c:
	  fix for tracker item ID: 3034259 Fix for confusing fatal error:
	  can't update LAME-tag frame!

2011-10-03 13:22  robert

	* frontend/parse.c, libmp3lame/id3tag.c:
	  making sure, the id3v1 tags get filled

2011-10-02 21:45  robert

	* doc/html/history.html, frontend/get_audio.c:
	  Bug fix for tracker item ID: 3368977 Wave Parser doesn't pad
	  chunks

2011-10-02 18:39  robert

	* frontend/portableio.c, frontend/portableio.h:
	  removing portableio.[ch] files, not needed anymore

2011-10-02 18:24  robert

	* frontend/parse.c, libmp3lame/id3tag.c:
	  fixing some compiler warnings

2011-10-02 17:13  robert

	* Makefile.MSVC, Makefile.unix, frontend/Makefile.am,
	  frontend/Makefile.in, frontend/get_audio.c, frontend/mp3rtp.c,
	  vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3rtp.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj:
	  replacing "portableio" by own code in get_audio.c

2011-10-02 14:52  robert

	* doc/html/detailed.html, doc/man/lame.1, frontend/main.c,
	  frontend/main.h, frontend/parse.c, include/lame.def,
	  include/lame.h, libmp3lame/id3tag.c:
	  clearify, unicode tags are utf-16

2011-09-28 19:11  robert

	* doc/html/history.html, frontend/parse.c, libmp3lame/id3tag.c,
	  vc_solution/vc9_lame.sln, vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3rtp.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_mpglib.vcproj:
	  bug fix for tracker item 3395813: id3tag_set_fieldvalue for URL
	  link frames Broken
	  
	  now one can add URL link frames to the ID3v2 tag, like:
	  --tv "WOAR=www.mypage.com" --tv "WXXX=get remixes
	  here=www.dj.org"

2011-09-25 17:08  robert

	* libmp3lame/version.h:
	  increasing patch level

2011-07-26 11:48  rbrito

	* frontend/parse.c:
	  frontend: Update licensing message.
	  
	  Patch supplied by Andres Mejia <mcitadel@gmail.com> according to
	  previous
	  discussions in the mailing lists `lame-dev` and
	  `pkg-multimedia-maintainers`.
	  
	  http://lists.alioth.debian.org/pipermail/pkg-multimedia-maintainers/2011-July/020417.html

2011-06-17 06:22  aleidinger

	* Makefile.am.global:
	  Remove ansi2knr part. It will be deprecated and subsequently
	  removed from
	  automake.
	  
	  Notified by: Stefano Lattarini <stefano.lattarini@gmail.com>

2011-05-24 20:45  robert

	* Makefile.unix, libmp3lame/psymodel.c, libmp3lame/quantize_pvt.c:
	  small change in addition of simultaneous maskings

2011-05-24 20:40  robert

	* mpglib/layer2.c:
	  workaround for tracker item 3301963:"assertion failed in layer 2
	  decoding"

2011-05-07 16:05  rbrito

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h, Dll/Example.cpp,
	  README, debian/copyright, frontend/brhist.c, frontend/brhist.h,
	  frontend/get_audio.c, frontend/get_audio.h,
	  frontend/gpkplotting.c, frontend/gpkplotting.h,
	  frontend/gtkanal.c, frontend/gtkanal.h, frontend/lame_main.c,
	  frontend/lametime.c, frontend/lametime.h, frontend/main.c,
	  frontend/main.h, frontend/mp3rtp.c, frontend/parse.c,
	  frontend/rtp.c, frontend/rtp.h, frontend/timestatus.c,
	  frontend/timestatus.h, include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/VbrTag.h, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/fft.h, libmp3lame/id3tag.c,
	  libmp3lame/l3side.h, libmp3lame/lame-analysis.h,
	  libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/newmdct.h, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/set_get.c, libmp3lame/set_get.h, libmp3lame/tables.c,
	  libmp3lame/tables.h, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h, libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/xmm_quantize_sub.c, libmp3lame/version.c,
	  libmp3lame/version.h, test/lame_test.cpp:
	  Make explicit that the licence is the Library GPLv2.
	  
	  According to Jonas Smedegaard, there is no Lesser GPLv2, only
	  Lesser GPLv2.1.
	  
	  Thanks to Andres.

2011-05-07 15:56  rbrito

	* frontend/parse.c:
	  Fix some warnings in 64-bit arches. Changing s/size_t/unsigned
	  int/
	  
	  This should not be a problem with the variables in question, as
	  they can't
	  be much larger than what an unsigned int is able to hold.
	  
	  Patch from:
	  
	  http://git.debian.org/?p=pkg-multimedia/lame.git;a=blob;f=debian/patches/07-field-width-fix.patch
	  
	  Sent by Andres.

2011-05-07 15:44  rbrito

	* configure.in:
	  Work round buggy softfloat optimization in ARM EABI compilers.
	  
	  A bug present in gcc-4.[123] up to 4.3.3 makes a buggy
	  optimization
	  for single-precision floats when -ffinite-math-only (enabled by
	  -ffast-math)
	  Refer to the following.
	  
	  https://sourceforge.net/tracker/index.php?func=detail&aid=2681646&group_id=290&atid=100290
	  https://trac.xiph.org/ticket/1526
	  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39501
	  
	  Thanks to Andres Mejia. This fixes SF bug 2681646.

2011-05-07 15:35  rbrito

	* libmp3lame/i386/nasm.h:
	  Disable executable patch.
	  
	  Patch from Russell Coker to disable executable stack from
	  Christian
	  Marillat's packaging. Sent by Andres Mejia.

2011-05-07 01:23  rbrito

	* README:
	  Tweak slightly the disclaimer of patents from the README file.
	  
	  This change was proposed by me with suggestions made by Andres
	  Mejia.

2011-05-06 20:14  rbrito

	* libmp3lame/lame.c:
	  libmp3lame: lame: Put some parentheses to shut up some static
	  analysis tools.
	  
	  The statement
	  
	  frames_left -= ((frame_num != gfc->ov_enc.frame_number) ? 1 : 0);
	  
	  could probably be written as
	  
	  
	  frames_left -= (frame_num != gfc->ov_enc.frame_number);
	  
	  unless we know of some platform where a boolean evaluation does
	  not result
	  in a 1 in case of it being true.

2011-05-06 19:59  rbrito

	* README:
	  Remove problematic restriction to the LGPL, to make it legally
	  OK.
	  
	  Also update the copyright years of the LAME team.

2011-04-20 13:03  aleidinger

	* config.rpath:
	  looks like this is needed now

2011-04-20 12:53  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, config.h.in, configure, debian/Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in,
	  macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, vc_solution/Makefile.in:
	  regen

2011-04-20 12:51  aleidinger

	* configure.in:
	  - correct the VBR check message to something more correct
	  (curses)
	  - check for libiconv and use it in the frontend

2011-04-19 15:36  robert

	* doc/man/lame.1, frontend/parse.c:
	  lame(1) manpage patch - ID: 3161674
	  Typo in --longhelp - ID: 3288886
	  replaced strnicmp, isn't portable

2011-04-15 20:17  robert

	* libmp3lame/lame.c:
	  Bug fix: when aborting an encode before any samples were fed into
	  the encoder, an assert was falsely triggered.

2011-04-14 20:48  robert

	* frontend/parse.c:
	  Allow -Z switch in none alpha builds (for now)

2011-04-14 19:36  robert

	* doc/html/history.html, frontend/parse.c, libmp3lame/presets.c,
	  libmp3lame/version.h:
	  3.99 beta phase

2011-04-08 23:50  robert

	* misc/mk_mp3.sh:
	  mk_mp3.sh - script to encode and tag CDs

2011-04-08 22:55  robert

	* libmp3lame/psymodel.c:
	  some code cleanup: unused vars, defines, etc.

2011-04-08 19:26  robert

	* frontend/lame_main.c, frontend/main.c, frontend/main.h:
	  Unicode support for extra options passed via environment varialbe
	  LAMEOPT

2011-04-07 23:03  robert

	* frontend/get_audio.c:
	  quick fix for tracker item 3277412: Segmentation fault at
	  layer2.c:105 with
	  --decode
	  (problem sample contains multiple ID3v2 tags)

2011-04-07 23:01  robert

	* libmp3lame/psymodel.c, libmp3lame/util.h:
	  - some code cleanup
	  - calculate short block thm from long blocks, in case we
	  eventually
	  may need it for pre-echo control of the next granule

2011-04-05 19:17  jaz001

	* doc/html/detailed.html:
	  corrected spelling.

2011-04-04 19:40  robert

	* libmp3lame/psymodel.c:
	  small update for the pre-echo control

2011-04-04 11:49  aleidinger

	* ChangeLog:
	  Update upto now.

2011-04-04 11:48  aleidinger

	* make_changelog.sh:
	  Add the script to generate the changelog, not added to
	  Makefile.am as it
	  does not make sense to distribute it.
	  
	  I keep the "cvs log" output in '../log_output', the script
	  expects it there.

2011-04-04 09:54  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, config.h.in, configure, debian/Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in,
	  macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, vc_solution/Makefile.in:
	  regen

2011-04-04 09:49  aleidinger

	* configure.in:
	  - depend upon at least autoconf 2.68, this should fix the
	  variable type
	  size check in configure on Windows (mingw)
	  - readd check for termcap lib (needed for brhist)
	  - remove float8 (requested by Robert Hegemann)
	  - my gcc moans about -march=generic not being supported
	  (anymore), use
	  'native' instead of 'generic' in one place

2011-04-04 09:45  aleidinger

	* Makefile.am.global:
	  Depend upon automake 1.11.

2011-04-04 09:42  aleidinger

	* libmp3lame/i386/Makefile.am:
	  Create the directory in any case, seems to prevent a race
	  condition.

2011-04-04 09:41  aleidinger

	* configMS.h:
	  Remove superflous brhist define.

2011-04-04 09:40  aleidinger

	* INSTALL:
	  Remove superflous explication about brhist switch.

2011-04-04 09:36  aleidinger

	* depcomp, frontend/depcomp, install-sh, libmp3lame/depcomp,
	  ltmain.sh, misc/depcomp, missing, mpglib/depcomp:
	  Update files from automake 1.11 and libtool 2.2.10.

2011-04-03 22:26  robert

	* libmp3lame/psymodel.c, libmp3lame/version.h:
	  pre-echo control for short blocks is dependent on previous frame.
	  maybe we can relax it later, to speed-up things again.

2011-03-25 19:57  robert

	* frontend/parse.c:
	  dropping note about mpglib being GPL, because we use HIP for
	  decoding.

2011-03-24 20:19  robert

	* doc/html/detailed.html, doc/man/lame.1, frontend/parse.c,
	  libmp3lame/id3tag.c:
	  new switches controlling the string encoding mode for ID3v2 tags:
	  --id3v2-latin1 turns on ID3v2 tags, text encoding ISO-8859-1
	  --id3v2-ucs2 turns on ID3v2 only tags, text encoding UCS-2
	  
	  Similar experimental switches removed.

2011-03-24 10:25  robert

	* doc/html/history.html:
	  fixing glitch in Joseph's name

2011-03-24 10:21  robert

	* doc/html/history.html:
	  updating history with info about latest changes

2011-03-22 20:35  robert

	* ChangeLog, libmp3lame/version.h:
	  updating ChangeLog

2011-03-05 20:05  robert

	* libmp3lame/presets.c, libmp3lame/quantize_pvt.c:
	  adapting some parameters for vbr-old/cbr/abr, because of the now
	  used psy model

2011-03-05 18:54  robert

	* libmp3lame/encoder.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize_pvt.c:
	  now all modes use the same (latest) psy model
	  
	  code cleanup

2011-03-05 14:06  robert

	* mpglib/layer3.c:
	  fix for a possible out of bounds array access
	  
	  re ID: 3091776 crash at III_dequantize_sample
	  This bug may be fixed now, don't have original problem sample to
	  be sure.

2011-03-02 16:56  rbrito

	* debian/changelog:
	  debian: Update changelog.

2011-03-02 16:56  rbrito

	* debian/changelog:
	  Put my SF email or we get warnings from lintian.

2011-03-02 16:54  rbrito

	* debian/control:
	  Verified that the packaging conforms to the Policy version 3.9.1

2011-03-02 16:54  rbrito

	* debian/control:
	  Verified that the packaging conforms to the Policy version 3.9.0

2011-03-02 16:54  rbrito

	* debian/control:
	  Remove unused Conflicts: and Replaces: fields.
	  
	  They were never part of any stable release and 3rd party sites
	  were,
	  perhaps, faster than us in providing changed packages (when the
	  packaging
	  was not up-to-date).

2011-03-02 16:53  rbrito

	* debian/control:
	  Verified that the packaging conforms to the Policy version 3.8.4

2011-03-02 16:53  rbrito

	* debian/control:
	  Verified that the packaging conforms to the Policy version 3.8.3

2011-03-02 16:52  rbrito

	* debian/control:
	  Verified that the packaging conforms to the Policy version 3.8.2

2011-03-02 16:52  rbrito

	* debian/control:
	  Verified that the packaging conforms to the Policy version
	  3.8.1.0

2011-03-02 16:52  rbrito

	* debian/control:
	  Fix lintian's debhelper-but-no-misc-depends from the binary
	  packages.

2011-03-02 16:51  rbrito

	* debian/changelog:
	  Update debian/changelog.

2011-03-02 16:50  rbrito

	* debian/control:
	  debian/control: Change B-D on nasm to allow any i386 platform (no
	  amd64, though).

2011-03-02 16:50  rbrito

	* debian/control:
	  debian/control: Rewrap the Build-Depends field for legibility.
	  
	  This also buys us some diff-friendliness.

2011-03-02 16:48  rbrito

	* debian/changelog, debian/rules:
	  debian: Transition to the debhelper 7 format.
	  
	  This format is easier to manage and other people should find it
	  easy to
	  adapt to downstream necessities.

2011-03-01 00:34  robert

	* libmp3lame/lame.c, libmp3lame/vbrquantize.c, testcase.mp3:
	  - new VBR: mpeg2.5 at 8 kHz doesn't use some scalefactor bands

2011-02-27 21:08  robert

	* libmp3lame/lame.c:
	  small update for q_map

2011-02-27 01:52  robert

	* libmp3lame/lame.c, libmp3lame/presets.c:
	  last patch made manual lowpass setting impossable (for VBR)

2011-02-27 00:52  robert

	* Makefile.unix, libmp3lame/lame.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/version.h:
	  make use of resampling at lower VBR quality levels

2011-02-22 19:00  robert

	* libmp3lame/bitstream.c, libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/version.h:
	  Starting to use newer VBR psy-model for CBR and ABR encoding too
	  
	  bug tracker ID: 3187397 - Strange compression behavior

2011-02-22 17:58  robert

	* libmp3lame/bitstream.c, libmp3lame/lame.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/version.c:
	  some code cleanup

2011-02-13 22:54  rbrito

	* frontend/brhist.c:
	  frontend: remove unused variable and computation from brhist.c

2011-02-13 16:36  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/version.h:
	  make vbr_mt the default vbr mode

2011-02-13 13:50  robert

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/presets.c, libmp3lame/set_get.c:
	  new mono encoding switches:
	  -ml encodes left channel only
	  -mr encodes right channel only

2011-02-12 16:16  robert

	* frontend/main.h, libmp3lame/presets.c, libmp3lame/vbrquantize.c:
	  small fixes for compiler warnings
	  (vbrquantize.c) bug fix, out of bounds memory write

2011-02-11 20:17  robert

	* libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/vbrquantize.c,
	  libmp3lame/version.h:
	  - reintroduction of fast encoding mode (-q7 / -f) (vbr_mt /
	  vbr_mtrh)
	  - minor fix for sfb21 encoding (vbr_mt / vbr_mrth)
	  - small tunings

2010-12-08 15:28  robert

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  fixing some compile issues when compiling as C++ code
	  
	  bug tracker ID: 3125235: make -f Makefile.mingw32 fails

2010-12-05 12:25  robert

	* Dll/BladeMP3EncDLL.c:
	  fixing some compile issues when compiling as C++ code
	  
	  bug tracker ID: 3125235: make -f Makefile.mingw32 fails

2010-12-04 20:43  jaz001

	* doc/html/detailed.html:
	  spelling and better description on -b, -F and -t settings

2010-12-04 13:30  robert

	* libmp3lame/gain_analysis.h:
	  fixing some compile issues with strict compiler settings.
	  
	  bug tracker ID: 3025801: gain_analysis.h needs int constants

2010-12-04 12:55  robert

	* Dll/BladeMP3EncDLL.c:
	  fixing some compile issues with strict compiler settings.
	  
	  bug tracker ID: 3125235: make -f Makefile.mingw32 fails

2010-12-03 19:46  jaz001

	* doc/html/about.html:
	  fixed listening tests URL.

2010-11-21 20:45  jaz001

	* doc/html/detailed.html, doc/html/usage.html:
	  yet more small detail changes.

2010-11-21 20:31  jaz001

	* doc/html/detailed.html:
	  more details about the q switch and the vbr-new/old
	  some formating changes

2010-11-21 20:07  jaz001

	* doc/html/abr.html, doc/html/cbr.html, doc/html/detailed.html,
	  doc/html/usage.html:
	  updates/fixes to the documentation.

2010-10-30 13:21  robert

	* Dll/LameDll_vc8.vcproj, Dll/Makefile.am, Dll/Makefile.in,
	  Makefile.am, Makefile.in, frontend/Makefile.am,
	  frontend/Makefile.in, frontend/lame_vc8.vcproj,
	  frontend/mp3x_vc8.vcproj, lame_vc8.sln, libmp3lame/Makefile.am,
	  libmp3lame/Makefile.in, libmp3lame/libmp3lame_vc8.vcproj,
	  mpglib/Makefile.am, mpglib/Makefile.in, mpglib/mpglib_vc8.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj:
	  VC8 no more

2010-09-30 20:58  jaz001

	* doc/html/Makefile.am, doc/html/Makefile.in:
	  fixed the erroneous naming of files. (fixes: 3078255)

2010-08-17 23:39  rbrito

	* README:
	  Put a notice that Mark Taylor is not the current maintainer, but
	  The
	  LAME team is.

2010-08-17 23:36  rbrito

	* README:
	  Update the URL to point to the Sourceforge webpage.

2010-08-17 23:35  rbrito

	* README:
	  Update the date of the README file.

2010-08-17 23:34  rbrito

	* README:
	  Cosmetic: delete trailing whitespaces.

2010-08-17 23:32  rbrito

	* README:
	  Remove duplicate copyright entry for Mark Taylor.

2010-08-17 23:32  rbrito

	* README:
	  Update copyright years of the LAME Project.

2010-07-09 18:59  jaz001

	* doc/html/basic.html, doc/html/examples.html, doc/html/id3.html,
	  doc/html/lame.css, doc/html/modes.html, doc/html/node6.html,
	  doc/html/switchs.html:
	  delete old documentation

2010-07-09 18:58  jaz001

	* doc/html/detailed.html:
	  added anchors to the descriptions, so that one can link to them
	  with an URL.
	  Corrected a few errors.

2010-06-11 17:44  jaz001

	* doc/html/Makefile.am, doc/html/about.html, doc/html/abr.html,
	  doc/html/cbr.html, doc/html/contact.html,
	  doc/html/contributors.html, doc/html/detailed.html,
	  doc/html/images, doc/html/images/logo.gif,
	  doc/html/images/valid-css.png, doc/html/images/valid-xhtml10.png,
	  doc/html/index.html, doc/html/introduction.html,
	  doc/html/links.html, doc/html/list.html, doc/html/ms_stereo.html,
	  doc/html/styles, doc/html/styles/lame.css, doc/html/usage.html,
	  doc/html/vbr.html:
	  Initial commit of the new documentation. Makefile.in should be
	  regenerated.

2010-06-07 00:18  robert

	* libmp3lame/presets.c, libmp3lame/psymodel.c:
	  something to try later

2010-06-06 19:20  robert

	* frontend/rtp.c:
	  fix compile problem on Linux system

2010-06-06 01:04  robert

	* libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/set_get.c,
	  libmp3lame/util.h, libmp3lame/version.h:
	  some further tweaking of vbr new

2010-06-01 21:55  robert

	* libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h, libmp3lame/version.h:
	  more tuning on vbr_mt

2010-06-01 18:51  robert

	* libmp3lame/bitstream.c, libmp3lame/lame.c:
	  changing default buffer constraint for vbr_mt

2010-06-01 16:21  robert

	* libmp3lame/lame.c:
	  fixing some compiler warnings

2010-06-01 15:59  robert

	* frontend/get_audio.c, frontend/get_audio.h, frontend/lametime.c,
	  frontend/lametime.h, libmp3lame/lame.c, libmp3lame/presets.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.c, libmp3lame/util.h, libmp3lame/version.h:
	  some work on VBR scale tuning, accessible with --vbr-new

2010-05-02 20:29  robert

	* frontend/get_audio.c, frontend/get_audio.h, frontend/gtkanal.c,
	  libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/util.c,
	  libmp3lame/version.h:
	  work-in-progress: VBR scale tuning

2010-04-29 00:29  robert

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h, libmp3lame/version.h:
	  some work on VBR scale tuning, accessible with --vbr-new

2010-04-15 10:44  robert

	* libmp3lame/VbrTag.c, libmp3lame/lame.c, libmp3lame/util.h:
	  - calculate input transform matrix once on initialization
	  - calculate clip detect even when scaling was set; else one
	  doesn't notice that clipping still might happen.

2010-04-14 09:38  robert

	* libmp3lame/i386/Makefile.in:
	  Bug fix for tracker item ID: 2986823: Error when building Lame
	  with NASM support -still exists

2010-04-11 15:46  robert

	* Makefile.MSVC, libmp3lame/lame.rc, libmp3lame/set_get.h:
	  adding version/icon resource to VC6 binary builds

2010-04-11 13:07  robert

	* include/lame.def, include/lame.h, libmp3lame/lame.c:
	  add encoding functions passing double precision pcm values ( +/-
	  1 fulll scale)

2010-04-09 21:51  robert

	* doc/html/switchs.html, doc/man/lame.1, frontend/get_audio.c,
	  frontend/main.h, frontend/parse.c:
	  - unicode filenames now working with libsndfile on MS-Windows
	  - presets use the default VBR engine, no need for extra parameter
	  "fast"

2010-04-08 11:07  robert

	* frontend/get_audio.c, frontend/get_audio.h, frontend/lame_main.c,
	  frontend/mp3rtp.c, frontend/mp3x.c, frontend/parse.c,
	  libmp3lame/id3tag.c, libmp3lame/version.h:
	  enabling to read OGG Vorbis files via libsndfile

2010-04-06 11:13  robert

	* frontend/get_audio.h, frontend/lame_main.c, frontend/parse.c,
	  include/lame.def, include/lame.h, libmp3lame/id3tag.c,
	  libmp3lame/lame.c:
	  - preparing to use ieee float [-1,+1] as internal pcm sample
	  representation in LAME frontend
	  - removing id3v2 picture size limit
	  - fixing id3v2 TXXX frame storage

2010-03-23 11:22  robert

	* include/lame.h:
	  small typo

2010-03-22 23:12  robert

	* libmp3lame/quantize.c, libmp3lame/version.h:
	  vbrnew: pad frame to build up bitrservoir

2010-03-22 20:28  robert

	* doc/html/history.html:
	  updating latest changes

2010-03-22 20:20  robert

	* doc/html/history.html:
	  updating latest changes

2010-03-22 19:44  robert

	* vc_solution/vc9_mpglib.vcproj:
	  adding header file to project

2010-03-22 14:44  robert

	* mpglib/AUTHORS, mpglib/README:
	  update licence information/header in decoder files

2010-03-22 14:32  robert

	* LICENSE, mpglib/Makefile.am, mpglib/Makefile.in, mpglib/common.c,
	  mpglib/common.h, mpglib/dct64_i386.c, mpglib/dct64_i386.h,
	  mpglib/decode_i386.c, mpglib/decode_i386.h, mpglib/huffman.h,
	  mpglib/interface.c, mpglib/interface.h, mpglib/l2tables.h,
	  mpglib/layer1.c, mpglib/layer1.h, mpglib/layer2.c,
	  mpglib/layer2.h, mpglib/layer3.c, mpglib/layer3.h,
	  mpglib/mpg123.h, mpglib/mpglib.h, mpglib/tabinit.c,
	  mpglib/tabinit.h:
	  update licence information/header in decoder files

2010-03-21 22:22  robert

	* frontend/parse.c, libmp3lame/set_get.c, mpglib/common.c,
	  mpglib/layer3.c:
	  minor changes

2010-03-21 22:18  robert

	* libmp3lame/reservoir.c:
	  Bug fix for tracker item ID: 2973877: A problem regarding the new
	  drain code

2010-03-15 01:32  robert

	* frontend/gtkanal.c, vc_solution/vc9_lame.sln,
	  vc_solution/vc9_lame_mp3x.vcproj:
	  it makes no sense, to link mp3x against libmp3lame.dll, because
	  of strong dependencies between them. (pinfo struct)

2010-03-15 01:07  robert

	* frontend/parse.c, include/lame.def, include/lame.h,
	  libmp3lame/VbrTag.c, libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/tables.c, libmp3lame/tables.h, libmp3lame/util.c:
	  access functions for bitrate and samplerate tables. exporting of
	  global vars doesn't work well with DLLs.

2010-03-14 22:47  robert

	* frontend/get_audio.c, frontend/lame_main.c, frontend/main.c,
	  frontend/main.h, frontend/parse.c, frontend/timestatus.c,
	  libmp3lame/set_get.h, vc_solution/vc9_lame.sln:
	  some code refactorings

2010-03-14 19:48  robert

	* frontend/timestatus.c, libmp3lame/lame.rc,
	  vc_solution/vc9_lame_mp3x.vcproj:
	  adding resource file containing version and program icon to MP3x
	  project

2010-03-14 17:23  robert

	* Makefile.MSVC, Makefile.unix, configure.in, frontend/Makefile.am,
	  frontend/Makefile.in, frontend/brhist.c, frontend/lame_main.c,
	  frontend/timestatus.c, include/Makefile.am,
	  include/libmp3lame.sym, vc_solution/Makefile.am:
	  BRHIST is mandatory
	  Layer1 and Layer2 decoding is mandatory
	  adding libmp3lame.sym file, to use with libtool, declares
	  exported symbols

2010-03-14 16:15  robert

	* frontend/mp3rtp.c, libmp3lame/set_get.c:
	  fixing some compiler warnings

2010-03-14 15:39  robert

	* frontend/main.c, frontend/timestatus.c, libmp3lame/encoder.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/set_get.c, libmp3lame/util.h, testcase.mp3:
	  trying to get rid of remaining GCC compiler warnings

2010-03-13 21:10  robert

	* Makefile.MSVC:
	  additional target mp3rtp

2010-03-13 20:45  robert

	* Makefile.MSVC, Makefile.unix, frontend/Makefile.am,
	  frontend/Makefile.in, frontend/lame_main.c,
	  frontend/lame_vc8.vcproj, frontend/main.c, frontend/mp3rtp.c,
	  frontend/mp3x.c, frontend/mp3x_vc8.vcproj,
	  vc_solution/vc9_lame.sln, vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3rtp.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj:
	  main.c is now entry point for all frontend programs
	  similar to mp3x and mp3rtp, there is now lame_main.c for lame

2010-03-13 01:49  robert

	* doc/html/history.html:
	  support for WAVE_FORMAT_IEEE_FLOAT

2010-03-13 01:17  robert

	* frontend/get_audio.c:
	  support for WAVE_FORMAT_IEEE_FLOAT (for reading via libsndfile)

2010-03-12 14:00  robert

	* frontend/get_audio.c:
	  support for WAVE_FORMAT_IEEE_FLOAT (for reading)

2010-03-11 00:59  robert

	* frontend/console.c, frontend/get_audio.c, frontend/main.c,
	  frontend/main.h, frontend/mp3rtp.c, frontend/parse.c,
	  frontend/rtp.c, frontend/timestatus.c, include/lame.def,
	  include/lame.h, libmp3lame/id3tag.h:
	  experimental unicode support on Windows OS

2010-03-10 01:41  robert

	* frontend/main.c, frontend/timestatus.c:
	  Starting with Unicode support for LAME on Windows

2010-03-10 00:43  robert

	* frontend/timestatus.c:
	  small fix for time status display during encoding

2010-03-10 00:35  robert

	* frontend/get_audio.c, frontend/main.c, frontend/timestatus.c,
	  frontend/timestatus.h:
	  display of estimated number of frames should be more accurate
	  now, even when re-encoding mp3->mp3

2010-03-09 00:03  robert

	* libmp3lame/set_get.c:
	  calculation of totalframes more in line with current
	  enc-delay+padding

2010-03-08 00:17  robert

	* frontend/get_audio.c, frontend/main.c, frontend/timestatus.c,
	  libmp3lame/bitstream.c, libmp3lame/util.h:
	  - reducing some warnings
	  - lame_get_totalframes still returns wrong estimate when
	  re-encoding

2010-03-07 21:14  robert

	* frontend/get_audio.c, frontend/get_audio.h, frontend/main.c,
	  frontend/mp3rtp.c, frontend/mp3x.c, frontend/timestatus.c:
	  some more code refactoring to clean up encoder/decoder loops

2010-03-06 21:34  robert

	* doc/html/history.html, frontend/get_audio.c, frontend/main.c:
	  Bug fix for tracker item ID: 2962223: Encoder delay and padding
	  values ignored when transcoding

2010-03-06 02:15  robert

	* frontend/get_audio.c, frontend/get_audio.h, frontend/main.c,
	  frontend/parse.c:
	  some code refactorings

2010-03-05 02:01  robert

	* doc/html/history.html:
	  updating history text

2010-03-04 21:37  robert

	* frontend/rtp.c:
	  cleanup some pre-processor code

2010-03-04 20:03  robert

	* frontend/rtp.c:
	  more work on mp3rtp

2010-03-04 19:42  robert

	* frontend/mp3rtp.c, frontend/rtp.c, frontend/rtp.h:
	  more work on mp3rtp
	  should compile with VC9 and GCC now

2010-03-04 16:38  robert

	* frontend/mp3rtp.c, frontend/rtp.c, frontend/rtp.h,
	  include/lame.def, vc_solution/vc9_lame.sln,
	  vc_solution/vc9_lame_mp3rtp.vcproj:
	  mp3rtp compiles with VC9 now

2010-03-04 09:32  robert

	* vc_solution/vc9_lame_mp3rtp.vcproj:
	  adding mp3rtp project to VC solution

2010-03-04 09:23  robert

	* vc_solution/vc9_lame.sln:
	  adding mp3rtp project to VC solution

2010-03-03 00:32  robert

	* mpglib/layer3.c, mpglib/mpglib.h:
	  Eliminated some remaining global variable in layer3 decoder code.

2010-03-02 01:21  robert

	* frontend/brhist.h, frontend/console.h, frontend/get_audio.c,
	  frontend/gtkanal.c, frontend/lametime.h, frontend/main.c,
	  frontend/main.h, frontend/mp3rtp.c, frontend/parse.c,
	  frontend/parse.h, frontend/portableio.h, frontend/timestatus.c,
	  frontend/timestatus.h:
	  some code refactoring

2010-02-28 17:22  robert

	* ChangeLog:
	  updating Changelog

2010-02-28 16:52  robert

	* libmp3lame/bitstream.c:
	  For free format mp3 above 320 kbps, the maximum allowed bits was
	  limited too far.

2010-02-27 03:18  robert

	* doc/html/history.html:
	  updating history text

2010-02-26 00:55  robert

	* libmp3lame/i386/Makefile.am:
	  Tracker item ID: 2807676:
	  Error when building Lame with NASM support
	  
	  Patch reported by Bernhard Doebler.

2010-02-20 21:19  robert

	* test/lame_test.cpp, vc_solution/vc9_lame_clients.sln,
	  vc_solution/vc9_lame_test.vcproj:
	  adding some small LAME test project.

2010-02-20 21:08  robert

	* libmp3lame/lame.c:
	  Bug fix for tracker ID: 2824296
	  wrong enc_padding value in LAME 3.99a & 3.98.3 (from CVS)

2010-02-20 21:01  robert

	* libmp3lame/VbrTag.c:
	  Bug fix for tracker ID: 2872590
	  LameTAG: "Music length" missmatch in Lame3.98

2010-02-19 00:46  robert

	* libmp3lame/id3tag.c:
	  Bug fix for tracker ID: 2887359
	  Wrong length in ID3v2 tag when num_samples isn't set

2010-02-19 00:30  robert

	* Dll/BladeMP3EncDLL.c:
	  Bug fix for tracker ID: 2893101
	  Access Violation in BladeMP3EncDLL if UNICODE was defined.

2010-02-19 00:05  robert

	* doc/man/lame.1:
	  Bug fix for tracker ID: 2928684
	  Error in man page about "fast insane"

2010-02-13 23:21  robert

	* frontend/brhist.c, frontend/console.h, frontend/get_audio.c,
	  frontend/parse.c, frontend/timestatus.c, include/lame.h,
	  libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/lame.c, libmp3lame/reservoir.c, libmp3lame/version.h:
	  Revisiting the FhG decoder problem (FhG V1.5 build 50, ships with
	  MS Windows):
	  enabling the new drain code seems to solve that issue better,
	  than restricting the buffer size.

2010-02-06 23:01  robert

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/reservoir.c, libmp3lame/set_get.c, libmp3lame/util.h:
	  added new switch for control buffer constraint

2009-12-12 15:28  robert

	* frontend/console.c:
	  increasing size of term_buff array, it might have become too
	  small by now

2009-12-11 23:19  robert

	* libmp3lame/util.c:
	  small issue,: if-def out hip functions, when not using
	  decode-on-the-fly

2009-12-11 22:30  robert

	* frontend/parse.c:
	  Tracker ID: 2891879
	  Because of Windows API change, there was a problem with prority
	  switch. Should work again.

2009-11-22 22:29  rbrito

	* debian/rules:
	  Enable the frontend to be dynamically linked to libmp3lame.

2009-11-22 22:09  rbrito

	* debian/lame.doc-base:
	  Fix correct section of doc-base.

2009-11-22 21:27  rbrito

	* debian/changelog:
	  Consolidate changelog entries.

2009-11-22 21:18  rbrito

	* debian/source, debian/source/format:
	  Part of migration to new format (3.0 quilt) of Debian packages.

2009-09-17 06:42  rbrito

	* debian/lame.doc-base:
	  Update the section field.

2009-09-17 05:54  rbrito

	* debian/lame.doc-base:
	  Include Debian's doc-base configuration.

2009-05-16 10:20  rbrito

	* debian/rules:
	  Revert cleaning of files in the debian/rules file.

2009-05-16 10:19  rbrito

	* debian/Makefile.am, debian/rules:
	  Update the list of files in Makefile.am

2009-05-15 23:09  robert

	* mpglib/interface.c, mpglib/layer1.c, mpglib/layer1.h,
	  mpglib/layer2.c, mpglib/layer2.h, mpglib/layer3.c,
	  mpglib/layer3.h:
	  trying to handle the case, when audio data exceeds frame size

2009-05-15 23:05  robert

	* libmp3lame/bitstream.c, libmp3lame/lame.c:
	  starting to demangle bitstream buffer handling from
	  postprocessing features like "clip detect" and "replaygain".

2009-05-10 22:08  jwflynn1

	* dshow/Mpegac.h:
	  minor correction

2009-05-10 22:06  jwflynn1

	* dshow/Mpegac.cpp:
	  Allocator buffer count on the output pin had been hardcoded to 1.
	  This was changed so that it will now attempt to use the buffer
	  count suggested by the downstream filter or default to 1 if none
	  is specified.

2009-05-02 13:45  rbrito

	* debian/lame.files, debian/libmp3lame0-dev.docs,
	  debian/libmp3lame0-dev.files, debian/libmp3lame0.files:
	  Remove obsolete files. We need to fix the auto* files to use the
	  new files.

2009-05-02 13:44  rbrito

	* debian/changelog, debian/compat, debian/control,
	  debian/lame.install, debian/libmp3lame-dev.install,
	  debian/libmp3lame0.install, debian/rules:
	  Update Debian/Ubuntu packaging. Add new files.

2009-04-20 21:48  robert

	* frontend/console.c, frontend/get_audio.c, frontend/main.c,
	  frontend/parse.c, frontend/portableio.c, libmp3lame/fft.c,
	  mpglib/common.c, mpglib/common.h, mpglib/layer1.c,
	  mpglib/layer2.c, vc_solution/vc9_lame_config.vsprops:
	  fixing some more compiler warnings

2009-04-20 05:39  rbrito

	* debian/changelog, debian/control, debian/copyright, debian/rules,
	  debian/watch:
	  Updating the Debian packaging and doing some investigation
	  regarding symbols.

2009-04-19 22:18  robert

	* libmp3lame/id3tag.c, libmp3lame/takehiro.c, mpglib/layer1.c:
	  reducing compiler warnings

2009-04-19 18:10  robert

	* mpglib/interface.c, mpglib/layer2.c:
	  Starting to fix decoding bugs.

2009-04-18 18:33  robert

	* libmp3lame/lame.rc, mpglib/interface.c, mpglib/layer1.c,
	  mpglib/layer2.c, mpglib/mpg123.h:
	  Starting to fix decoding bugs.

2009-04-17 22:21  robert

	* frontend/console.c, frontend/console.h, frontend/get_audio.c,
	  frontend/main.c, frontend/parse.c, frontend/timestatus.c,
	  libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h,
	  mpglib/interface.c, mpglib/mpglib.h:
	  commandline option 'quiet' should suppress any message now

2009-04-17 11:24  robert

	* frontend/get_audio.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/util.c,
	  libmp3lame/util.h, mpglib/common.c, mpglib/common.h,
	  mpglib/interface.c, mpglib/layer3.c, mpglib/mpglib.h:
	  replaced fprintf calls by lame reporting function calls. this
	  allows to silence decoder. (it used to spit out error messages to
	  stderr)

2009-04-11 16:04  jwflynn1

	* dshow/README:
	  *** empty log message ***

2009-04-11 15:59  jwflynn1

	* dshow/Mpegac.cpp:
	  - Added support for the DirectShow IAMStreamConfig Interface to
	  the LAME encoder filter output pin.
	  
	  - Modified the DirectShow filter registration section so that the
	  LAME Encoder filter is correctly registered in the Audio
	  Compressors filter category. This will allow third-party encoding
	  applications using the DirectShow System Device Enumerator
	  Interface to correctly detect the LAME encoder when querying the
	  Audio Compressors filter group.
	  
	  - Modified the filter registration information so that the MP3
	  audio subtype is correctly reported as being supported on the
	  encoder output pin. This will allow third-party encoding
	  applications using the DirectShow IFilterMapper2 Interface to
	  recognize that the LAME encoder supports MP3 output.
	  
	  - Altered the Filter Merit Value that was being used when the
	  filter was registered so that it is now using the standard
	  DirectShow compressor filter merit value of MERIT_DO_NOT_USE
	  (0x200000). Previously, the filter was being registered using a
	  value of MERIT_SW_COMPRESSOR (0x100000), which was at a lower
	  priority (i.e. worse priority) than MERIT_DO_NOT_USE. This
	  prevented the LAME Encoder filter from being selected for use by
	  some third-party encoding applications.
	  
	  - Added code to calculate the frame length of the audio frames
	  used for the nBlockSize element of the WAVEFORMATEX output
	  structure. Previously this value was simply hard-coded to 1.

2009-04-11 15:51  jwflynn1

	* dshow/Mpegac.h:
	  Added support for the DirectShow IAMStreamConfig Interface

2009-04-11 15:47  jwflynn1

	* dshow/Property.rc:
	  Updated the embedded "FileVersion" and "ProductVersion" fields

2009-03-31 22:38  robert

	* libmp3lame/util.c, libmp3lame/util.h:
	  fix for tracker item 2723518: resampling in 3.98 and 3.99alpha

2009-03-30 14:38  robert

	* libmp3lame/lame.c:
	  ensure we pad with 288 samples at least

2009-03-30 13:42  robert

	* frontend/main.c, frontend/main.h, frontend/parse.c:
	  commandline option "--swap-channel" added, as requested. see
	  item: 1118412

2009-03-29 18:28  robert

	* libmp3lame/psymodel.c:
	  take away sandpaper safety margin for lower samplerates.
	  It seems samplerates which have convolution bands containing
	  single lines are affected only (at least I don't have any
	  testcase for lower samplerates to justify the safety margin right
	  now)

2009-03-29 17:35  robert

	* libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  workaround for FFMPEG bug, which uses to call lame_encode_flush
	  more than once in loop

2009-03-02 23:27  robert

	* libmp3lame/VbrTag.c:
	  GetVbrTag only works for Layer3, so ignore it when Layer1 or 2.

2009-02-14 21:59  robert

	* Makefile.unix:
	  adding CFG=RH_INTEL for Intel C(++) compiler setup

2009-02-13 15:03  robert

	* libmp3lame/fft.c, libmp3lame/vector/xmm_quantize_sub.c,
	  vc_solution/vc9_libmp3lame.vcproj:
	  fix crash in VC9 Release Build

2009-02-13 03:47  robert

	* libmp3lame/fft.c, libmp3lame/quantize.c, libmp3lame/takehiro.c,
	  vc_solution/arch_sse2.vsprops:
	  small changes

2009-02-12 14:32  robert

	* libmp3lame/quantize.c, libmp3lame/takehiro.c,
	  libmp3lame/vector/xmm_quantize_sub.c:
	  some tiny changes

2009-02-12 11:35  robert

	* libmp3lame/fft.c, libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/xmm_quantize_sub.c,
	  vc_solution/arch_sse2.vsprops:
	  first try to add fht using SSE intrinsics

2009-02-11 00:22  robert

	* vc_solution/vc9_lame_config.vsprops:
	  adding configuration for SSE2 target architectures

2009-02-07 11:19  robert

	* Dll/Makefile.mingw32:
	  applying patch from item tracker: [ 2406420 ] compile
	  lame_enc.dll under cygwin

2009-02-07 00:44  robert

	* libmp3lame/id3tag.c, libmp3lame/vector/xmm_quantize_sub.c:
	  small changes

2009-02-06 23:27  robert

	* libmp3lame/id3tag.c:
	  silence some GCC compiler warning

2009-02-06 23:26  robert

	* libmp3lame/VbrTag.c, libmp3lame/lame.c:
	  bug-fix: low bitrate CBR incorrectly tried to write LAME-Tag; var
	  inconsistency

2009-02-06 23:22  robert

	* libmp3lame/vector/xmm_quantize_sub.c:
	  - fix compiler warning about strict-aliasing problem
	  - fix possible use of uninitialized vars

2009-02-06 23:20  robert

	* Makefile.unix:
	  adding work-configuration for SSE2 optimization under SuSE 32bit

2009-02-06 00:21  robert

	* vc_solution/arch_sse2.vsprops, vc_solution/arch_x87.vsprops,
	  vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_mpglib.vcproj:
	  adding configuration for SSE2 target architectures

2009-02-06 00:21  robert

	* configMS.h, libmp3lame/vbrquantize.c,
	  vc_solution/arch_nasm.vsprops, vc_solution/vc9_lame.sln,
	  vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_mpglib.vcproj:
	  adding configuration for SSE2 target architectures

2009-01-18 15:57  robert

	* frontend/get_audio.c, frontend/gtkanal.c, frontend/lametime.c,
	  frontend/main.c, libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/vector/xmm_quantize_sub.c, mpglib/interface.c,
	  mpglib/mpglib.h:
	  replacing C++ style comments by C style comments

2009-01-18 13:59  robert

	* doc/html/switchs.html, doc/man/lame.1:
	  fix for bug tracker item [ 2051870 ] documentation clarifications

2009-01-18 13:10  robert

	* doc/man/lame.1:
	  fix for bug tracker item 2423650

2008-12-26 11:58  robert

	* libmp3lame/presets.c, libmp3lame/psymodel.c:
	  disabling testcode

2008-12-02 03:23  robert

	* libmp3lame/presets.c, libmp3lame/psymodel.c:
	  testing some low quality tuning (VBR-new only)

2008-11-30 00:20  robert

	* Makefile.unix, libmp3lame/presets.c, libmp3lame/psymodel.c:
	  testing some different short block switching and change in
	  masking tonal
	  samples (VBR-NEW only)

2008-11-09 14:16  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  debian/Makefile.in, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in, vc_solution/Makefile.in:
	  regen

2008-11-09 14:13  aleidinger

	* aclocal.m4, config.h.in, configure:
	  regen (with autoconf 2.62)

2008-11-09 14:10  aleidinger

	* libmp3lame/Makefile.am:
	  Add missing files, remove deleted ones.

2008-11-09 14:06  aleidinger

	* vc_solution/Makefile.am:
	  Add missing automake Makefile.

2008-11-09 14:05  aleidinger

	* configure.in:
	  Add possibility to build frontends with dynamic libmp3lame.

2008-11-09 14:00  aleidinger

	* frontend/Makefile.am:
	  Remove -static in preparation for corresponding configure option.

2008-11-09 13:56  aleidinger

	* ltmain.sh:
	  Update to version from libtool 1.5.26.

2008-11-09 13:50  aleidinger

	* ACM/ADbg/Makefile.am, ACM/Makefile.am, ACM/tinyxml/Makefile.am,
	  Dll/Makefile.am, Makefile.am, debian/Makefile.am,
	  dshow/Makefile.am, frontend/Makefile.am, include/Makefile.am,
	  libmp3lame/i386/Makefile.am, mpglib/Makefile.am:
	  Add files which are new, and remove deleted ones.

2008-11-06 07:10  rbrito

	* doc/html/history.html:
	  Update history file with my changes.

2008-11-06 06:56  rbrito

	* debian/rules:
	  Actually remove the rpath issue part.

2008-11-06 06:55  rbrito

	* debian/changelog, debian/control:
	  Include suggestions of Fabian on the packaging.
	  Clean up the packaging a bit.
	  Update the mainline version package.

2008-11-02 15:26  robert

	* libmp3lame/quantize_pvt.c, libmp3lame/version.h:
	  start of 3.99 development cycle:
	  - demangling "calc_xmin"

2008-11-01 21:39  robert

	* ACM/ADbg/ADbg_vc7.vcproj, ACM/lameACM_vc7.vcproj,
	  ACM/tinyxml/tinyxml_vc7.vcproj, dshow/Mpegac.cpp, dshow/REG.CPP,
	  dshow/REG.H, dshow/aboutprp.cpp:
	  - Name clash with our CRegKey class and ATL::CRegKey class solved
	  by putting ours in namespace Lame.

2008-11-01 21:21  robert

	* vc_solution/vc9_lame_acm.vcproj,
	  vc_solution/vc9_lame_acm_adbg.vcproj,
	  vc_solution/vc9_lame_acm_tinyxml.vcproj,
	  vc_solution/vc9_lame_dll.vcproj,
	  vc_solution/vc9_lame_dll_example.vcproj,
	  vc_solution/vc9_lame_dshow.vcproj,
	  vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_mpglib.vcproj:
	  moving common VS projects settings into own property page

2008-10-23 23:26  robert

	* mpglib/common.c, mpglib/common.h, mpglib/dct64_i386.c,
	  mpglib/dct64_i386.h, mpglib/decode_i386.c, mpglib/decode_i386.h,
	  mpglib/huffman.h, mpglib/interface.c, mpglib/interface.h,
	  mpglib/l2tables.h, mpglib/layer1.c, mpglib/layer1.h,
	  mpglib/layer2.c, mpglib/layer2.h, mpglib/layer3.c,
	  mpglib/layer3.h, mpglib/mpg123.h, mpglib/mpglib.h,
	  mpglib/tabinit.c, mpglib/tabinit.h:
	  fixing indentation

2008-10-23 23:20  robert

	* mpglib/huffman.h, mpglib/interface.c, mpglib/interface.h,
	  mpglib/layer1.h, mpglib/layer2.c, mpglib/layer2.h,
	  mpglib/layer3.c, mpglib/layer3.h, mpglib/mpg123.h,
	  mpglib/tabinit.c, mpglib/tabinit.h:
	  removing some tabs
	  preparing for indent

2008-10-23 23:17  robert

	* frontend/brhist.h, frontend/get_audio.c, frontend/get_audio.h,
	  frontend/main.c, frontend/main.h, frontend/portableio.c,
	  frontend/portableio.h, mpglib/common.c, mpglib/common.h,
	  mpglib/dct64_i386.c, mpglib/dct64_i386.h, mpglib/decode_i386.c,
	  mpglib/decode_i386.h, mpglib/huffman.h:
	  removing some tabs
	  preparing for indent

2008-10-21 23:09  robert

	* mpglib/TODO, mpglib/interface.c, mpglib/layer3.c,
	  mpglib/mpg123.h, mpglib/mpglib.h:
	  re-applying some of our changes to decoding lib

2008-10-21 22:35  robert

	* mpglib/README, mpglib/common.c, mpglib/common.h,
	  mpglib/dct64_i386.c, mpglib/dct64_i386.h, mpglib/decode_i386.c,
	  mpglib/decode_i386.h, mpglib/huffman.h, mpglib/interface.c,
	  mpglib/interface.h, mpglib/l2tables.h, mpglib/layer1.c,
	  mpglib/layer1.h, mpglib/layer2.c, mpglib/layer2.h,
	  mpglib/layer3.c, mpglib/layer3.h, mpglib/mpg123.h,
	  mpglib/mpglib.h, mpglib/tabinit.c, mpglib/tabinit.h:
	  replacing mpglib by libhip

2008-10-21 22:34  robert

	* include/lame.h:
	  fix small issues with C99

2008-10-12 19:47  robert

	* libmp3lame/lame.rc, libmp3lame/logoe.ico, libmp3lame/version.c,
	  libmp3lame/version.h, vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj:
	  Adding resources to EXE and DLL

2008-10-11 18:58  robert

	* frontend/gtkanal.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/util.h:
	  fixing mp3x build

2008-10-11 18:04  robert

	* doc/html/history.html, frontend/gtkanal.c, include/lame.def,
	  include/lame.h, libmp3lame/id3tag.c,
	  libmp3lame/mpglib_interface.c, mpglib/layer3.c, mpglib/mpglib.h,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj:
	  Library API change: lame_decode functions are now obsolete but
	  still present, please use hip_decode instead.
	  
	  The reason for this change is: lame_decode functions use a single
	  global variable within the library to store decoder setup,
	  hip_decode functions don't.
	  
	  The encoder now uses hip_decode internally and it is now possible
	  to use clipdetect feature while reencoding mp3 to mp3.

2008-10-07 21:57  robert

	* doc/html/history.html, frontend/parse.c:
	  - let ignore-tag-errors effect albumart option too
	  - updating history

2008-09-21 15:29  robert

	* doc/man/lame.1:
	  fixing Rogerios patch

2008-09-19 07:02  rbrito

	* doc/man/lame.1:
	  Fixing warnings generated by nroff.

2008-09-14 15:21  robert

	* frontend/parse.c:
	  replacing stricmp by local_strcasecmp

2008-09-14 11:55  robert

	* USAGE, doc/html/history.html, doc/html/id3.html, doc/man/lame.1,
	  frontend/parse.c, libmp3lame/id3tag.c:
	  Fixing some ID3 tagging issues:
	  - some sloppier search for genre names
	  - new switch --pad-id3v2-size <n> which allows some user defined
	  padding

2008-09-12 19:41  robert

	* frontend/parse.c, include/lame.def, include/lame.h,
	  libmp3lame/id3tag.c, libmp3lame/id3tag.h:
	  new frontend switch "--pad-id3v2-size <value>":
	  allow id3v2 padding of 0 to 128000 bytes

2008-09-11 00:49  robert

	* frontend/parse.c:
	  replace common suffixes by MP3/WAV when no output filename was
	  given

2008-09-08 22:46  robert

	* frontend/main.c, frontend/parse.c, include/lame.h,
	  libmp3lame/mpglib_interface.c:
	  eleminating some compiler warnings

2008-08-31 19:31  robert

	* include/lame.h, libmp3lame/i386/nasm.h:
	  fixing the fix to get it assembling for ELF again

2008-08-31 17:52  robert

	* include/lame.h:
	  remove some debugging noise

2008-08-31 16:14  robert

	* frontend/get_audio.c, frontend/gtkanal.c, frontend/parse.c,
	  include/lame.def, include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/util.c, libmp3lame/util.h, mpglib/layer3.c,
	  mpglib/mpglib.h:
	  Fix for API problem:
	  lame decoding functionality used one global data structure in
	  libmp3lame for all decoding ativities. This is a big problem with
	  dynamic linking!
	  New API functions added, which use handles to work with.
	  Old API functions are still there, but are deprecated.

2008-08-31 12:28  robert

	* libmp3lame/i386/choose_table.nas, libmp3lame/i386/fft3dn.nas,
	  libmp3lame/i386/fftsse.nas, libmp3lame/i386/nasm.h:
	  fixing none PIC assembling problem

2008-08-27 10:27  rbrito

	* libmp3lame/i386/Makefile.am, libmp3lame/i386/choose_table.nas,
	  libmp3lame/i386/fft3dn.nas, libmp3lame/i386/fftsse.nas,
	  libmp3lame/i386/nasm.h:
	  Committing the changes from Gentoo and PaX Team to the HEAD.

2008-08-16 12:49  robert

	* doc/man/lame.1:
	  adding MPEG-2.5 bitrate table

2008-08-13 21:01  robert

	* libmp3lame/tables.h:
	  revert

2008-08-13 20:38  robert

	* libmp3lame/tables.c, libmp3lame/tables.h:
	  fix compilation via configure/make

2008-08-11 20:56  robert

	* libmp3lame/takehiro.c:
	  loop rewritten; result of discussion with Michael Stoner

2008-08-08 20:34  robert

	* libmp3lame/takehiro.c:
	  tiny speed-up

2008-08-07 20:55  robert

	* Makefile.unix, doc/html/switchs.html, doc/man/lame.1,
	  frontend/parse.c, libmp3lame/lame.c:
	  making some switches developer only switches

2008-08-07 14:53  robert

	* mpglib, mpglib/common.c, mpglib/dct64_i386.h, mpglib/huffman.h,
	  mpglib/interface.c, mpglib/l2tables.h, mpglib/layer1.h,
	  mpglib/mpg123.h, mpglib/mpglib.h, mpglib/tabinit.h:
	  sync with HIP development

2008-08-07 14:52  robert

	* libmp3lame/lame.c:
	  we did init bitstream (Xing header) before final decisions were
	  made

2008-08-06 17:12  robert

	* doc/html/history.html, frontend/parse.c:
	  fix for tracker item [ 2039648 ] potential memory leak in
	  parse_args() function in parse.c

2008-08-06 16:14  robert

	* frontend/parse.c:
	  internal switches are available only when compiled with
	  preprocessor symbol _ALLOW_INTERNAL_OPTIONS
	  defined.

2008-08-06 16:11  robert

	* USAGE, doc/html/switchs.html, doc/man/lame.1, frontend/parse.c:
	  removing some obsolete switches

2008-08-05 17:45  robert

	* doc/html/history.html, frontend/parse.c, libmp3lame/lame.c:
	  fix for bugtracker item [ 2029282 ] Frequency filtering API
	  broken in 3.98

2008-08-05 14:19  robert

	* doc/html/history.html:
	  updating history

2008-08-05 12:33  robert

	* libmp3lame/id3tag.c:
	  fix for bug: LAME writes ID3v2 tag when some ID3v1 field length
	  is too short to hold the text, even in case ID3v1 only is
	  specified.

2008-08-05 12:32  robert

	* libmp3lame/bitstream.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  changes after doing some linting

2008-08-04 19:53  robert

	* libmp3lame/VbrTag.c, libmp3lame/VbrTag.h, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/encoder.h, libmp3lame/fft.c,
	  libmp3lame/fft.h, libmp3lame/gain_analysis.c,
	  libmp3lame/gain_analysis.h, libmp3lame/id3tag.c,
	  libmp3lame/id3tag.h, libmp3lame/lame-analysis.h,
	  libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/mpglib_interface.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/set_get.c, libmp3lame/tables.c, libmp3lame/tables.h,
	  libmp3lame/takehiro.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c, libmp3lame/vbrquantize.h:
	  changes after doing some linting

2008-07-28 17:18  robert

	* libmp3lame/lame.c, misc/abx.c:
	  fixing swapped parameters on calloc call

2008-07-28 16:57  robert

	* libmp3lame/psymodel.c:
	  fixing small typo

2008-07-28 16:38  robert

	* include/lame.h, libmp3lame/VbrTag.c, libmp3lame/VbrTag.h,
	  libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/encoder.c, libmp3lame/encoder.h, libmp3lame/id3tag.c,
	  libmp3lame/id3tag.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/machine.h,
	  libmp3lame/newmdct.c, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/quantize.c,
	  libmp3lame/quantize.h, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/reservoir.c,
	  libmp3lame/reservoir.h, libmp3lame/set_get.c,
	  libmp3lame/takehiro.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  reorganizing the huge lame_internal_flags structure

2008-07-23 14:59  robert

	* libmp3lame/util.c:
	  removed extra line.
	  tracker item: [ 2022101 ] tiny bug(?) in fill_buffer_resample()

2008-07-23 14:46  robert

	* libmp3lame/lame.c:
	  fix for tracker item: [ 2022035 ] encoder_padding value and
	  resampling

2008-07-19 13:34  rbrito

	* HACKING:
	  Fixed HACKING documentation.

2008-07-19 13:09  rbrito

	* HACKING:
	  Fix documentation for set_get/get_set confusion.

2008-07-19 12:54  rbrito

	* doc/html/history.html, frontend/mp3rtp.c, frontend/rtp.c:
	  Fixed compilation of mp3rtp and closed bug on sourceforge
	  tracker.

2008-07-18 22:43  robert

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/gain_analysis.c,
	  libmp3lame/id3tag.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/vbrquantize.c:
	  - removing some unused variables
	  - fix a possible 'divide by zero' in vbrpsy_mask_add

2008-07-12 15:59  rbrito

	* Makefile.unix:
	  Tidying up the Makefile.unix file. Possibly more to come.

2008-07-12 14:32  rbrito

	* debian/changelog:
	  Mention changes to the build system on debian/ubuntu/whatever
	  systems.

2008-07-12 14:28  rbrito

	* debian/rules:
	  Include changes so that the linker only includes libraries
	  directly needed.

2008-07-12 14:26  rbrito

	* Makefile.unix:
	  Include a clue for emacs to know that this is a makefile.

2008-07-08 07:14  rbrito

	* doc/man/lame.1:
	  Small updates to the manpage

2008-07-07 23:50  robert

	* doc/html/history.html, libmp3lame/presets.c:
	  linear interpolate preset param sf21mod

2008-07-07 10:20  rbrito

	* Makefile.unix:
	  Fixed typo in one option.

2008-07-07 09:32  rbrito

	* Makefile.unix:
	  Include flags for compilation for newer machines running Linux.
	  Comment the use of gtk1.2 for Linux at least (don't know about
	  *BSD).

2008-07-07 02:18  rbrito

	* doc/html/history.html:
	  Many fixes to the HTML file (too numerous to mention here). Now,
	  it
	  validates at w3.org (and let us keep it this way, please).

2008-07-07 00:17  rbrito

	* misc/abx.c:
	  Finally fix the endianness issue and be able to close the
	  reported bug.

2008-07-07 00:16  rbrito

	* misc/abx.c:
	  Fix a memory leak (detected by running valgrind).

2008-07-07 00:14  rbrito

	* misc/abx.c:
	  Remove inactive/commented coded that doesn't seem to be
	  necessary.

2008-07-05 11:51  robert

	* lame_clients_vc6.dsw, lame_projects_vc6.dsp, lame_vc6.dsw:
	  we don't support VC6 anymore

2008-07-05 07:50  robert

	* ACM/ADbg/ADbg.dsp, ACM/lameACM_vc6.dsp, ACM/tinyxml/test.dsp,
	  ACM/tinyxml/test.dsw, ACM/tinyxml/tinyxml.dsp,
	  Dll/Example_vc6.dsp, Dll/Example_vc6.dsw, Dll/LameDll_vc6.dsp,
	  dshow/dshow.dsp, dshow/dshow.dsw, frontend/lame_vc6.dsp,
	  frontend/mp3x_vc6.dsp, libmp3lame/libmp3lame_dll_vc6.dsp,
	  libmp3lame/libmp3lame_vc6.dsp, mpglib/mpglib_vc6.dsp:
	  we don't support VC6 anymore

2008-07-05 07:38  robert

	* configure.in, libmp3lame/version.h:
	  Main branch at 3.99 alpha

2008-06-30 23:31  robert

	* configMS.h, dshow/Property.rc, dshow/aboutprp.cpp:
	  updating LAME version property

2008-06-29 21:40  rbrito

	* configure, configure.in, debian/changelog:
	  New updated options for compiling with expopt (to be cleaned up
	  latter).
	  Fixes to the Debian packaging for the pending release.

2008-06-29 03:25  rbrito

	* doc/html/history.html:
	  Fixed my name on an older entry of the history.

2008-06-28 22:36  robert

	* doc/html/Makefile.am, doc/html/Makefile.in:
	  removing presets.html, it is not in repository anymore

2008-06-27 20:47  robert

	* Makefile.unix, testcase.mp3:
	  updating testcase for 3.98 stable release

2008-06-26 21:26  robert

	* vc_solution/vc9_lame.sln, vc_solution/vc9_lame_acm.vcproj,
	  vc_solution/vc9_lame_acm_adbg.vcproj,
	  vc_solution/vc9_lame_acm_tinyxml.vcproj,
	  vc_solution/vc9_lame_dll.vcproj,
	  vc_solution/vc9_lame_dll_example.vcproj,
	  vc_solution/vc9_lame_dshow.vcproj,
	  vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_mpglib.vcproj:
	  removing blanks from configuration and project names (because of
	  NASM rule)

2008-06-26 10:26  bouvigne

	* doc/html/index.html, doc/html/presets.html, doc/man/lame.1,
	  frontend/parse.c:
	  simplified --preset docs

2008-06-26 10:09  bouvigne

	* doc/html/history.html, doc/html/index.html:
	  updated the html docs for 3.98

2008-06-26 08:57  bouvigne

	* libmp3lame/psymodel.c:
	  revert my calc_energy changes

2008-06-25 21:59  robert

	* mpglib/layer2.c, mpglib/layer3.c, mpglib/mpg123.h:
	  cosmetic changes, some more compiler (gcc) warnings eliminated

2008-06-25 12:45  aleidinger

	* ChangeLog:
	  recent changes

2008-06-25 12:43  aleidinger

	* doc/html/history.html:
	  a little bit of history

2008-06-25 12:40  aleidinger

	* configure:
	  regen

2008-06-25 12:37  aleidinger

	* configure.in:
	  fix --diable-cpml switch

2008-06-25 12:17  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  debian/Makefile.in, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  regen (math lib change)

2008-06-25 12:13  aleidinger

	* configure.in:
	  export the math lib, allow to disable the use of the alpha
	  optimized math lib

2008-06-25 12:12  aleidinger

	* libmp3lame/Makefile.am:
	  add the math lib explicitely to the build of the lib

2008-06-25 08:20  robert

	* frontend/parse.c, mpglib/decode_i386.c, mpglib/interface.c,
	  mpglib/interface.h, mpglib/layer1.c, mpglib/layer2.c,
	  mpglib/mpglib.h:
	  Some cosmetic changes, now the vc9 compiles the workspace
	  vc9_lame without any warning

2008-06-24 19:07  rbrito

	* doc/html/history.html:
	  Add some modifications of mine to the history

2008-06-24 18:39  rbrito

	* debian/changelog, debian/control, debian/rules:
	  Updates to the Debian packaging.

2008-06-24 17:06  rbrito

	* misc/abx.c:
	  Quick and dirty fix for some memory leak (detected via Valgrind).

2008-06-22 18:42  robert

	* doc/html/contributors.html, doc/html/history.html:
	  tracker item [ lame-Patches-1990643 ] DShow direct file writer
	  output
	  patch by Anton Sergunov (setosha)

2008-06-22 18:31  robert

	* dshow/Encoder.cpp, dshow/Encoder.h, dshow/Mpegac.cpp,
	  dshow/Mpegac.h:
	  tracker item [ lame-Patches-1990643 ] DShow direct file writer
	  output
	  patch by Anton Sergunov (setosha)

2008-06-22 18:29  robert

	* vc_solution/vc9_lame_acm.vcproj,
	  vc_solution/vc9_lame_acm_adbg.vcproj,
	  vc_solution/vc9_lame_acm_tinyxml.vcproj,
	  vc_solution/vc9_lame_dll_example.vcproj,
	  vc_solution/vc9_lame_dshow.vcproj:
	  linking against multithreaded (debug) DLL

2008-06-17 20:57  robert

	* frontend/parse.c, libmp3lame/presets.c:
	  bug fix: preset INSANE wasn't a CBR preset anymore, but preset
	  help said it should be

2008-06-10 19:24  rbrito

	* misc/abx.c:
	  Fixes for consistency (lack of PATH on some commands)

2008-06-10 19:14  rbrito

	* misc/abx.c:
	  Removing spurious whitespaces at the end of lines.

2008-06-10 16:10  rbrito

	* misc/abx.c:
	  Corrected the abx.c file to work on Linux x86-64 (it should work
	  on ia32 too).
	  
	  I think that I can remove latter the commented lines. They were
	  prejudicial
	  to playing the file on Linux (i.e., a 1min file was played in
	  less than 5s,
	  which is not exactly something that useful for doing an abx
	  test).

2008-05-24 08:40  aleidinger

	* ChangeLog:
	  Add nearly a year of history.

2008-05-24 08:37  aleidinger

	* Makefile.in, configure:
	  regen

2008-05-24 08:31  aleidinger

	* macosx/English.lproj/Makefile.in,
	  macosx/LAME.xcodeproj/Makefile.in, macosx/Makefile.in:
	  regen

2008-05-24 08:28  aleidinger

	* configure.in, macosx/English.lproj/Makefile.am,
	  macosx/LAME.xcodeproj/Makefile.am, macosx/Makefile.am:
	  Add some more macosx stuff

2008-05-22 17:32  audiofileeng

	* macosx/English.lproj, macosx/English.lproj/InfoPlist.strings,
	  macosx/LAME.xcodeproj, macosx/LAME.xcodeproj/project.pbxproj:
	  adding files in subdirectories in macosx

2008-05-22 06:26  aleidinger

	* Makefile.am, configure.in, macosx/Makefile.am:
	  add the new macosx directory

2008-05-21 03:05  audiofileeng

	* macosx, macosx/Info.plist, macosx/LAME_Prefix.pch:
	  added XCode framework project for Mac OS X

2008-05-19 07:41  bouvigne

	* libmp3lame/psymodel.c, libmp3lame/util.h:
	  faster calc_energy: don't loop if there is a single line within
	  the band

2008-05-18 17:49  bouvigne

	* libmp3lame/psymodel.c:
	  faster calc_energy: dont' loop if there is a single line within
	  the band

2008-05-14 07:43  rbrito

	* debian/watch:
	  New file for detecting new versions of the package.

2008-05-14 07:42  rbrito

	* debian/changelog, debian/control, debian/copyright,
	  debian/lame.docs, debian/libmp3lame0-dev.docs,
	  debian/libmp3lame0-dev.files, debian/rules:
	  Updates to the Debian packaging.

2008-04-28 20:32  robert

	* frontend/get_audio.c:
	  fixing LIBSNDFILE compile problem found by Rogério Brito

2008-04-23 01:50  robert

	* frontend/main.c, frontend/parse.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/machine.h,
	  libmp3lame/psymodel.c:
	  fix for bug-tracker item:[ 1948101 ] Memory not freed (lame-3.97,
	  lame-3.98b8)
	  plus, we didn't call lame_close in decode-only case

2008-04-22 23:01  robert

	* libmp3lame/machine.h, libmp3lame/newmdct.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  mpglib/interface.c, mpglib/layer1.c:
	  fixing a problem if you are using --fast-math compiler switch

2008-04-13 22:03  robert

	* doc/html/history.html, libmp3lame/version.h:
	  updating history, we are now at 3.98 beta 8

2008-04-13 21:27  robert

	* frontend/get_audio.c, frontend/main.h, frontend/parse.c,
	  libmp3lame/psymodel.c:
	  found some problem in Naoki's additive masking code and fixed it
	  for VBR_NEW

2008-04-13 13:45  robert

	* frontend/get_audio.c, testcase.mp3:
	  changed compile-time test for machine byte order into run-time
	  test,
	  this should solve issues PPC users seem to have

2008-04-13 02:52  robert

	* libmp3lame/psymodel.c:
	  some SNR testcode to try with -Z3

2008-04-12 18:18  robert

	* frontend/parse.c, frontend/timestatus.c, include/lame.def,
	  include/lame.h, libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/fft.c,
	  libmp3lame/gain_analysis.c, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/presets.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/reservoir.c, libmp3lame/set_get.c,
	  libmp3lame/tables.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c:
	  extended VBR quality scale: 0, ..., 9.999 as floating point value

2008-04-06 15:07  robert

	* doc/html/history.html, frontend/main.c:
	  updating history.html

2008-04-06 02:38  robert

	* libmp3lame/lame_global_flags.h, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/set_get.c,
	  libmp3lame/version.h, mpglib/layer3.c:
	  some simpler spreading function for VBR NEW

2008-04-05 17:38  robert

	* Makefile.unix, frontend/brhist.c, frontend/brhist.h,
	  frontend/main.c, frontend/parse.c, frontend/timestatus.c:
	  removing "#ifdef/#endif" from code bocks, which were always
	  compiled in

2008-04-05 17:13  robert

	* debian/copyright, lame.spec.in, libmp3lame/id3tag.c, misc/abx.c,
	  misc/lameGUI.html:
	  changing some links from "www.mp3dev.org/mp3" to "www.mp3dev.org"
	  and from "www.sulaco.org/mp3" to "www.mp3dev.org"

2008-03-20 22:28  robert

	* include/lame.h:
	  keep 'lame_mp3_tags_fid' alive at least for next stable release

2008-03-12 11:23  robert

	* libmp3lame/psymodel.c, vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_mpglib.vcproj:
	  bug fix: stack corruption in vbrpsy_compute_masking_s

2008-03-11 21:57  robert

	* libmp3lame/set_get.c, libmp3lame/vbrquantize.c:
	  fixing some more compiler warnings

2008-03-11 21:17  robert

	* libmp3lame/VbrTag.c, libmp3lame/fft.c, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/quantize.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c, vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj:
	  fixing some more compiler warnings

2008-03-11 00:56  robert

	* libmp3lame/set_get.c:
	  fixing some more compiler warnings

2008-03-11 00:21  robert

	* vc_solution/vc9_libmp3lame.vcproj:
	  update for nasm file config

2008-03-10 23:47  robert

	* libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/gain_analysis.c, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize.h, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/set_get.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h:
	  fixing some compiler warnings

2008-03-09 22:37  robert

	* frontend/parse.c:
	  for now, keep old ID3 tag switches passing raw character encoded
	  strings
	  some experimental switches for Linux systems:
	  
	  (ID3v1/ID3v2)
	  --lTtitle <some song title to be stored as Latin-1 string>
	  --lArtist <some artist to be stored as Latin-1 string>
	  --lAlbum <some album title to be stored as Latin-1 string>
	  --lGenre <some genre to be stored as Latin-1 string>
	  --lComment <some comment to be stored as Latin-1 string>
	  --lFieldvalue <some fieldvalue to be stored as Latin-1 string>
	  
	  (ID3v2)
	  --uTitle <some song title to be store as UCS-2 string>
	  --uArtist <some artist to be stored as UCS-2 string>
	  --uAlbum <some album title to be stored as UCS-2 string>
	  --uGenre <some genre to be stored as UCS-2 string>
	  --uComment <some comment to be stored as UCS-2 string>

2008-03-09 22:29  robert

	* include/lame.h:
	  under Linux we need some C-library includes here (why not on
	  Win32?)

2008-03-09 19:31  robert

	* libmp3lame/takehiro.c, libmp3lame/vbrquantize.c:
	  fixing some compiler warnings about:
	  - integer constant overflow
	  - signed/unsigned mismatch

2008-03-09 17:13  robert

	* Dll/BladeMP3EncDLL.c, dshow/aboutprp.cpp, frontend/brhist.c,
	  frontend/mp3x.c, frontend/parse.c, include/lame.h:
	  removing unneccessary stdlib includes from lame.h

2008-03-09 16:33  robert

	* vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_nasm.rules:
	  adding NASM build rule

2008-02-23 20:58  robert

	* frontend/parse.c, libmp3lame/id3tag.c:
	  adding ICONV character conversion (NOT for WIN32), because ID3
	  tags have to
	  be in ISO Latin-1 or UCS-2 text encoding. Previously we wrote
	  bytes into ID3
	  tags, as we got them from the commandline, in locale text
	  encoding.
	  
	  There are some experimental switches to store UCS-2 encoded ID3
	  tags:
	  --uArtist, --uAlbum, --uTitle, --uGenre and --uComment

2008-02-23 12:58  robert

	* libmp3lame/id3tag.c, libmp3lame/id3tag.h, libmp3lame/util.c:
	  locate and replace {TXXX,COMM} field if LANGUAGE and DESCRIPTION
	  match new one

2008-02-22 02:48  robert

	* libmp3lame/id3tag.c, libmp3lame/id3tag.h, libmp3lame/util.c:
	  preparing unicode id3v2 tags

2008-02-22 02:38  robert

	* libmp3lame/lame.c, libmp3lame/vbrquantize.c,
	  libmp3lame/version.c:
	  fixing some compiler warnings

2008-02-20 03:04  robert

	* vc_solution/vc9_lame_dll_example.vcproj:
	  VC9 Solution files

2008-02-20 02:40  robert

	* vc_solution/vc9_lame_dshow.vcproj:
	  VC9 Solution files

2008-02-20 01:53  robert

	* vc_solution/vc9_lame_acm.vcproj:
	  VC9 Solution files

2008-02-20 00:38  robert

	* vc_solution, vc_solution/vc9_lame.sln,
	  vc_solution/vc9_lame_acm.vcproj,
	  vc_solution/vc9_lame_acm_adbg.vcproj,
	  vc_solution/vc9_lame_acm_tinyxml.vcproj,
	  vc_solution/vc9_lame_clients.sln,
	  vc_solution/vc9_lame_dll.vcproj,
	  vc_solution/vc9_lame_dll_example.vcproj,
	  vc_solution/vc9_lame_dshow.vcproj,
	  vc_solution/vc9_lame_lame.vcproj,
	  vc_solution/vc9_lame_mp3x.vcproj,
	  vc_solution/vc9_libmp3lame.vcproj,
	  vc_solution/vc9_libmp3lame_dll.vcproj,
	  vc_solution/vc9_mpglib.vcproj:
	  VC9 Solution files

2008-02-15 01:27  robert

	* Dll/BladeMP3EncDLL.c:
	  use new API function for LAME-tag frame update

2008-02-15 00:23  robert

	* libmp3lame/VbrTag.c:
	  let the dummy Xing frame have a valid mp3 sync word

2008-02-10 18:38  robert

	* include/lame.def, include/lame.h, libmp3lame/id3tag.c:
	  fixes:
	  id3tag.c - some block scope missing
	  lame.h, lame.def: - some exported function was missing

2008-02-10 17:36  robert

	* frontend/main.c, include/lame.def, include/lame.h,
	  libmp3lame/VbrTag.c, libmp3lame/VbrTag.h, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/set_get.c, libmp3lame/version.h:
	  Adding some new API functions
	  - to en-/dis- able automatic ID3 tag writing into audio stream
	  (default on):
	  
	  void CDECL lame_set_write_id3tag_automatic(lame_global_flags *
	  gfp, int);
	  int CDECL lame_get_write_id3tag_automatic(lame_global_flags
	  const* gfp);
	  
	  - to query ID3 tags into a buffer:
	  
	  size_t CDECL lame_get_id3v1_tag(
	  lame_global_flags * gfp, unsigned char* buffer, size_t size);
	  size_t CDECL lame_get_id3v2_tag(
	  lame_global_flags * gfp, unsigned char* buffer, size_t size);
	  
	  - to query the final Xing/LAME tag frame into a buffer:
	  size_t CDECL lame_get_lametag_frame(
	  const lame_global_flags *, unsigned char* buffer, size_t size);
	  
	  Those new functions allow to encode a complete input file into
	  RAM first,
	  before writing it into a file/socket/whatever...

2008-02-03 14:28  robert

	* libmp3lame/psymodel.c:
	  VBRPSY: don't use ATH in 'additive masking' function

2008-01-09 20:46  rbrito

	* debian/changelog, debian/compat, debian/control, debian/rules:
	  Updated the Debian packaging.

2008-01-06 16:46  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c:
	  VBR-NEW can be mixed with free format now. The result will be a
	  free format
	  stream using VBR-NEW quantization routine.
	  
	  note: following free format bitrates are maximum bitrates which
	  make sense:
	  samplerate: 8 11 12 16 22 24 32 44 48
	  kbps 109 151 164 218 301 327 435 600 640(652)
	  
	  Using the maximum bitrate one may look and see how the VBR code
	  behaves
	  when the bitrate pressure is lowered to the theoretical minimum,
	  like:
	  lame -V0 -b600 --freeformat x.wav x.mp3

2008-01-03 00:03  aleidinger

	* configure:
	  regenerate with autoconf 2.61

2008-01-03 00:02  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  debian/Makefile.in, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, libmp3lame/vector/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in:
	  regenerate with automake 1.10

2008-01-03 00:01  aleidinger

	* aclocal.m4:
	  update with stuff from more recent autotools

2008-01-03 00:00  aleidinger

	* ltmain.sh:
	  update to a libtool 1.5.24 version

2008-01-02 23:58  aleidinger

	* configure.in:
	  After installing sndfile:
	  - fix SNDFILE_xxx
	  
	  Give a better failure description for some errors (sizeof check).
	  
	  If there are intXX_t but no uintXX_t, use the intXX_t with
	  "unsigned" instead
	  of trying to find the right K&R-C integral type.
	  
	  Use macho as the format type for NASM on OSX, but disable the
	  nasm use by
	  default there, as there seems to be a problem with the nasm code
	  on OSX.

2008-01-02 22:00  aleidinger

	* configure.in:
	  - If SNDFILE_xxx are defined and the fileio is sndfile, then
	  assume the
	  user is knowing what he does (some systems don't have pkg-config
	  but
	  have sndfile).
	  - Fix the problem case where SNDFILE_CFLAGS has spaces at the
	  beginning.

2008-01-02 20:30  robert

	* libmp3lame/id3tag.c:
	  fix for tracker item: [ 1861450 ] TLEN tag fixed on encode from
	  stdin with ID3v2 tags

2007-12-26 22:04  robert

	* ACM/ACMStream.cpp, Dll/BladeMP3EncDLL.c:
	  removing obsolete function calls

2007-12-26 13:12  robert

	* Makefile.MSVC, frontend/parse.c, include/lame.h,
	  libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/set_get.c, libmp3lame/set_get.h, libmp3lame/util.h:
	  removing dead code

2007-12-25 16:15  robert

	* libmp3lame/vbrquantize.c:
	  some small code simplification

2007-12-25 00:16  robert

	* libmp3lame/presets.c, libmp3lame/psymodel.c:
	  clean code from obsolete branches

2007-12-24 16:00  robert

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/set_get.c, libmp3lame/takehiro.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  - removing legacy ISO quantization code
	  - vbr-new now has only two -qx settings left: q0 and q5
	  - vbr-new can't use legacy GPSYCHO anymore
	  
	  TODO:
	  + remove legacy GPSYCHO code, as we don't use it anymore
	  + remove superflouos psy-model selection switches
	  + remove superflouos psy-model configuration switches like
	  --cwlimit

2007-12-20 14:44  robert

	* libmp3lame/vbrquantize.c:
	  in rare cases, where one channel is completely zero, some illegal
	  "scalefactor
	  selection info" might be used in VBR-NEW. fixed

2007-12-17 20:54  robert

	* doc/html/history.html:
	  fixing typo

2007-12-17 00:08  robert

	* libmp3lame/vbrquantize.c:
	  fixing typo

2007-12-16 17:56  robert

	* doc/html/history.html, libmp3lame/psymodel.c:
	  disabling the use of the "interchannel masking feature" in
	  dual-channel case

2007-12-16 17:18  robert

	* libmp3lame/psymodel.c, libmp3lame/vbrquantize.c:
	  VBR-NEW:
	  -quantization: another "out of bits" strategy
	  -psymodel: dropped "inter channel effect", because the effect
	  doesn't save
	  many bits, but raises the chance of annoying artefacts a lot.

2007-12-14 02:52  robert

	* libmp3lame/psymodel.c:
	  found new test sample SNAPS which needs some more short blocks

2007-12-09 22:47  robert

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.h:
	  merger from test branch:
	  - features a new psy model, a modification from NSPSY
	  
	  VBR NEW uses the new psy model, unless you call lame with
	  --nspsytune, or
	  the developer only switch --psymodel x, with x in {1,2}
	  
	  features of the new psy model:
	  - speed: it does determine the resulting block type before doing
	  the fft
	  and other psy stuff and will calc long/short blocks only as
	  necessary
	  - interchannel masking effect: it will be calculated after the
	  mid-side fix
	  and it's working on convolution bands, instead of scalefactor
	  bands
	  - mid-side fix: calculated on convolution bands, instead of sf
	  bands
	  - mask_adjust feature: it's now used earlier in the convolution
	  calculation

2007-11-26 00:14  robert

	* Makefile.unix, frontend/get_audio.c, frontend/parse.c,
	  libmp3lame/psymodel.c, libmp3lame/vbrquantize.c:
	  psymodel.c:
	  - bug fix in init_numlines: it returned sometimes a too large
	  npart_[sl]
	  - fixing some more compiler warnings
	  get_audio.c:
	  - some fix in error-report in case of unsupported input format

2007-11-21 21:30  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.h:
	  bug fix for bin_search_StepSize. It returned illegal values,
	  leading to
	  assertion failures in fast encoding mode, which only adjusts
	  global step size.

2007-11-01 19:52  robert

	* frontend/get_audio.c:
	  enabling --signed / --unsigned switches for 8 bit raw pcm input
	  data

2007-11-01 19:25  robert

	* frontend/get_audio.c, frontend/parse.c:
	  [ 1823881 ] lame 3.97 produces complete garbage
	  now invalid WAV/AIFF files are rejected, instead of trying to
	  encode them as
	  raw pcm files.

2007-11-01 19:23  robert

	* doc/man/lame.1:
	  [ 1158193 ] Man page doesn't tell about stereo format in raw
	  [ 1158253 ] Data format is not described in the manpage
	  added some info to manpage

2007-11-01 19:16  robert

	* libmp3lame/lame.c:
	  bitrate_index may have been uninitialized under some rare
	  circumstances

2007-10-24 02:04  robert

	* libmp3lame/psymodel.c:
	  back to the roots. Maybe I'll branch from here

2007-10-23 23:23  robert

	* libmp3lame/vbrquantize.c:
	  turn of scalefac scale feature

2007-10-23 02:46  robert

	* libmp3lame/psymodel.c:
	  yeah, ok it's C and not C++
	  open block scope for local variable

2007-10-23 02:31  robert

	* libmp3lame/psymodel.c, libmp3lame/vbrquantize.c:
	  dropping Naoki's idea of "simultaneous masking addition" for the
	  newer VBR code
	  for now

2007-10-22 12:00  robert

	* libmp3lame/presets.c, libmp3lame/psymodel.c:
	  re-establishing previous mask adjustments

2007-10-21 00:45  robert

	* libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/version.h:
	  recalibration of VBR new's compression level settings, because of
	  recent bug fixes

2007-10-19 00:10  robert

	* Makefile.unix:
	  moving GTK config script to MP3x target compilation, it's not
	  needed outside MP3x

2007-10-18 02:54  robert

	* frontend/get_audio.c:
	  seems, VC compiler doesn't like redundant semicolon, fixed

2007-10-18 02:09  robert

	* libmp3lame/machine.h, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/set_get.c:
	  PSY: the newer vbr code uses tonalality estimation from long
	  blocks for short blocks

2007-10-16 01:07  robert

	* doc/html/history.html, frontend/get_audio.c:
	  Fix for bugtracker item: [ 1813496 ] AIFF parsing bug

2007-10-14 20:18  robert

	* frontend/get_audio.c, libmp3lame/id3tag.h, libmp3lame/util.c:
	  some more compiler and compiler-warning fixes

2007-10-14 19:54  robert

	* frontend/get_audio.c, frontend/get_audio.h, frontend/mp3x.c,
	  frontend/parse.c, frontend/portableio.c, include/lame.h,
	  libmp3lame/bitstream.c, libmp3lame/id3tag.c, libmp3lame/lame.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/set_get.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vector/xmm_quantize_sub.c:
	  eliminating some of the several hundred compiler warnings

2007-10-14 14:16  robert

	* doc/html/history.html:
	  updating history to latest changes:
	  - bug fix scalefactor selection in vbr new code
	  - feature added: WAVE_FORMAT_EXTENSIBLE

2007-10-14 13:33  robert

	* frontend/get_audio.c:
	  feature request [ 1811483 ] WAVE_FORMAT_EXTENSIBLE support
	  
	  We still support PCM wave files only with our little wave reader,
	  it is just that we now accept them when written in wave format
	  extensible.

2007-10-13 20:34  robert

	* libmp3lame/vbrquantize.c:
	  Bug-fix: the quantization-find-routine could under some
	  circumstances return
	  some way too large quantization step size. For example, the song
	  Piku from
	  The Chemical Brothers had almost 550 scalefactors wrong. For most
	  songs this
	  fix makes no difference.

2007-10-08 20:59  robert

	* frontend/brhist.c:
	  replaced variable 'lines' by 'lines_used', because term.h has
	  some macro
	  definition for 'lines'.

2007-09-29 19:17  robert

	* libmp3lame/vbrquantize.c:
	  - Removing dead code.
	  - Avoiding unnecessary search for quantization step sizes. This
	  speeds-up lower quality encodings.
	  If energy is below masking threshold, the resulting step size
	  will be at 255. So no need to search what we already know.

2007-08-29 13:44  aleidinger

	* configure:
	  regen

2007-08-29 13:43  aleidinger

	* configure.in:
	  Don't change the include/library path, if there's no
	  corresponding soundfile
	  path set.
	  
	  This is supposed to fix the compile problem with soundfile
	  Rogério sees on
	  Debian.

2007-08-20 21:54  robert

	* Makefile.MSVC:
	  removing duplicate target_dir references

2007-08-12 18:30  robert

	* Makefile.MSVC:
	  let targets link against multi threaded C-library

2007-08-12 00:07  robert

	* doc/html/history.html, libmp3lame/vbrquantize.c,
	  libmp3lame/version.h:
	  small speedup in VBR NEW
	  
	  we are now at 3.98 beta 5, a release candidate for 3.98 stable?

2007-08-12 00:04  robert

	* logo:
	  Removing the 6 MB fat logo files from encoder sources.
	  We can expect the logo files to be almost static and
	  having them in a separate directory seems to be the
	  better choice.

2007-08-11 23:06  robert

	* testcase.mp3:
	  updating "testcase.mp3" for LAME 3.98 beta 5

2007-08-11 21:44  robert

	* libmp3lame/psymodel.c:
	  Short block masking: the CBR/ABR code seems to rely on the old
	  masking definition. The VBR NEW code on the other hand needs the
	  lowered masking.

2007-07-30 06:51  aleidinger

	* ChangeLog:
	  Add the changelog for the last ~10 months.

2007-07-30 06:39  aleidinger

	* configure:
	  regen with autoconf 2.61

2007-07-30 06:37  aleidinger

	* configure.in:
	  oops, previous fix for the gcc version pattern matching was
	  wrong...

2007-07-30 06:30  aleidinger

	* configure.in:
	  - fix the pattern matching for the gcc version check
	  - improve the optimization stuff for gcc 4.x

2007-07-30 06:28  aleidinger

	* config.h.in:
	  regen with autoconf 2.61

2007-07-30 06:28  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, debian/Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  libmp3lame/vector/Makefile.in, mac/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in:
	  regen with automake 1.10

2007-07-29 22:46  robert

	* doc/html/history.html:
	  adding latest bug fixes to history

2007-07-29 22:08  robert

	* ACM/ADbg/ADbg.dsp, ACM/lameACM_vc6.dsp, ACM/tinyxml/tinyxml.dsp,
	  Dll/LameDll_vc6.dsp, dshow/dshow.dsp, frontend/lame_vc6.dsp,
	  frontend/mp3x_vc6.dsp, include/lame.def, lame_vc6.dsw,
	  libmp3lame/libmp3lame_dll_vc6.dsp, libmp3lame/libmp3lame_vc6.dsp,
	  mpglib/mpglib_vc6.dsp:
	  updating VC6 project files
	  
	  from MSDN: DLLs should be linking against multi-threaded MS VC
	  rountime library

2007-07-29 16:02  robert

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  Fixing a possible resource leak

2007-07-28 12:48  aleidinger

	* configure:
	  regen (new LAME version)

2007-07-28 00:41  robert

	* frontend/get_audio.c:
	  task tracker item [ 1445175 ] Input being stdin fails in Windows
	  on WAV files
	  and item [ 1576370 ] Fix for piping in a non PCM file via stdin
	  under Windows
	  
	  The workaround tests the actual file descriptor being a pipe or
	  not.
	  In case being a pipe, the number of bytes to skip are simply
	  read.

2007-07-27 08:49  bouvigne

	* README:
	  3.98

2007-07-24 17:46  bouvigne

	* frontend/brhist.c, frontend/brhist.h, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/gpkplotting.c,
	  frontend/gpkplotting.h, frontend/gtkanal.c, frontend/gtkanal.h,
	  frontend/lametime.c, frontend/lametime.h, frontend/main.c,
	  frontend/main.h, frontend/parse.c, frontend/timestatus.c,
	  frontend/timestatus.h, include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/VbrTag.h, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/fft.h, libmp3lame/id3tag.c,
	  libmp3lame/l3side.h, libmp3lame/lame-analysis.h,
	  libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/newmdct.h, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/set_get.c, libmp3lame/set_get.h, libmp3lame/tables.c,
	  libmp3lame/tables.h, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h, libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/xmm_quantize_sub.c, libmp3lame/version.c,
	  libmp3lame/version.h:
	  Library GPL -> Lesser GPL

2007-07-24 15:26  robert

	* frontend/get_audio.c, frontend/parse.c:
	  bug item [ 1160757 ] --big-endian and --little-endian don't work
	  with -m s -a
	  and item [ 1160741 ] --big-endian and --little-endian are swapped
	  
	  the --big-endian/--little-endian switches do now work with our
	  minimal
	  wave and pcm reader too.
	  
	  The problems reported in the above bug tracker items should be
	  solved now

2007-07-22 18:49  robert

	* libmp3lame/lame.c:
	  bug fix: [ 1746336 ] Incorrect Bitrate with ABR und --resample,
	  LAME 3.98b4
	  
	  There was a typo in patch Revision 1.294 "restrict mpeg 2.5 to
	  64kbps"
	  which set the average bitrate up to the maximum allowed bitrate.

2007-07-20 23:23  robert

	* libmp3lame/bitstream.c:
	  fixing task tracker item: [ 1755981 ] Percent signs in print
	  format not escaped

2007-07-16 20:06  bouvigne

	* include/lame.h:
	  gpl -> lgpl (we forgot about it a long time ago)

2007-07-10 21:05  robert

	* dshow/Property.rc:
	  updating version info

2007-07-01 20:05  robert

	* mpglib/common.c, mpglib/common.h, mpglib/interface.c,
	  mpglib/layer1.c, mpglib/layer2.c, mpglib/layer2.h,
	  mpglib/mpg123.h:
	  we will always use Layer1 and 2 decoding

2007-07-01 16:17  robert

	* doc/html/history.html:
	  word wrapping fixed

2007-07-01 16:16  robert

	* doc/html/history.html:
	  adding latest changes for Task tracker item [ 1719593 ] and [
	  1742623 ]

2007-07-01 14:39  robert

	* frontend/get_audio.c, frontend/get_audio.h, frontend/gtkanal.c,
	  frontend/main.c, frontend/parse.c:
	  trying to fix Bugs item #1742623

2007-06-30 17:14  robert

	* USAGE, frontend/parse.c, include/lame.h, libmp3lame/id3tag.c:
	  bug fix: [ 1719593 ] Track numbers > 255 not allowed even with
	  --id3v2-only

2007-06-27 21:23  robert

	* configMS.h, configure.in:
	  changing version number 3.97 to 3.98

2007-06-27 21:22  robert

	* include/lame.def:
	  adding newly id3tag functions to DLL export definition

2007-06-26 01:33  robert

	* libmp3lame/VbrTag.c:
	  I missed the swapped arguments in the patch by Jonathan Stott,
	  now fixed.

2007-06-26 00:57  robert

	* doc/html/contributors.html, doc/html/history.html,
	  libmp3lame/VbrTag.c, libmp3lame/lame.c:
	  Patch by Jonathan Stott:
	  
	  Bug tracker items: [ 1590693 ] ID3v2 tag not writing, [ 1636267 ]
	  ID3v2 tags overwritten
	  If the output file is opened 'write-only', then LAME can't update
	  the LAME tag.
	  In this case LAME silently overwrote the first bytes of the file
	  and an
	  optional ID3v2 tag disappeared. Now an error message will be
	  printed and
	  no data is written in this case.

2007-06-23 18:43  robert

	* libmp3lame/id3tag.c, libmp3lame/util.c, libmp3lame/util.h:
	  fixing another possible memory leak

2007-06-23 17:39  robert

	* frontend/main.c, frontend/parse.c, libmp3lame/id3tag.c,
	  libmp3lame/id3tag.h, libmp3lame/util.c, libmp3lame/version.h:
	  Some simplifications on ID3 tag usage:
	  - specifying an unlisted ID3v1 genre does not count as an error
	  anymore, it is mapped to "Other" and passed as-is to an ID3v2
	  genre tag
	  - an ID3 track argument is passed directly to an ID3v2 TRCK tag,
	  allowing for example "002/012"

2007-06-23 15:16  robert

	* libmp3lame/id3tag.c, libmp3lame/util.c:
	  Now ID3 tag are local copies of given parameters. This allows to
	  pass strings located on the program stack and you don't have to
	  think about the lifetime of your strings you pass as arguments.

2007-06-23 13:27  robert

	* libmp3lame/util.c:
	  Bug fix: Nyaochi's "ID3v2 extension" patches introduced some
	  memory leaks

2007-06-04 22:31  robert

	* doc/html/history.html:
	  ups: moved history entry "TLEN bug" up into correct list

2007-06-04 22:25  robert

	* doc/html/history.html:
	  added: bugfix TLEN computation

2007-06-04 22:04  robert

	* libmp3lame/id3tag.c:
	  bug fix: [ 1730875 ] TLEN calculation wrong
	  fixing the calculation of "playlength in ms" written into the
	  ID3v2 tag TLEN

2007-05-30 23:17  rjamorim

	* logo, logo/logoERGB.ai, logo/logoEncodedRGB.ai, logo/logoPDF.pdf,
	  logo/logoRGB.ai, logo/logoRGB.eps, logo/logoRGB.svg,
	  logo/logoe.ai, logo/logoe.eps, logo/logoe.svg, logo/logofull.ai,
	  logo/logofull.eps, logo/logofull.svg, logo/logomini.ai,
	  logo/logomini.eps, logo/logomini.svg:
	  Initial import of the Lame logo in several vectorial formats

2007-05-25 03:38  robert

	* libmp3lame/vbrquantize.c:
	  VBR_NEW: more tuned 'out of bits' strategy

2007-05-24 13:26  bouvigne

	* doc/html/contributors.html, doc/html/history.html,
	  doc/html/id3.html, doc/man/lame.1, frontend/parse.c:
	  support for total tracks count in the frontend (id3 tags)

2007-05-24 13:10  bouvigne

	* libmp3lame/id3tag.c:
	  adding a total track count should add an id3 V2 tag

2007-05-24 12:58  bouvigne

	* doc/html/history.html, doc/html/id3.html, frontend/main.c,
	  frontend/parse.c, frontend/parse.h, include/lame.h,
	  libmp3lame/id3tag.c, libmp3lame/id3tag.h:
	  Albumart (APIC ID3v2.3 frame) patch by Nyaochi

2007-05-24 12:29  bouvigne

	* doc/html/contributors.html, doc/html/history.html,
	  doc/html/id3.html, frontend/parse.c, include/lame.h,
	  libmp3lame/id3tag.c, libmp3lame/id3tag.h:
	  Ability to set user-defined ID3v2.3 frame (patch from Nyaochi)

2007-05-23 14:44  bouvigne

	* doc/html/contributors.html, doc/html/history.html,
	  doc/html/index.html:
	  Updated history

2007-05-22 10:10  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/version.h:
	  ok, now the real bug fix, forget the last commit:
	  the ABR routine did not apply format specific bitrate constrains
	  when allocating bits; now it does

2007-05-22 08:39  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  bug fix: reduce side was given a too large 'max bits' parameter,
	  resulting in assertion aborts

2007-05-21 22:20  robert

	* libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  bug fix for feature request:
	  "adjust user min/max bitrate if outside of allowed range (ex: -V9
	  -B320)
	  Requested by Benski"
	  
	  Some input files with unusual sample frequencies couldn't be
	  encoded at CBR anymore, no matter what resample/bitrate options
	  where passed to LAME.
	  (Testsample: FX319.WAV)

2007-05-20 18:07  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/version.h:
	  VBR_NEW: 'out of bits' situation improved

2007-05-17 22:25  robert

	* ACM/ACM.cpp, libmp3lame/VbrTag.c:
	  Bugs item #1693461: fixing memory leaks in LAME ACM

2007-05-16 02:43  robert

	* libmp3lame/version.h:
	  we are now at 3.98 beta 1 (of many more to come)

2007-05-16 01:30  robert

	* frontend/parse.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/set_get.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c:
	  - some code clean-up
	  - bug fix: in rare cases 'on_pe' returned target bits violating
	  the 4095 bit
	  maximum constrain for a granule
	  - VBR-NEW: modified strategy to handle 'out of bits' cases

2007-05-13 18:31  robert

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/id3tag.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/util.h, libmp3lame/version.h:
	  bug fix: [ 1711980 ] LAME writes invalid Xing header when ID3
	  tags exist

2007-03-12 21:56  bouvigne

	* libmp3lame/util.c:
	  adjust user min/max bitrate if outside of allowed range (ex: -V9
	  -B320)
	  
	  Requested by Benski

2007-01-13 18:02  robert

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/util.h:
	  bug fix.
	  bug tracker entry: [ 1634092 ] encoder.c doesn't compile with
	  --disable-brhist

2007-01-09 23:10  robert

	* Makefile.unix:
	  adding vector/xmm_quantize_sub.c

2007-01-09 22:24  bouvigne

	* libmp3lame/libmp3lame_vc8.vcproj:
	  directory x86_64 -> vector

2007-01-09 12:02  aleidinger

	* libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/psymodel.c:
	  fix the easy warnings

2007-01-09 10:37  aleidinger

	* libmp3lame/x86_64:
	  Remove all files, they reside in the vector directory now.
	  
	  This breaks the VC build until Gabriel or Robert catch up with
	  those
	  changes.

2007-01-09 10:31  aleidinger

	* libmp3lame/quantize.c:
	  include the right intrinsics header and only if we need it

2007-01-09 10:29  aleidinger

	* Makefile.MSVC:
	  use the code in the vector directory

2007-01-09 10:27  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  config.h.in, configure, debian/Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in:
	  regen

2007-01-09 10:24  aleidinger

	* libmp3lame/lame.c, libmp3lame/quantize.c:
	  remove the CPUCCODE check, it is not needed anymore

2007-01-09 10:23  aleidinger

	* libmp3lame/i386/Makefile.am:
	  the quoting does not work out as expected, remove it

2007-01-09 10:20  aleidinger

	* configure.in:
	  Use the vector lib on amd64 and x86.
	  
	  This only works with gcc when -msse is specified in CFLAGS at
	  configure
	  time. Currently I get a sigbus with gcc 3.4.6 with the vector
	  lib.
	  
	  Remove the HAVE_GTK define, it is not used anymore.

2007-01-09 10:18  aleidinger

	* libmp3lame/Makefile.am:
	  add the vector lib and remove the x86_64 part

2007-01-09 10:15  aleidinger

	* libmp3lame/vector, libmp3lame/vector/Makefile.am,
	  libmp3lame/vector/Makefile.in, libmp3lame/vector/lame_intrin.h,
	  libmp3lame/vector/xmm_quantize_sub.c:
	  this is a copy of the x86_64 parts with modified automake stuff

2007-01-07 21:28  robert

	* Makefile.MSVC:
	  do not build legacy BLame DLL on make target all

2007-01-07 20:00  aleidinger

	* libmp3lame/i386/Makefile.in:
	  regen

2007-01-07 20:00  aleidinger

	* libmp3lame/i386/Makefile.am:
	  fftsse.nas is used now.

2007-01-07 18:03  robert

	* frontend/lame_vc6.dsp, frontend/mp3x_vc6.dsp, lame_vc6.dsw,
	  libmp3lame/libmp3lame_dll_vc6.dsp, libmp3lame/libmp3lame_vc6.dsp,
	  mpglib/mpglib_vc6.dsp:
	  VC6 project files updated, added project to generate
	  Libmp3lame-dynamic.lib/dll

2007-01-07 14:51  robert

	* Makefile.MSVC:
	  Makefile with some basic Libmp3lame-dynamic.lib generation
	  ability

2007-01-07 14:31  robert

	* include/lame.def:
	  export definition for Libmp3lame-dynamic.lib

2007-01-07 14:30  robert

	* Makefile.unix, frontend/parse.c, include/lame.h,
	  libmp3lame/fft.c, libmp3lame/lame.c, mpglib/mpg123.h:
	  adding FFTSSE assembler code to the HAVE_NASM conditional code

2007-01-07 14:28  robert

	* ., frontend, lame_clients_vc6.dsw:
	  collecting all LAME client projects in lame_clients_vc6.dsw
	  workspace.

2007-01-07 14:24  robert

	* dshow/README, dshow/dshow.dsp:
	  DirectShow Filter links against Libmp3lame-dynamic.lib now

2007-01-07 14:23  robert

	* Dll/LameDll_vc6.dsp:
	  BLAME DLL links against Libmp3lame-dynamic.lib now

2007-01-07 14:22  robert

	* ACM/ADbg/ADbg.dsp, ACM/lameACM_vc6.dsp, ACM/tinyxml/tinyxml.dsp:
	  ACM links against Libmp3lame-dynamic.lib now

2007-01-07 14:21  robert

	* ACM/ADbg/ADbg.dsp, ACM/lameACM_vc6.dsp:
	  ACM links against Libmp3lame-dynamic.lib now

2006-12-28 21:56  robert

	* Dll/BladeMP3EncDLL.c:
	  we are now at 3.98a11

2006-12-28 21:49  robert

	* dshow/dshow.dsp, libmp3lame/version.h:
	  we are now at 3.98a11

2006-12-28 21:41  robert

	* doc/html/history.html:
	  update history

2006-12-28 20:41  robert

	* dshow/Encoder.cpp, dshow/Encoder.h, dshow/Property.rc,
	  dshow/dshow.dsp, dshow/dshow.dsw:
	  DirectShow filter uses LAME Library API now

2006-12-27 22:19  robert

	* libmp3lame/reservoir.c:
	  small change in the bitreservoir handling, enable the old
	  --strictly-enforce-ISO switch again

2006-12-27 22:09  robert

	* doc/html/history.html:
	  updating history entries

2006-12-25 21:37  robert

	* ACM/ACM.cpp, ACM/ACM.h, ACM/ACMStream.cpp, ACM/ACMStream.h,
	  ACM/AEncodeProperties.cpp, ACM/DecodeStream.h, ACM/acm.rc,
	  ACM/lameACM_vc6.dsp, ACM/main.cpp:
	  some changes, now I can compile the ACM stuff at least

2006-12-25 16:15  robert

	* libmp3lame/reservoir.c:
	  relaxing the FhG bitreservoir restriction a little bit:
	  at 320 kbps it seems to be possible to use the reservoir if we
	  limit it to almost the size of bits used for sideinfo

2006-12-23 13:54  aleidinger

	* libmp3lame/x86_64/Makefile.in:
	  regen

2006-12-23 13:54  aleidinger

	* libmp3lame/x86_64/Makefile.am:
	  fix the build on non x86 CPUs

2006-12-21 09:19  aleidinger

	* libmp3lame/x86_64/quantize_sub.c:
	  Add a comment to the #endif so that you know where it belongs.

2006-12-21 09:16  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in, configure,
	  debian/Makefile.in, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, libmp3lame/x86_64/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in:
	  regen

2006-12-21 09:14  aleidinger

	* libmp3lame/Makefile.am:
	  Handle CPU specific C code in a generic way.

2006-12-21 09:09  aleidinger

	* libmp3lame/i386/Makefile.am:
	  Quote some parts in case directory names contain spaces.

2006-12-21 09:08  aleidinger

	* aclocal.m4, config.h.in:
	  regen

2006-12-21 09:04  aleidinger

	* configure.in:
	  Additional stuff for x86_64. First cut...

2006-12-21 09:03  aleidinger

	* acinclude.m4:
	  Properly quote some parts according to the new rules auf aclocal.

2006-12-21 09:02  aleidinger

	* libmp3lame/x86_64/Makefile.am:
	  First cut on adding the x86_64 code. Only tested on IA32.

2006-12-21 09:00  aleidinger

	* libmp3lame/lame.c, libmp3lame/quantize.c:
	  On x86 we also have the xmmintrin header, but we don't use such
	  code there.
	  So not only check for the header, but also for a CPUCCODE define.
	  This may
	  be refined later depending on further inline C code showing up or
	  not.

2006-12-21 02:18  robert

	* libmp3lame/machine.h, libmp3lame/version.h:
	  we are now at patch level 10

2006-12-21 02:17  robert

	* libmp3lame/psymodel.c:
	  fixing some divide by zero error

2006-12-20 00:17  robert

	* frontend/get_audio.c, frontend/parse.c:
	  let LAME be more strict when it comes to unrecognized program
	  options or unsupported input file formats

2006-12-19 18:50  bouvigne

	* libmp3lame/libmp3lame_vc8.vcproj:
	  include paths

2006-12-19 01:23  robert

	* frontend/parse.c:
	  add some help text entries

2006-12-19 01:23  robert

	* libmp3lame/lame.c:
	  disabling temporal masking effect by default for vbr-mtrh, it
	  works against other tunings in this mode

2006-12-18 23:11  robert

	* libmp3lame/psymodel.c:
	  fixing some index problems in convolution loops

2006-12-17 23:21  robert

	* libmp3lame/version.h:
	  we are at 3.98 alpha 9

2006-12-17 21:16  robert

	* libmp3lame/psymodel.c:
	  fixing two possible out of range index errors accessing fftenergy
	  array
	  for GPSYCHO too

2006-12-17 20:53  robert

	* libmp3lame/psymodel.c, libmp3lame/util.h:
	  fixing two possible out of range index errors accessing fftenergy
	  array
	  
	  convert partition to scalefac:
	  as we replaced ISO tables with own calculations we forgot to add
	  proper
	  weightening for the convolution band where the transition from
	  one scalefactor
	  to th next scalefactor happens. Now added.

2006-12-17 20:49  robert

	* libmp3lame/bitstream.c:
	  assert didn't match valid input range: bitrate 8 - 640 kbps

2006-12-15 22:16  robert

	* libmp3lame/version.h:
	  we are now at 3.98 alpha 8

2006-12-15 22:15  robert

	* libmp3lame/psymodel.c:
	  2 bug fixes
	  a) in some cases we did access uninitialized data in conversion
	  from partition bands to scalefac bands
	  b) in nsspy-code the conversion from partition to scalefac bands
	  for short blocks did not match the long block case, nor gpsycho.

2006-12-07 10:50  robert

	* frontend/main.c, libmp3lame/util.c:
	  disabling the "process affinity kludge"

2006-11-27 20:28  robert

	* libmp3lame/lame.c:
	  - removed unused function parameter nVbrScale in function
	  PutVbrTag

2006-11-20 20:02  robert

	* libmp3lame/VbrTag.c, libmp3lame/VbrTag.h:
	  - removed unused function parameter nVbrScale in function
	  PutVbrTag
	  - reduced compilation warnings

2006-11-15 02:25  robert

	* libmp3lame/lame.c:
	  bug fix: [ 1596306 ] lame 3.97: "fatal error during
	  initialization" with this wav
	  
	  an invalid MPEG samplerate was returned by optimum_samplefreq
	  function

2006-11-15 02:21  robert

	* Makefile.MSVC, libmp3lame/quantize.c:
	  fixing compile problem with quantize_sub.c

2006-11-12 11:44  bouvigne

	* doc/html/history.html:
	  history

2006-11-11 17:58  bouvigne

	* libmp3lame/libmp3lame_vc8.vcproj, libmp3lame/quantize.c,
	  libmp3lame/x86_64, libmp3lame/x86_64/lame_intrin.h,
	  libmp3lame/x86_64/quantize_sub.c:
	  intrinsics functions are now in a separate directory (as in v4)

2006-11-11 17:32  bouvigne

	* frontend/mp3x_vc8.vcproj:
	  bring back mp3x to life (vc8)

2006-11-11 17:14  bouvigne

	* Dll/LameDll_vc8.vcproj, README.WINGTK, frontend/lame_vc8.vcproj,
	  lame_vc8.sln, libmp3lame/libmp3lame_vc8.vcproj,
	  mpglib/mpglib_vc8.vcproj:
	  bring back mp3x to life (vc8)

2006-11-08 01:53  robert

	* frontend/get_audio.c, frontend/main.c, frontend/main.h,
	  frontend/parse.c:
	  feature request: [ 1588283 ] Flushing output stream in lame.exe
	  
	  'flush' option added

2006-11-07 01:35  robert

	* libmp3lame/gain_analysis.c, libmp3lame/id3tag.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/util.c:
	  fixing some compiler warnings, removing some unused
	  lame_global_flags elements

2006-11-07 00:36  robert

	* frontend/parse.c, libmp3lame/VbrTag.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/quantize.c,
	  libmp3lame/set_get.c, libmp3lame/util.h:
	  removing dead code:
	  side channel sparsing didn't work as hoped

2006-11-06 22:39  robert

	* frontend/brhist.c, frontend/get_audio.c, frontend/parse.c:
	  eliminating some compiler warnings

2006-11-05 18:54  robert

	* frontend/console.c:
	  bug tracker item:
	  [ 1585942 ] lame not --silent when TERM not set
	  
	  in case LAME was build with TERMCAP defined and no TERM
	  environment is
	  defined, now we do not issue an error message and silently
	  fallback to the default
	  behaviour as if LAME was compiled without TERMCAP defined.

2006-11-05 17:49  robert

	* frontend/timestatus.c:
	  include string header file added
	  unused variable removed in encoder_progress_end

2006-11-05 16:08  robert

	* frontend/main.c, frontend/timestatus.c, frontend/timestatus.h:
	  refactoring progress display code out of encoding loop
	  it's located in timestatus.c now
	  
	  Windows specific:
	  fixed parameter type for SetProcessAffinityMask
	  Gabriel: do we need similar things for Win64 too?

2006-11-05 13:37  robert

	* frontend/console.c:
	  dead code removed

2006-11-05 02:30  robert

	* libmp3lame/libmp3lame_vc8.vcproj:
	  using fftsse.nas

2006-11-05 01:55  robert

	* Makefile.MSVC:
	  using fftsse.nas

2006-11-04 18:59  robert

	* Makefile.MSVC, Makefile.unix, frontend/parse.c, libmp3lame/fft.c,
	  libmp3lame/i386/fft3dn.nas, libmp3lame/i386/fftsse.nas,
	  libmp3lame/i386/nasm.h, libmp3lame/lame.c, libmp3lame/version.h:
	  we are now at patch level 7
	  
	  changes in lame frontend switches
	  -k removed,
	  call lowpass and highpass if you need to change them
	  --short/noshort/allshort - degraded into DEVELOPER ONLY switches
	  normal users shouldn't use them
	  -X -Z degraded to DEVELOPER ONLY switches
	  -X is too tough to communicate to end users
	  -Z isn't used actualy
	  
	  fft3dnow and fftsse nasm code taken from LAME4 branch
	  but not added to the configure stuff/VC project files

2006-10-18 20:29  bouvigne

	* Dll/LameDll_vc8.vcproj:
	  added back the .def config

2006-10-05 21:07  bouvigne

	* libmp3lame/lame.c:
	  error in a test (noticed by Robert)

2006-09-30 19:55  kylev

	* lame.spec.in:
	  we distribute a .gz, not a .bz2, and there really isn't a public
	  URL for the file

2006-09-30 15:37  aleidinger

	* Dll/Makefile.in, Makefile.in, configure, frontend/Makefile.in,
	  libmp3lame/Makefile.in, mpglib/Makefile.in:
	  regen

2006-09-30 15:35  aleidinger

	* configure.in:
	  Only check for sndfile if it is requested.

2006-09-30 09:17  bouvigne

	* Dll/Makefile.am, Makefile.am, frontend/Makefile.am,
	  libmp3lame/Makefile.am, mpglib/Makefile.am:
	  updated the .am files to change vc7 to vc8

2006-09-24 16:17  bouvigne

	* doc/html/history.html:
	  updated history

2006-09-23 21:46  bouvigne

	* doc/html/history.html:
	  release 3.97

2006-09-23 19:30  aleidinger

	* configure:
	  regen

2006-09-23 19:29  aleidinger

	* config.h.in, configure.in:
	  check for xmmintrin.h

2006-09-23 08:25  bouvigne

	* Dll/BladeMP3EncDLL.c:
	  fix in boolean comparison for Delphi progs (bug #1536751)

2006-09-22 19:18  aleidinger

	* ChangeLog:
	  update from CVS history

2006-09-22 18:51  aleidinger

	* configMS.h:
	  Borland C seems to follow the lead of MSC.
	  
	  Bug ID: 1563522

2006-09-10 14:25  bouvigne

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/tables.c,
	  libmp3lame/util.c:
	  restrict mpeg 2.5 to 64kbps

2006-09-03 15:07  bouvigne

	* Dll/LameDll_vc7.vcproj, frontend/lame_vc7.vcproj,
	  frontend/mp3x_vc7.vcproj, lame_vc7.sln,
	  libmp3lame/libmp3lame_vc7.vcproj, mpglib/mpglib_vc7.vcproj:
	  removed unmaintained VC7 project files

2006-07-30 13:53  bouvigne

	* configMS.h, libmp3lame/util.c:
	  enable SSE code when using x64

2006-07-30 09:51  bouvigne

	* Dll/LameDll_vc8.vcproj, frontend/lame_vc8.vcproj, lame_vc8.sln,
	  libmp3lame/libmp3lame_vc8.vcproj, mpglib/mpglib_vc8.vcproj:
	  x64 builds

2006-07-30 09:32  bouvigne

	* Dll/LameDll_vc8.vcproj:
	  updated project

2006-07-30 09:11  bouvigne

	* Dll/LameDll_vc8.vcproj, lame_vc8.sln:
	  added lame DLL VC8 project

2006-07-14 16:06  bouvigne

	* frontend/lame_vc8.vcproj, lame_vc8.sln,
	  libmp3lame/libmp3lame_vc8.vcproj, mpglib/mpglib_vc8.vcproj:
	  VC8 projects

2006-07-14 15:58  bouvigne

	* libmp3lame/psymodel.c:
	  slight change in the short block detection safety threshold

2006-06-28 23:05  robert

	* libmp3lame/lame.c:
	  full outer loop search disabled because of audible distortions it
	  may generate

2006-06-28 21:05  robert

	* libmp3lame/psymodel.c:
	  normalizing the spreading function (vbr-new)

2006-06-24 15:19  robert

	* libmp3lame/lame.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/version.h:
	  enabling the conditional RH_ATHAA_FIX code for vbr-new
	  we are now at patch level 6

2006-06-24 01:57  robert

	* libmp3lame/lame.c, libmp3lame/quantize_pvt.c:
	  tuning ATHAA_FIX for high quality VBR new presets

2006-06-23 23:40  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/vbrquantize.c:
	  tuning ATHAA_FIX for midrange VBR new presets

2006-06-23 23:09  robert

	* libmp3lame/machine.h:
	  always compile in assertions for alpha builds

2006-06-23 23:06  robert

	* Makefile.MSVC:
	  updating LibSndFile support, it's used by default now

2006-06-23 20:52  aleidinger

	* libmp3lame/version.h:
	  bump the patchlevel for the updated sndfile support

2006-06-23 20:49  aleidinger

	* doc/html/history.html:
	  document the updated libsndfile support

2006-06-23 20:46  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, configure, debian/Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in:
	  regen

2006-06-23 20:41  aleidinger

	* configure.in, frontend/Makefile.am, frontend/get_audio.c:
	  Add support for a recent sndfile library.
	  
	  Submitted by: Erik de Castro Lopo <erikd@mega-nerd.com> (sndfile
	  author)

2006-06-18 23:02  robert

	* libmp3lame/lame.c:
	  developer only switch "--tune x" mow modifies current mask_adjust
	  settings
	  instead of assigning x to it

2006-06-18 21:32  robert

	* Makefile.unix, libmp3lame/quantize.h:
	  adding RH_TEST_ATHAA_FIX to the Makefile config RH
	  matching declaration and implementation of
	  trancate_smallsspectrums

2006-06-18 20:49  robert

	* libmp3lame/bitstream.c:
	  fixing asserts

2006-06-18 20:48  robert

	* libmp3lame/psymodel.c:
	  bugfix: in NSPSY highpass filter, out of bounds access in fircoef
	  bugfix: loss of fraction in equal loudness weighting

2006-06-18 19:20  robert

	* libmp3lame/quantize_pvt.c:
	  small update to the ATHAA problem fix

2006-06-18 18:16  robert

	* Makefile.MSVC, libmp3lame/machine.h, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/reservoir.c:
	  trying to fix ATHAA problem for low volume files
	  to enable the code, one has to define RH_TEST_ATHAA_FIX at
	  compile time

2006-06-18 16:45  robert

	* frontend/gtkanal.c:
	  needs console.h

2006-06-18 16:44  robert

	* libmp3lame/bitstream.c:
	  fixed assertion

2006-06-16 00:16  robert

	* libmp3lame/quantize.c, libmp3lame/util.h, mpglib/interface.c:
	  few more include dependencies sorted

2006-06-15 23:51  robert

	* frontend/gtkanal.c, frontend/mp3x.c, libmp3lame/VbrTag.c,
	  libmp3lame/VbrTag.h, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/fft.c, libmp3lame/fft.h,
	  libmp3lame/gain_analysis.c, libmp3lame/gain_analysis.h,
	  libmp3lame/id3tag.c, libmp3lame/id3tag.h, libmp3lame/l3side.h,
	  libmp3lame/lame-analysis.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/lameerror.h,
	  libmp3lame/machine.h, libmp3lame/mpglib_interface.c,
	  libmp3lame/newmdct.c, libmp3lame/newmdct.h, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/set_get.c, libmp3lame/set_get.h, libmp3lame/tables.c,
	  libmp3lame/tables.h, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/version.c, libmp3lame/version.h, mpglib/interface.c,
	  mpglib/layer3.c, mpglib/mpglib.h:
	  some more header dependencies, indentation

2006-06-15 23:07  robert

	* libmp3lame/VbrTag.h, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/fft.c, libmp3lame/fft.h,
	  libmp3lame/gain_analysis.c, libmp3lame/gain_analysis.h,
	  libmp3lame/lame.c, libmp3lame/libmp3lame_vc6.dsp,
	  libmp3lame/machine.h, libmp3lame/mpglib_interface.c,
	  libmp3lame/newmdct.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/tables.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h, libmp3lame/version.c,
	  libmp3lame/version.h:
	  making arguments and variables const where possible

2006-06-15 16:03  robert

	* libmp3lame/VbrTag.c, libmp3lame/VbrTag.h, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/fft.c, libmp3lame/fft.h,
	  libmp3lame/gain_analysis.h, libmp3lame/id3tag.c,
	  libmp3lame/id3tag.h, libmp3lame/l3side.h,
	  libmp3lame/lame-analysis.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/machine.h,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/presets.c, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/quantize.c,
	  libmp3lame/quantize.h, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/reservoir.c,
	  libmp3lame/set_get.c, libmp3lame/set_get.h, libmp3lame/tables.c,
	  libmp3lame/tables.h, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/version.c:
	  tried to avoid includes within include headers to make
	  dependencies more obvious

2006-06-03 13:34  aleidinger

	* ChangeLog:
	  Add nearly a year of history.

2006-06-03 13:14  aleidinger

	* doc/html/history.html:
	  add todays history

2006-06-03 13:12  aleidinger

	* ltconfig, mkinstalldirs, stamp-h.in:
	  Remove obsolete files (at least after the use of the new
	  autotools,
	  they are not created anymore).

2006-06-03 13:11  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.am.global,
	  Makefile.in, aclocal.m4, config.h.in, configure, configure.in,
	  debian/Makefile.in, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.am, libmp3lame/i386/Makefile.in,
	  ltmain.sh, mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in:
	  Update to:
	  - libtool 1.5.22
	  - automake 1.9
	  - autoconf 2.59

2006-06-03 13:03  aleidinger

	* libmp3lame/lame.c:
	  - Fix bug introduced in last revision.
	  - Don't segfault in lame_close() in case of an error on
	  initialisation.

2006-06-03 11:41  aleidinger

	* libmp3lame/lame.c:
	  Plug some memory leaks in the error case.
	  
	  Noticed by: Christophe Fergeau <teuf@users.sf.net>
	  SF-Bug: 1443630

2006-06-03 11:05  aleidinger

	* lame.spec.in:
	  - Remove requirements for specific gcc versions, since modern
	  ones "just work".
	  - Remove out-dated hyper-optimizations (some of which weren't
	  valid compiler
	  flags anymore).
	  - Update to current RPM techniques and macros (inspired by
	  freshrpms.net spec).
	  
	  Submitted by: Kyle VanderBeek <kylev@kylev.com>
	  SF-Patch: 1488468

2006-05-04 19:50  bouvigne

	* libmp3lame/reservoir.c, libmp3lame/version.h:
	  restrict max mp3 buf to solve compatibility issue with FhG
	  directshow filter

2006-05-04 19:43  bouvigne

	* libmp3lame/psymodel.c:
	  attack detection: workaround against "low freq" periodic signals
	  (like trumpet)

2006-04-16 09:02  takehiro

	* configure.in:
	  better socket library configuration.

2006-04-16 07:06  takehiro

	* config.h.in, configure.in:
	  remove winsock check. It seems we do not need it anymore.

2006-04-08 07:12  takehiro

	* libmp3lame/id3tag.c:
	  fix genre tag spelling as reported by CrazySoul <hicrazysoul at
	  gmail com>

2006-04-08 07:08  takehiro

	* doc/html/id3.html:
	  fix spelling as reported by CrazySoul <hicrazysoul at gmail com>

2006-02-18 16:35  takehiro

	* frontend/get_audio.c:
	  remove redundant description.

2006-02-18 16:35  takehiro

	* frontend/main.c:
	  fix decoding problem when the output is stdout, reported by
	  john.edwards33
	  (sf.net bugid #1430719)

2006-02-03 06:52  vitaly-ivanov

	* dshow/README:
	  typo fix

2006-01-30 15:41  takehiro

	* frontend/get_audio.c:
	  fix error message when unsupported format AIFF file.
	  (related bugid #1414515,
	  http://sourceforge.net/tracker/index.php?func=detail&aid=1414515&group_id=290&atid=100290
	  )

2006-01-22 17:52  robert

	* include/lame.h, libmp3lame/encoder.c, libmp3lame/lame-analysis.h,
	  libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/util.h, libmp3lame/version.h:
	  bug fixes: psy code used long block ATH in short block case
	  workaround: quantization routines of new vbr code failed on
	  strong tonal
	  samples, due to over optimistic masking promises from psy model

2006-01-21 13:10  takehiro

	* libmp3lame/id3tag.c:
	  patch #1379187, "id3tag_write_v2 doesn't allocate enough memory
	  for track"

2005-12-21 21:41  robert

	* frontend/parse.c:
	  bug fix: implementation of the priority commandline switch did
	  not what
	  the documentation says (windows only)

2005-12-18 18:26  robert

	* frontend/brhist.c:
	  fixing display glitch with very long files

2005-12-05 15:57  takehiro

	* libmp3lame/id3tag.c:
	  fix id3v2 genre tag problem, reported by wylda@volny.
	  now LAME won't crash with "--id3v2-only --tg 14".

2005-11-29 21:21  robert

	* frontend/mp3rtp.c, frontend/mp3x.c:
	  commit was pending, needed for John33's decoding patch

2005-11-29 19:49  robert

	* doc/html/history.html:
	  updating history

2005-11-28 10:09  vitaly-ivanov

	* dshow/Encoder.h, dshow/Mpegac.cpp, dshow/Mpegac.h,
	  dshow/PropPage.cpp, dshow/PropPage.h, dshow/PropPage_adv.cpp,
	  dshow/PropPage_adv.h, dshow/Property.rc, dshow/README,
	  dshow/iaudioprops.h, dshow/resource.h:
	  Property page update - moved "set duration" check box to the
	  advanced page, added "allow sample overlap" check box (uncheck it
	  if downstream filter is Avi Mux, otherwise use it for better
	  video/audio synchronization - especially when capturing). Copying
	  terms update (README file)

2005-11-27 15:02  bouvigne

	* doc/html/history.html, doc/man/lame.1, frontend/parse.c,
	  libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/fft.c, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/presets.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/reservoir.c, libmp3lame/set_get.c,
	  libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c:
	  merge from 3.97 branch

2005-11-22 22:15  robert

	* frontend/console.c, frontend/console.h, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/gtkanal.c, frontend/lametime.c,
	  frontend/main.c, frontend/main.h, frontend/parse.c,
	  frontend/rtp.c:
	  John33's lame decoding fix

2005-11-20 14:24  takehiro

	* libmp3lame/gain_analysis.c:
	  fix built problem with Sun Studio 10, reported by Moinak Ghosh

2005-11-12 18:52  bouvigne

	* libmp3lame/gain_analysis.c:
	  speedup of silent parts analysis

2005-11-12 18:33  bouvigne

	* doc/html/history.html:
	  history from 3.97 branch

2005-11-12 18:21  bouvigne

	* Makefile.MSVC:
	  added console.c

2005-11-09 18:56  aleidinger

	* frontend/rtp.c:
	  Fix bug #1349332

2005-11-09 18:40  aleidinger

	* libmp3lame/id3tag.c:
	  Fix bug #1351021.

2005-11-05 16:35  aleidinger

	* doc/html/history.html:
	  add todays work

2005-11-05 16:20  aleidinger

	* libmp3lame/id3tag.c, libmp3lame/id3tag.h:
	  Add support for total number of tracks per album.
	  
	  Submitted by: Kyle VanderBeek <kylev@users.sourceforge.>
	  
	  While I'm here add some abort() calls in case we overflow some
	  arrays
	  (this is only compile tested, so pass me the pointy hat if it
	  fails).

2005-11-05 15:54  aleidinger

	* libmp3lame/id3tag.c:
	  Add id3v2 TLEN support.
	  
	  Submitted by: Linus Walleij <snirkel@users.sourceforge.>

2005-11-05 15:52  aleidinger

	* frontend/console.c:
	  some compilers really want the void nowadays...

2005-11-05 15:51  aleidinger

	* frontend/Makefile.in:
	  regen

2005-11-05 15:49  aleidinger

	* frontend/console.h:
	  make some pseudo prototypes to real prototypes (fixes gcc
	  warnings)

2005-11-05 15:48  aleidinger

	* frontend/Makefile.am:
	  console.[ch] is now mandatory, so build it

2005-11-05 15:47  aleidinger

	* include/lame.h:
	  remove redundant function prototypes (we need only one prototype)

2005-11-01 20:17  robert

	* frontend/console.c:
	  bug fix: FILE pointer may be NULL for DEBUGF

2005-11-01 13:54  robert

	* frontend/mp3rtp.c, frontend/rtp.c:
	  getting mp3rtp working again

2005-11-01 13:10  robert

	* frontend/mp3x.c:
	  replacing (f)printf by console_printf, error_printf and
	  report_printf. fixes the problem that some messages would be send
	  to stdout instead of stderr. added internal switch '--debug-file
	  <debugfile.txt>', allows to debug the code more comfortably by
	  writing DEBUGF(xyz) at code hot spots.

2005-11-01 13:01  robert

	* frontend/.indent.pro, frontend/amiga_mpega.c, frontend/brhist.c,
	  frontend/brhist.h, frontend/console.c, frontend/console.h,
	  frontend/get_audio.c, frontend/get_audio.h,
	  frontend/gpkplotting.c, frontend/gpkplotting.h,
	  frontend/gtkanal.c, frontend/gtkanal.h, frontend/lame_vc6.dsp,
	  frontend/lame_vc7.vcproj, frontend/lametime.c,
	  frontend/lametime.h, frontend/main.c, frontend/main.h,
	  frontend/mp3rtp.c, frontend/mp3x.c, frontend/mp3x_vc6.dsp,
	  frontend/mp3x_vc7.vcproj, frontend/parse.c, frontend/parse.h,
	  frontend/portableio.c, frontend/portableio.h, frontend/rtp.c,
	  frontend/rtp.h, frontend/timestatus.c, frontend/timestatus.h:
	  replacing (f)printf by console_printf, error_printf and
	  report_printf. fixes the problem that some messages would be send
	  to stdout instead of stderr. added internal switch '--debug-file
	  <debugfile.txt>', allows to debug the code more comfortably by
	  writing DEBUGF(xyz) at code hot spots.

2005-11-01 12:51  robert

	* Makefile.unix, configMS.h:
	  *** empty log message ***

2005-10-31 13:02  bouvigne

	* libmp3lame/quantize.c:
	  fixed bug 1227632

2005-10-31 13:01  bouvigne

	* libmp3lame/lame.c:
	  fixed bug 1337556

2005-10-31 13:00  bouvigne

	* frontend/main.c:
	  spelling of Takehiro's name

2005-10-10 07:06  rbrito

	* COPYING, debian/changelog, debian/copyright, debian/rules,
	  doc/man/lame.1:
	  Updated manpage to include accented characters;
	  Updated debian packaging;
	  Updated documentation to include the new FSF address.

2005-09-29 21:22  robert

	* libmp3lame/i386/fft3dn.nas:
	  taken from lame4 branch:
	  fix possible link problem on cygwin

2005-09-25 22:09  glessard

	* libmp3lame/quantize.c:
	  GCC 4 on OS X was choking on line 615.
	  Given the assignment on line 630, clearly 'work' isn't a pointer
	  to an array
	  of constants anyway.

2005-09-20 19:51  robert

	* libmp3lame/util.c:
	  fix wrong typing

2005-09-20 19:41  robert

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h:
	  moved loop from quantize.c to vbrquantize.c

2005-09-20 19:38  robert

	* libmp3lame/util.c, libmp3lame/util.h:
	  malloc with n-bytes alignment

2005-09-20 09:19  robert

	* frontend/parse.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  added internal developer switch to select between ISO and x^3/4
	  quantization
	  
	  --quantization-type n n=0: no change, n=1: ISO, n=2: x^3/4

2005-09-18 22:57  robert

	* libmp3lame/vbrquantize.c, libmp3lame/vbrquantize.h,
	  libmp3lame/version.h:
	  vbr new noise shaping: let lame abort in exceptional cases
	  
	  now on alpha 2

2005-09-18 21:38  robert

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize.h, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  vbr-new: bug fix for "searchGlobalStepSizeMax"
	  
	  demangling vbr_old and vbr_new code

2005-09-11 10:20  bouvigne

	* libmp3lame/version.h:
	  welcome to 3.98

2005-09-11 09:57  bouvigne

	* lame_vc7.sln:
	  missing file

2005-09-11 09:14  bouvigne

	* libmp3lame/version.h:
	  3.97b

2005-09-08 22:57  takehiro

	* doc/html/history.html:
	  fixed style.

2005-09-08 22:57  takehiro

	* doc/html/history.html:
	  added my changes for 3.97.

2005-09-08 02:44  rbrito

	* doc/man/lame.1:
	  Fixed typo.

2005-09-07 21:15  bouvigne

	* testcase.mp3:
	  updated testcase for 3.97b1

2005-09-06 18:24  bouvigne

	* doc/html/history.html:
	  typo

2005-09-05 21:24  bouvigne

	* doc/html/history.html, doc/html/index.html:
	  updated history for 3.97b

2005-09-05 20:49  bouvigne

	* frontend/parse.c:
	  -Vx short help

2005-09-05 04:29  rbrito

	* debian/changelog, debian/rules:
	  Update generation of package with full optimization, now that
	  compilation with GCC 4 works.

2005-09-04 19:52  aleidinger

	* ChangeLog:
	  Update with recent history.

2005-09-04 19:45  aleidinger

	* configure:
	  Regen.

2005-09-04 19:35  aleidinger

	* config.h.in:
	  Add the bits for largefile support.
	  
	  Submitted by: Diego "Flameeyes" Pettenò <flameeyes@gentoo.org>
	  Bug ID: 1271744

2005-09-04 19:11  aleidinger

	* configure.in:
	  Remove some flags which gcc 4.x doesn't understand anymore.
	  
	  Noticed by: Rogério Brito

2005-09-04 19:07  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, config.h.in, debian/Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in:
	  Regen.

2005-09-04 19:05  aleidinger

	* depcomp, frontend/depcomp, install-sh, libmp3lame/depcomp,
	  ltmain.sh, misc/depcomp, missing, mkinstalldirs, mpglib/depcomp:
	  Update to recent autotools.

2005-09-04 19:02  aleidinger

	* frontend/Makefile.am:
	  Play catchup with the autotools.

2005-09-04 18:54  aleidinger

	* configure.in:
	  Initialize autoconf correctly (with lame mailinglist address).

2005-09-04 11:52  rbrito

	* doc/man/lame.1:
	  Fixed typo in manpage.

2005-09-02 18:09  bouvigne

	* frontend/parse.c:
	  fixed a stupid bug with --preset medium/standard/extreme

2005-09-01 18:12  bouvigne

	* libmp3lame/VbrTag.c:
	  320kbps cbr was flagged as "non optimal" in the info header

2005-08-28 17:30  bouvigne

	* frontend/parse.c, libmp3lame/version.h:
	  Changed the way "--preset xxx" is handled,
	  as I've seen many users passing "--preset standard --vbr-new".
	  Presets are now evaluated during initialisation, so this kind of
	  command
	  line now properly works.

2005-08-28 17:02  bouvigne

	* Makefile.MSVC, frontend/parse.c, libmp3lame/lame.c,
	  libmp3lame/version.c, libmp3lame/version.h:
	  64bits build from within win64 platform sdk

2005-08-27 08:25  takehiro

	* configure, configure.in:
	  fix for configure problem reported by Steven P. Ulrick.

2005-08-25 12:14  takehiro

	* debian/Makefile.am, debian/Makefile.in:
	  follow the file name change.

2005-08-21 17:32  bouvigne

	* ACM/ADbg/ADbg_vc7.vcproj, ACM/ADbg/Makefile.am, ACM/Makefile.am,
	  ACM/lameACM_vc7.vcproj, ACM/tinyxml/Makefile.am,
	  ACM/tinyxml/tinyxml_vc7.vcproj, Dll/LameDll_vc7.vcproj,
	  Dll/Makefile.am, Makefile.am, frontend/Makefile.am,
	  frontend/lame_vc7.vcproj, frontend/mp3x_vc7.vcproj,
	  libmp3lame/Makefile.am, libmp3lame/libmp3lame_vc7.vcproj,
	  mpglib/Makefile.am, mpglib/mpglib_vc7.vcproj:
	  added project files for VC7

2005-08-20 17:48  bouvigne

	* doc/html/switchs.html:
	  typo

2005-08-16 07:56  rbrito

	* debian/changelog, debian/control:
	  Make libmp3lame0 provide and replace liblame0, not conflict with
	  it.

2005-08-16 07:11  rbrito

	* debian/changelog, debian/control, debian/libmp3lame0-dev.files:
	  Fixed development package to not include shared libraries.
	  Other misc fixes.
	  
	  Yet more to come.

2005-08-15 04:25  rbrito

	* debian/changelog, debian/control, debian/liblame0-dev.docs,
	  debian/liblame0-dev.files, debian/liblame0.files,
	  debian/libmp3lame0-dev.docs, debian/libmp3lame0-dev.files,
	  debian/libmp3lame0.files, debian/rules:
	  Update Debian packaging so that the library packages follow the
	  SONAME
	  rules.
	  
	  Also incorporate some changes from Christian Marillat's packages
	  in the
	  basic lame package.
	  
	  More changes to come soon.

2005-08-07 14:07  takehiro

	* doc/man/lame.1:
	  when the input is raw, the input file will be treated as native
	  endian.
	  (bugid 1158189)

2005-08-07 13:24  bouvigne

	* doc/html/switchs.html:
	  update of stereo modes description

2005-08-05 15:22  takehiro

	* doc/man/lame.1:
	  added "per sample" for the description of --bitwidth option.
	  (bugid #1158196)

2005-08-05 09:22  vitaly-ivanov

	* dshow/Mpegac.cpp, dshow/Mpegac.h:
	  Another timestamp fix. Added additional resynchronization points
	  (now there are 4 resynch points) in order to implement absolutely
	  accurate in-out synchronization

2005-08-04 01:42  rbrito

	* debian/changelog, debian/control, debian/rules:
	  Updates for Debian packaging.

2005-08-04 01:40  rbrito

	* doc/man/lame.1:
	  Documentation of switches: --{un,}signed and
	  --{big,little}-endian.

2005-08-01 13:10  vitaly-ivanov

	* dshow/Mpegac.cpp, dshow/Property.rc:
	  Output timestamps fix. There was a serious bug when input audio
	  samples had time gaps between each other - DirectShow timestamps
	  for output samples were calculated wrong in this case

2005-07-28 13:05  rbrito

	* doc/html/basic.html, doc/html/contributors.html,
	  doc/html/examples.html, doc/html/history.html, doc/html/id3.html,
	  doc/html/switchs.html:
	  Minor spelling corrections and other fixes to documentation.

2005-07-28 13:02  rbrito

	* debian/rules:
	  Safe version for compilation with GCC 4.

2005-07-27 21:49  rbrito

	* doc/html/presets.html, doc/html/switchs.html:
	  Spelling fixes for documentation.

2005-07-27 21:38  rbrito

	* doc/man/lame.1:
	  Include description of the "medium" preset in the manpage.
	  
	  Fix some spelling errors on the manpage.

2005-07-27 08:43  rbrito

	* debian/changelog, debian/control, debian/lame.manpages,
	  debian/rules, doc/man/lame.1:
	  Small fixes for lame manpage to make lintian happy. More updates
	  to come.

2005-07-27 07:41  rbrito

	* debian/changelog:
	  Preparation for the new beta release. More changes to come.

2005-07-10 16:13  bouvigne

	* libmp3lame/presets.c, libmp3lame/version.h:
	  alpha 11: update of abr/cbr 160-320

2005-06-29 12:26  vitaly-ivanov

	* dshow/Encoder.cpp, dshow/PropPage.cpp, dshow/PropPage_adv.cpp,
	  dshow/Property.rc:
	  Joint / Standard stereo selection fix. Got rid of automatic joint
	  stereo encoding when the bit rate is relatively low, now it's up
	  to the user

2005-06-25 08:15  vitaly-ivanov

	* dshow, dshow/Encoder.h, dshow/Property.rc, dshow/aboutprp.cpp:
	  *** empty log message ***

2005-06-24 10:04  vitaly-ivanov

	* dshow, dshow/Encoder.cpp, dshow/Encoder.h, dshow/Mpegac.cpp,
	  dshow/Mpegac.def, dshow/Mpegac.h, dshow/PropPage.cpp,
	  dshow/PropPage.h, dshow/PropPage_adv.cpp, dshow/PropPage_adv.h,
	  dshow/Property.rc, dshow/README, dshow/REG.CPP, dshow/UIDS.H,
	  dshow/aboutprp.cpp, dshow/aboutprp.h, dshow/dshow.dsp,
	  dshow/dshow.dsw, dshow/iaudioprops.h, dshow/resource.h:
	  *** empty log message ***

2005-06-24 09:37  vitaly-ivanov

	* dshow/REG.H:
	  *** empty log message ***

2005-05-29 16:10  takehiro

	* debian/changelog, debian/control:
	  fixed version number (upstream -> downstream)
	  
	  patch from Jack, from
	  http://sourceforge.net/tracker/index.php?func=detail&aid=1201172&group_id=290&atid=100290

2005-05-25 14:17  takehiro

	* frontend/get_audio.c, frontend/parse.c:
	  fixed output message even when "--silent"
	  (bug #1145932)

2005-05-25 13:56  takehiro

	* libmp3lame/lame.c:
	  fix format strings to print out the "float" value.
	  (bug #1207470)

2005-05-25 13:47  takehiro

	* frontend/parse.c:
	  removed -Z option description, which has no effects currently.

2005-03-20 17:28  bouvigne

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/presets.c,
	  libmp3lame/quantize.c, libmp3lame/quantize.h:
	  allow masking adjustements in ABR/CBR

2005-03-19 18:34  bouvigne

	* libmp3lame/version.h:
	  alpha 10

2005-03-19 17:51  bouvigne

	* libmp3lame/presets.c:
	  updated vbr 0/1/2/3 presets

2005-03-19 15:04  bouvigne

	* libmp3lame/quantize_pvt.c:
	  also use nssfb21 in short blocks

2005-03-19 15:03  bouvigne

	* libmp3lame/quantize.c:
	  typo

2005-03-19 14:44  bouvigne

	* libmp3lame/quantize.c:
	  updated psfb21 analog silence detection

2005-03-18 05:26  rbrito

	* debian/changelog, debian/control, debian/rules, doc/man/lame.1:
	  Slight modifications to debian packaging (with more to come).
	  Minor changes to the frontend manpage.

2005-03-15 00:26  robert

	* libmp3lame/vbrquantize.c:
	  yet another -Y thing: calc_xmin does not compute upto
	  max_nonzero_coeff
	  in the -Y case.

2005-03-14 22:32  robert

	* libmp3lame/vbrquantize.c:
	  bugfix for vbr-new when sub block gain violates IXMAXVAL
	  constrain in -Y case

2005-03-14 21:53  robert

	* libmp3lame/quantize_pvt.c:
	  removed an assert in calc_noise_core_c
	  cod_info->count1 == 0 seems to be a common case

2005-03-13 17:20  robert

	* libmp3lame/presets.c:
	  removing placebo -bx from fast presets
	  it's a placebo with vbr-old too, but keep it that way a little
	  bit longer

2005-03-13 17:01  robert

	* frontend/mp3x.c:
	  forgot to commit this one, removed unused parameter from usage
	  parameter list

2005-03-13 14:46  robert

	* libmp3lame/encoder.c, libmp3lame/set_get.c, libmp3lame/version.c,
	  libmp3lame/version.h:
	  removing unused ATH auto adjust types. JDs is now the only one

2005-03-13 14:43  robert

	* ACM/ACM.cpp:
	  version

2005-03-13 14:42  robert

	* frontend/main.c, frontend/parse.c, frontend/parse.h:
	  print helptext for internal switches only when internal switches
	  are available

2005-03-12 23:01  robert

	* libmp3lame/vbrquantize.c:
	  vbr-new: fixing the -Y bug, that sfb21 wasn't quantized at all

2005-03-12 17:39  bouvigne

	* libmp3lame/lame.c:
	  fix: q2 and higher were NOT enabling substep shaping 2

2005-03-12 17:38  bouvigne

	* mpglib/mpg123.h:
	  fix for replaygain-accurate

2005-03-08 23:07  robert

	* Makefile.unix, frontend/parse.c, libmp3lame/VbrTag.c,
	  libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/presets.c,
	  libmp3lame/set_get.c, libmp3lame/util.h:
	  removing some unused member of lame_internal_flags structure

2005-03-06 18:59  bouvigne

	* libmp3lame/takehiro.c:
	  faster noquant_count_bits

2005-03-06 16:56  bouvigne

	* doc/html/history.html:
	  history

2005-03-06 16:17  bouvigne

	* libmp3lame/presets.c, libmp3lame/version.h:
	  updated V3 and V2 presets

2005-03-05 18:26  bouvigne

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c, libmp3lame/vbrquantize.c:
	  slightly faster quantization (not totally sure if it is really
	  faster)
	  
	  Also fixed a severe bug in ISO quantization when not using the
	  IEEE hack

2005-03-05 13:45  bouvigne

	* configMS.h, libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/util.h:
	  update of the intrinsics coding style

2005-03-05 12:40  bouvigne

	* libmp3lame/quantize_pvt.c:
	  minor speed increase in calc_noise

2005-02-28 22:45  bouvigne

	* libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h:
	  calc_noise is back to the C version

2005-02-27 22:03  bouvigne

	* libmp3lame/quantize_pvt.c:
	  calc_noise:
	  fixed an assertion failure in frames with no big_values

2005-02-27 20:43  bouvigne

	* libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h:
	  sse version of calc_noise (speedup is very small)

2005-02-27 15:09  bouvigne

	* configMS.h, libmp3lame/quantize.c, libmp3lame/util.h:
	  slightly faster init_xrpow_core_sse

2005-02-26 06:21  takehiro

	* frontend/get_audio.c:
	  fix mp3 frame # estimation (bugid #934072).

2005-02-25 19:01  bouvigne

	* configMS.h:
	  intrinsics support for ICL

2005-02-25 01:21  robert

	* frontend/brhist.c, frontend/timestatus.c:
	  RH_HIST: don't print unused channel modes and block types

2005-02-24 00:27  robert

	* frontend/brhist.c:
	  fix scrolling problem I introduced with the last commit

2005-02-23 17:58  bouvigne

	* libmp3lame/quantize_pvt.c:
	  slightly faster calc_noise

2005-02-22 02:06  robert

	* libmp3lame/psymodel.c:
	  Compile time option to enable GPSYCHO like block type switching
	  in NSPSYTUNE.
	  Due to differences in convolution, the switching does not happen
	  exactly
	  as in GPSYCHO. (disabled by default, edit psymodel.c to activate)

2005-02-19 22:47  robert

	* frontend/brhist.c, frontend/brhist.h, frontend/main.c:
	  some statistics about block type distribution

2005-02-19 18:48  bouvigne

	* configMS.h:
	  conditionnal intrinsics support for MSVC (probably also works for
	  ICL)

2005-02-19 15:32  bouvigne

	* libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h:
	  SSE intrinsic version of init_xrpow
	  It is now using 1% of total time (using 4% in the standard
	  version)

2005-02-13 14:42  bouvigne

	* libmp3lame/psymodel.c:
	  minor speedup

2005-02-07 22:02  robert

	* configure.in, libmp3lame/Makefile.am, libmp3lame/Makefile.in:
	  fixing mingw32 configure problems

2005-02-06 23:05  robert

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h:
	  some MSVC warning level 4 issues resolved

2005-02-06 20:23  robert

	* libmp3lame/quantize.c, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c:
	  resolved compiler warnings

2005-02-06 19:49  robert

	* libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  lame.c: fixed lost break in lame_init_parms
	  reduced calloc/free calls (where called once per frame, now once
	  per run)
	  machine.h: removed IIPOW20
	  util.c, util.h: removed VBR_t, added in_buffer_[0/1] to
	  lame_internal_flags
	  vbrquantize.c: fixed some small rounding problem in quantize
	  routines

2005-02-05 14:14  robert

	* libmp3lame/vbrquantize.c:
	  eleminating dead code
	  expanding preprocessor defines
	  works probably not with GPsycho anymore, but NSPSYTUNE only
	  (G.not tested)
	  doesn't try to use 126 bits at minimum per granule anymore, less
	  bloating
	  on mono like signals where the difference signal is almost zero.

2005-02-05 13:32  bouvigne

	* libmp3lame/version.h:
	  bumped alpha

2005-02-05 13:31  bouvigne

	* libmp3lame/lame.c:
	  enabled ath adjustment for abr/cbr

2005-02-04 18:34  bouvigne

	* libmp3lame/lame.c:
	  sorry, I shouldn't have commited this one

2005-02-01 10:29  bouvigne

	* libmp3lame/bitstream.c, libmp3lame/lame.c, libmp3lame/takehiro.c:
	  reduced warnings

2005-01-30 22:31  robert

	* Makefile.unix, libmp3lame/vbrquantize.c:
	  scalefactor allocation should work now as expected

2005-01-30 22:27  robert

	* libmp3lame/machine.h, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  off by one bug:
	  we do access iipow20 by values in [0,Q_MAX2] inclusive,
	  that makes Q_MAX2+1 entries in total

2005-01-29 19:40  bouvigne

	* libmp3lame/presets.c:
	  updated vbr presets 9 to 4

2005-01-29 18:42  bouvigne

	* libmp3lame/takehiro.c:
	  fixed a crash

2005-01-29 18:11  bouvigne

	* libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c:
	  quantize_xrpow: also re-use previously computed data in short
	  blocks

2005-01-25 21:47  bouvigne

	* libmp3lame/quantize.c:
	  X9: more bits into reservoir

2005-01-23 16:25  bouvigne

	* libmp3lame/takehiro.c:
	  oops, quantize_xrpow optimization has been accidentally disabled
	  in november

2005-01-23 14:27  bouvigne

	* libmp3lame/lame.c:
	  reinstated old q1 (increase only 1 sf per iteration)

2005-01-22 18:15  bouvigne

	* libmp3lame/quantize_pvt.c:
	  problem with q1 and q0 finally found.
	  I am unable to trace it prior to "the big 3.94 merge".

2005-01-20 16:38  takehiro

	* libmp3lame/quantize.c:
	  to clarify the relationship of "if-then-else" and suppress
	  warnings from gcc,
	  added braces.

2005-01-13 19:18  bouvigne

	* libmp3lame/version.h:
	  bump alpha

2005-01-13 19:17  bouvigne

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.h:
	  use bits in quant_comp

2005-01-13 18:54  bouvigne

	* libmp3lame/presets.c:
	  updated ath values for abr/cbr

2005-01-13 18:20  bouvigne

	* frontend/gtkanal.c, libmp3lame/lame-analysis.h,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  Replaced X9 by the former X10 (ssd).
	  The abr/cbr modes are using this comparison

2005-01-10 09:23  takehiro

	* libmp3lame/quantize_pvt.c:
	  seems it cannot build on Linux and some Unix.
	  to fix this, I changed max() -> Max().

2005-01-09 17:31  bouvigne

	* frontend/gtkanal.c, libmp3lame/lame-analysis.h:
	  display SSD in mp3x

2005-01-09 17:31  bouvigne

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  do not compute noise if best quant has no distorted band and
	  current quant
	  is bigger than the best one

2005-01-08 11:47  bouvigne

	* libmp3lame/version.h:
	  bumped alpha

2005-01-08 11:46  bouvigne

	* libmp3lame/psymodel.c:
	  limited ATH effect in short blocks.
	  This is fundamentally wrong, but otherwise it produces bad
	  quality

2005-01-08 10:58  bouvigne

	* libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h:
	  updated VBR presets to match the cbr/abr changes
	  (also corrected the --cbr behavior)

2005-01-08 10:56  bouvigne

	* doc/html/history.html:
	  history

2004-12-31 12:07  takehiro

	* frontend/main.c:
	  merge 1.73.2.34 -> 1.73.2.35
	  fix it outputs CR even when --silent/--quiet
	  (bug #973302)

2004-12-31 11:26  takehiro

	* frontend/mp3rtp.c:
	  fix buffer overrun(same in my experimental branch).

2004-12-27 16:13  bouvigne

	* libmp3lame/lame.c:
	  cbr/abr: adjusted lowpass

2004-12-27 15:32  bouvigne

	* libmp3lame/presets.c:
	  abr/cbr: adjusted m/s and short/long

2004-11-29 23:41  takehiro

	* libmp3lame/psymodel.c:
	  fix ATHlower is always "0".

2004-11-28 19:10  bouvigne

	* libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c:
	  some FLOAT -> int

2004-11-28 18:21  takehiro

	* Dll/BladeMP3EncDLL.c:
	  added brancket to fix bugid#1003342 on sf.net

2004-11-17 19:54  bouvigne

	* libmp3lame/quantize.c, libmp3lame/util.h:
	  noise shaping type 3:
	  amplify bands within 50% of distortion,
	  then refine by amplifying the most distorted band

2004-10-24 14:26  bouvigne

	* libmp3lame/version.h:
	  bump alpha

2004-10-24 14:25  bouvigne

	* libmp3lame/lame.c:
	  reworked -q values:
	  noise_shaping_amp 2 removed
	  full Huffman search removed in q0
	  full outer loop search enabled in q0

2004-10-24 14:22  bouvigne

	* libmp3lame/quantize.c, libmp3lame/util.h:
	  full outer loop search:
	  allows to search for better quantization until global_gain==255

2004-08-19 19:49  bouvigne

	* doc/html/history.html:
	  added 3.96.1

2004-08-19 19:36  bouvigne

	* doc/html/modes.html:
	  spelling, clarifications

2004-07-21 18:46  bouvigne

	* ACM/ACM.cpp:
	  fix by Dmitriy Startsev

2004-07-11 18:12  bouvigne

	* ACM/acm.rc:
	  updated version number because of recent fixes

2004-07-11 18:11  bouvigne

	* ACM/ACMStream.cpp:
	  fixed a potential crash

2004-07-11 18:05  bouvigne

	* ACM/ACM.cpp:
	  Should fix the display issue under win95

2004-07-11 17:11  bouvigne

	* ACM/ACM.cpp:
	  fix crash when used through directshow.
	  It still does not work in this configuration, but at least it
	  does not crash anymore.

2004-06-27 13:02  bouvigne

	* libmp3lame/psymodel.c:
	  fixed short block detection using mid and side channels

2004-06-26 15:35  bouvigne

	* libmp3lame/encoder.c:
	  lame_encode_frame_init

2004-06-20 17:28  bouvigne

	* doc/html/history.html, libmp3lame/lame.c:
	  restored use of padding when not using bit reservoir

2004-06-18 19:15  bouvigne

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/l3side.h, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h, libmp3lame/version.h:
	  FLOAT8 -> FLOAT

2004-06-06 15:31  bouvigne

	* frontend/gpkplotting.c, frontend/gtkanal.c:
	  more explicit display of noise values

2004-06-06 14:15  bouvigne

	* frontend/gtkanal.c:
	  display more scalefactor band splitting lines.
	  (as in exp branch)

2004-06-06 09:48  bouvigne

	* doc/html/history.html:
	  history

2004-06-05 15:46  bouvigne

	* libmp3lame/quantize.c:
	  small speedup: init_xrpow only up to max non-zero coeff

2004-05-31 17:35  bouvigne

	* libmp3lame/takehiro.c:
	  reduced the number of calls to quantization function

2004-05-27 23:55  robert

	* libmp3lame/version.h:
	  now at 3.97 alpha 2

2004-05-27 23:52  robert

	* libmp3lame/takehiro.c, libmp3lame/vbrquantize.c:
	  bug fix for lame crashing occasionally when using --vbr-new
	  PS: scalefactor allocation is still not optimal

2004-05-26 21:37  bouvigne

	* libmp3lame/takehiro.c:
	  fixed a stupid bug

2004-05-26 21:04  bouvigne

	* libmp3lame/takehiro.c:
	  fixed a bug in quantize_xrpow, but it is still crashing
	  Oddly, cbr/abr are working fine, and low vbr settings are also
	  working fine.
	  -V2 is not working

2004-05-23 13:16  bouvigne

	* libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c, libmp3lame/util.h:
	  Use function pointer to select between quantization methods.
	  This should ease plugin of vectorized versions

2004-05-23 12:47  bouvigne

	* libmp3lame/takehiro.c:
	  factorized code which is selecting which lines to quantize

2004-05-23 12:43  bouvigne

	* configMS.h:
	  version

2004-05-14 19:29  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/ddk/Makefile.in,
	  ACM/tinyxml/Makefile.in, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, configure, debian/Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in:
	  regen with new auto* version

2004-05-14 19:20  aleidinger

	* depcomp, frontend/depcomp, install-sh, libmp3lame/depcomp,
	  missing, mpglib/depcomp:
	  update to a new version

2004-05-14 19:12  aleidinger

	* testcase.mp3:
	  update it to have a more sane number

2004-05-14 19:11  aleidinger

	* Makefile.am:
	  Tell the users that they do not have to care about the output of
	  make test.

2004-04-25 12:33  bouvigne

	* frontend/gtkanal.c:
	  fixed mp3x monopolizing processor in its idle loop

2004-04-25 06:13  takehiro

	* libmp3lame/psymodel.c:
	  fix bug id 921798
	  http://sourceforge.net/tracker/index.php?func=detail&aid=921798&group_id=290&atid=100290
	  
	  This is reported as bug for "BCC", but it may happen for every
	  compiler.

2004-04-24 18:29  olcios

	* libmp3lame/gain_analysis.h:
	  changed some constants' types to long

2004-04-24 17:47  olcios

	* libmp3lame/gain_analysis.c, libmp3lame/gain_analysis.h:
	  A better fix for the FP x86 problem. Provided by Andrew Church.
	  Modified by olcios.

2004-04-14 22:15  robert

	* frontend/get_audio.c, mpglib/common.c, mpglib/common.h,
	  mpglib/dct64_i386.c, mpglib/decode_i386.c, mpglib/interface.c,
	  mpglib/layer1.c, mpglib/layer2.c, mpglib/layer2.h,
	  mpglib/layer3.c, mpglib/layer3.h, mpglib/mpg123.h,
	  mpglib/mpglib.h, mpglib/tabinit.c:
	  fixing the decoding issue with the reported "Christmas" song.
	  I've taken Takehiro's modifications and getting the same output
	  as he does on his 4.0 branch. Imho this is not the best solution
	  to the given problem, but it works for now.

2004-04-12 12:29  aleidinger

	* debian/lame.docs:
	  PRESETS.draft does not exists anymore.

2004-04-11 15:39  glessard

	* frontend/main.c, include/lame.h, libmp3lame/set_get.c:
	  Added two sets of set/get functions to the API. They modify and
	  read the
	  nogap_total and nogap_current variables of the internal_flags
	  struct.
	  Added calls to these functions to the gapless-encoding loop of
	  the frontend.

2004-04-11 15:06  aleidinger

	* ChangeLog:
	  update

2004-04-11 15:04  aleidinger

	* configure:
	  regen

2004-04-11 15:00  aleidinger

	* config.h.in:
	  Add some missing LARGEFILE pieces. This patch is also in the
	  generated
	  lame-3.96 tarball, it just enters the CVS repository for 3.97 for
	  organizational reasons.

2004-04-11 14:58  aleidinger

	* configure.in:
	  Welcome to lame 3.97.

2004-04-11 13:05  bouvigne

	* libmp3lame/version.h:
	  3.97a1

2004-04-11 13:03  bouvigne

	* libmp3lame/version.h:
	  release 3.96

2004-04-11 13:03  bouvigne

	* doc/html/history.html:
	  history

2004-04-11 13:02  bouvigne

	* debian/changelog:
	  updated debian changelog

2004-04-11 13:01  bouvigne

	* libmp3lame/presets.c:
	  use X9 to select abr/cbr quantization,
	  use sfscale up to 160kbps

2004-04-11 12:50  bouvigne

	* Makefile.in:
	  manually updated Makefile.in

2004-04-03 17:28  bouvigne

	* frontend/gtkanal.c, libmp3lame/lame-analysis.h,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  Add X9 quantization selection method.
	  It increases applaud.wav substancially.
	  Not enabled by default

2004-03-28 13:11  bouvigne

	* libmp3lame/version.h:
	  3.96 beta 2

2004-03-28 13:10  bouvigne

	* doc/html/history.html:
	  history

2004-03-28 13:10  bouvigne

	* libmp3lame/presets.c:
	  updated minimal bitrate for V1 and V2

2004-03-23 22:17  robert

	* Makefile.unix, libmp3lame/vbrquantize.c:
	  rewritten vbr-new
	  
	  the scalefactor allocation functions take into account now, that
	  some
	  minimal scalefactor has to be used (the less IXMAX_VAL case)
	  
	  the output of vbr-new is not bit identical to previous ones, but
	  the
	  code is much more straight forward now.

2004-03-23 22:11  robert

	* libmp3lame/util.c:
	  patch for resampling code, found by Takehiro
	  
	  this patch removes some unneccessary integer rounding
	  resulting mp3s where bit identical, but resampling is faster now

2004-03-23 03:07  olcios

	* frontend/parse.c:
	  fixed a compile-time issue in --longhelp

2004-03-23 01:57  olcios

	* USAGE, doc/html/switchs.html, doc/man/lame.1, frontend/main.c,
	  frontend/parse.c, include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c,
	  libmp3lame/util.h:
	  1. added --noreplaygain switch that disables RG 2. added
	  --replaygain-fast switch that complements --replaygain-accurate.
	  The new switch is enabled by default in the frontend. 3. Made RG
	  analysis no longer default in libmp3lib. RG can still be fully
	  controlled via the libmp3lib interface, but for other interfaces,
	  like DLL or ACM, that provide no RG control, RG analysis is now
	  _disabled_. 4. updated documentation

2004-03-19 22:09  olcios

	* libmp3lame/gain_analysis.c:
	  changed the pre-processor condition related to the x86's extended
	  FP bug when compiled by GCC with optimizations

2004-03-19 18:18  olcios

	* libmp3lame/gain_analysis.c:
	  improved comment

2004-03-19 13:17  olcios

	* libmp3lame/gain_analysis.c:
	  workaround for the gcc ceil() bug when compiled with
	  optimizations

2004-03-18 22:00  olcios

	* libmp3lame/gain_analysis.h:
	  fixed compile-time problem with uint32_t on gcc

2004-03-18 14:27  aleidinger

	* libmp3lame/gain_analysis.c, libmp3lame/gain_analysis.h:
	  Replace Uint32_t and Int32_t with uint32_t and int32_t. There are
	  only few
	  instances of them used, so there's no need to typedef new types
	  for them.
	  
	  I hope this fixes problems on OSX reported on lame-dev.
	  
	  While I'm here, fix a cast (rgData->sampleWindow is long, not
	  int).

2004-03-17 18:56  aleidinger

	* libmp3lame/bitstream.c, libmp3lame/gain_analysis.c,
	  libmp3lame/gain_analysis.h, libmp3lame/lame.c:
	  - remove config.h (prerequisite) include from gain_analysis.h
	  - make sure prerequisites get included in every file before
	  gain_analysis.h

2004-03-14 18:22  bouvigne

	* libmp3lame/gain_analysis.h:
	  solved the potential data type problem

2004-03-11 23:43  olcios

	* libmp3lame/bitstream.c:
	  fixed a bug related to the usage of mpglib when decoding on the
	  fly

2004-03-11 19:35  olcios

	* libmp3lame/lame.c:
	  fixed compile-time problem when mpglib is disabled

2004-03-10 20:49  robert

	* doc/html/contributors.html, doc/html/history.html:
	  history

2004-03-10 00:50  robert

	* mpglib/interface.c:
	  patch for freeformat decoding bug

2004-03-09 19:04  bouvigne

	* libmp3lame/lame.c:
	  typo

2004-03-09 00:35  robert

	* libmp3lame/vbrquantize.c:
	  found a sample which did not give bit identical result with the
	  previous
	  patch. found out why and corrected it, so now the results for
	  find_scalefac_ave_[x34|ISO] are identical to the original code,
	  but --preset fast standard is ~7 percent faster now.

2004-03-08 00:54  robert

	* libmp3lame/vbrquantize.c:
	  faster "find_scalefac_ave_[x34|ISO]":
	  this patch gave bit identical results on all tested samples
	  compared to
	  previous version, so I'm pretty much confident that I don't
	  introduce
	  a new bug here.
	  the speeding up of this functions results in an overall speed
	  increase of
	  almost 7 percent for --preset fast standard

2004-03-07 23:55  robert

	* Makefile.MSVC, Makefile.unix:
	  in case of single precision (FLOAT8==float) we have to make sure
	  that
	  the mpglib uses REAL==float too. otherwise the decoder will moan
	  about
	  too small buffers

2004-03-06 15:36  bouvigne

	* libmp3lame/version.h:
	  entering beta stage

2004-03-06 15:34  bouvigne

	* doc/html/history.html:
	  history

2004-03-01 22:21  robert

	* configMS.h, libmp3lame/version.h:
	  we are now at 3.96 alpha 4

2004-03-01 22:20  robert

	* libmp3lame/psymodel.c:
	  bug fix: the highly optimized loops did an off by one access
	  resulting
	  in often invalid energy and threshold calculations. (both psy
	  models
	  affected). added a workaround and some assertions.

2004-02-29 19:55  robert

	* libmp3lame/bitstream.c, libmp3lame/takehiro.c:
	  fixing the "huffman decode overrun" and "mp3x problem"

2004-02-29 18:20  bouvigne

	* doc/html/contributors.html, doc/html/history.html:
	  no message

2004-02-29 17:17  takehiro

	* mpglib/layer1.c:
	  fix layer1 decoding bug

2004-02-26 17:20  takehiro

	* frontend/get_audio.c:
	  fixed Layer2 decoding bug reported by Goran Markovic

2004-02-25 22:10  robert

	* libmp3lame/util.c:
	  fixing some aliasing problem in fastlog2 calculation

2004-02-25 21:46  robert

	* libmp3lame/vbrquantize.c:
	  getting rid of a warning about a function being defined but not
	  used

2004-02-25 21:46  robert

	* libmp3lame/gain_analysis.c:
	  fixing code relying on compiler implementation details

2004-02-25 21:41  robert

	* frontend/parse.c, libmp3lame/VbrTag.c, libmp3lame/lame.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/set_get.c, libmp3lame/util.c:
	  getting rid of some compiler warnings:
	  - about too long string constants
	  - about C++ stylish comments
	  there is still a warning about the use of a C99 hexadecimal
	  floating
	  constant in portableio.c. will this compile on older C-compilers?

2004-02-24 01:57  olcios

	* libmp3lame/lame.c:
	  fixed a bug with --resample and ReplayGain analysis of input
	  data. With certain parameters the data sent to analysis was
	  mangled (some samples missing, some doubled, some in the wrong
	  order). With most natural music the error wasn't vast (below
	  1dB), but with e.g. a sine wave the gain error was audible.

2004-02-23 19:48  bouvigne

	* libmp3lame/gain_analysis.h:
	  now works under gcc, but I am not sure about the typedefs on
	  different platforms.

2004-02-23 18:53  bouvigne

	* libmp3lame/gain_analysis.h, libmp3lame/util.h:
	  moved replaygain_t to gain_analysis.h

2004-02-22 21:17  bouvigne

	* libmp3lame/bitstream.c, libmp3lame/gain_analysis.c,
	  libmp3lame/gain_analysis.h, libmp3lame/lame.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  made ReplayGain reentrant

2004-02-22 21:16  bouvigne

	* libmp3lame/quantize_pvt.c:
	  yet another "usefullsize" bug (found by passing the python tests)

2004-02-22 21:14  bouvigne

	* test/CBRABR.op, test/VBR.op, test/misc.op, test/shortCBRABR.op,
	  test/shortVBR.op:
	  Updated test files.
	  As the cbr/abr and vbr ones are quite long, I also added some
	  short ones.

2004-02-17 00:37  robert

	* libmp3lame/lame.c:
	  bug fix: in mono mode all block type where of short type due to
	  forced
	  block type coupling

2004-02-16 21:51  robert

	* libmp3lame/gain_analysis.c:
	  bug fix by John Edwards (48 kHz replay gain segmentation fault)

2004-02-15 21:39  robert

	* libmp3lame/lame.c, libmp3lame/vbrquantize.c:
	  standard ISO quantization for vbr-mtrh

2004-02-15 19:05  robert

	* libmp3lame/lame.c, libmp3lame/takehiro.c:
	  bug fix: usefullsize may become negativ due to some optimizations

2004-02-15 01:20  robert

	* libmp3lame/takehiro.c:
	  a simple -V9 -f line lets LAME crash,
	  there seems to be something wrong with scalefactor width
	  at least my inserted assertions indicate this

2004-02-10 02:04  robert

	* libmp3lame/vbrquantize.c:
	  try to make some use of max_nonzero_coeff in vbr-mtrh

2004-02-08 19:00  bouvigne

	* libmp3lame/presets.c:
	  reduced the number of short blocks

2004-02-08 14:08  bouvigne

	* Makefile.am, lame.dsw:
	  removed old workspace

2004-02-08 12:44  bouvigne

	* libmp3lame/quantize_pvt.c, libmp3lame/takehiro.c:
	  inverted a while statement, some int ->unsigned.
	  I hope that they will be safe this time.

2004-02-08 03:10  robert

	* libmp3lame/takehiro.c:
	  sorry Gabriel, reverting your signed->unsigned patch due to
	  exceptions
	  while encoding awe32_20sec.wav using --preset standard

2004-02-08 03:06  robert

	* libmp3lame/quantize_pvt.c:
	  sorry Gabriel, reverting your signed->unsigned patch because of
	  exceptions in awe32_20sec.wav using --preset standard

2004-02-08 03:01  robert

	* libmp3lame/vbrquantize.c:
	  bug fix: scalefactors may have been read from uninitialized
	  memory

2004-02-08 02:12  robert

	* libmp3lame/takehiro.c, libmp3lame/vbrquantize.c:
	  calculating xrpow_max for vbr_mtrh too

2004-02-07 19:38  bouvigne

	* libmp3lame/quantize_pvt.c, libmp3lame/takehiro.c:
	  some int -> unsigned int

2004-02-06 16:56  takehiro

	* frontend/get_audio.c, mpglib/interface.c:
	  mp3 decoding patch from Dmitriy Startsev

2004-02-04 20:17  bouvigne

	* libmp3lame/version.h:
	  bump alpha because of the fixed crash

2004-02-04 20:16  bouvigne

	* libmp3lame/takehiro.c:
	  fixed a crash in quantize_xrpow occuring when using low bitrates
	  (32kbps)

2004-02-02 15:20  olcios

	* USAGE, doc/html/switchs.html:
	  reworded the RG part. Discussed with: Alexander Leidinger

2004-02-02 14:17  olcios

	* testcase.mp3:
	  oops, committed an experimental file, reverting back to 1.80

2004-02-02 13:28  aleidinger

	* configure:
	  regen

2004-02-02 13:25  aleidinger

	* configure.in:
	  Add a check for 64 bit off_t's ("LARGEFILE" support).
	  
	  This may not be enough as lame uses ints and longs where other
	  types
	  may be more correct, but it may work in some cases.
	  
	  HOWTO provided by: Dan Nelson <dnelson@allantgroup.com>

2004-02-02 13:02  aleidinger

	* doc/man/lame.1:
	  - reformatted the replaygain part (less than 80 characters per
	  line)
	  - reworded some parts
	  - add a note about the decoder requierement for the --decode,
	  --replaygain-accurate and --clipdetect options
	  
	  Discussed with: Aleksander Korzynski

2004-02-02 12:59  aleidinger

	* config.h.in, configure:
	  regen

2004-02-02 12:57  aleidinger

	* configMS.h:
	  Add the DECODE_ON_THE_FLY define in the HAVE_MPGLIB case.

2004-02-02 12:57  aleidinger

	* configure.in, include/lame.h:
	  - enable bitrate histogram by default
	  discussed with Gabriel in Sept. 03
	  
	  - define DECODE_ON_THE_FLY for the replaygain feature in the
	  HAVE_MPGLIB
	  case too (moved from include/lame.h, I think it's cleaner this
	  way)
	  discussed with Aleksander Korzynski

2004-02-02 12:38  aleidinger

	* libmp3lame/id3tag.c:
	  Patch from Ti Kan <ti@amb.org>:
	  ---snip---
	  Below is a patch to make Lame 3.95.1 compile and run on AIX,
	  HP-UX,
	  IRIX, Solaris, UnixWare, SCO Open Server and SunOS 4.1.x.
	  
	  The change involves the call to snprintf() in
	  libmp3lame/id3tag.c.,
	  but snprintf() is not available on many of these platforms. I
	  know
	  that in general replacing snprintf() with sprintf() removes a
	  layer
	  of string buffer overflow protection, but in this case it should
	  be safe because the size of string is bounded.
	  ---snip---

2004-01-31 20:05  olcios

	* frontend/main.c, libmp3lame/lame.c, testcase.mp3:
	  disabled peak sample detection when LAME tag is not written into
	  the file

2004-01-29 16:39  bouvigne

	* doc/html/history.html:
	  history

2004-01-29 16:31  bouvigne

	* libmp3lame/version.h:
	  bump alpha because of the fixed crashes

2004-01-29 16:26  bouvigne

	* libmp3lame/takehiro.c:
	  fixed bug #886989

2004-01-26 14:00  olcios

	* USAGE, doc/html/switchs.html, doc/man/lame.1, frontend/parse.c:
	  updated documentation and --longhelp on RG

2004-01-26 12:05  olcios

	* frontend/main.c, frontend/main.h, frontend/parse.c,
	  include/lame.h, libmp3lame/VbrTag.c, libmp3lame/VbrTag.h,
	  libmp3lame/bitstream.c, libmp3lame/lame.c, libmp3lame/set_get.c,
	  libmp3lame/util.h:
	  removed ReportLameTagProgress (it is not LAME API). Fixed
	  --replaygain-accurate and --clipdetect.

2004-01-21 13:50  aleidinger

	* ACM/Makefile.in, Dll/Makefile.in, Makefile.in,
	  frontend/Makefile.in, libmp3lame/Makefile.in, mpglib/Makefile.in:
	  regen

2004-01-21 13:48  aleidinger

	* frontend/main.h:
	  only include limits.h if it is present on the system

2004-01-21 10:17  bouvigne

	* libmp3lame/quantize.c:
	  missing {} in last fix

2004-01-21 09:41  bouvigne

	* ACM/ACM.dsw, ACM/Makefile.am, ACM/lameACM.dsp, Dll/Example.dsp,
	  Dll/Example.dsw, Dll/MP3EncDll.dsp, Dll/MP3EncDll.dsw,
	  Dll/Makefile.am, Makefile.B32, Makefile.DJGPP, Makefile.am,
	  PRESETS.draft, README.B32, README.DJGPP, frontend/Makefile.am,
	  frontend/lame.dsp, frontend/lame.dsw, frontend/mp3x.dsp,
	  frontend/mp3x.dsw, libmp3lame/Makefile.am,
	  libmp3lame/libmp3lame.dsp, mpglib/Makefile.am, mpglib/mpglib.dsp:
	  removed old unmaintained files

2004-01-20 18:41  bouvigne

	* libmp3lame/quantize.c:
	  fixed a crash:
	  sometimes quantization could overflow because of big xr_pow
	  values

2004-01-15 19:05  aleidinger

	* frontend/parse.c:
	  We should not need errno... introduced in the last commit, remove
	  it.

2004-01-15 19:04  aleidinger

	* frontend/parse.c:
	  Different fix for the priority parsing bug, uses less code.
	  
	  Tested by: Gabriel

2004-01-15 13:54  bouvigne

	* ACM/TODO, doc/html/history.html, doc/html/index.html,
	  doc/html/switchs.html:
	  doc, history

2004-01-15 11:15  bouvigne

	* include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  made lame.h identical with or without DECODE_ON_THE_FLY.
	  made gfp identical with or without DECODE_ON_THE_FLY.

2004-01-15 01:24  olcios

	* USAGE, frontend/parse.c, libmp3lame/lame.c:
	  changed the behaviour of --clipdetect to force RG analysis on
	  decoded data (not input data). Updated --longhelp and
	  documentation

2004-01-14 14:38  bouvigne

	* libmp3lame/psymodel.c:
	  minor code change in mask_add

2004-01-14 14:16  bouvigne

	* libmp3lame/psymodel.c:
	  minor code change in mask_add

2004-01-13 19:23  bouvigne

	* frontend/parse.c:
	  isnumber ->is_number

2004-01-13 18:35  glessard

	* frontend/get_audio.c:
	  Removed a spurious fprintf that occurred during AIFF-C decoding.

2004-01-13 17:36  bouvigne

	* USAGE:
	  ReplayGain ->89dB

2004-01-13 17:30  takehiro

	* frontend/get_audio.c:
	  fix mp3 decoding...

2004-01-13 13:30  bouvigne

	* frontend/parse.c:
	  fixed issue when --priority is not followed by a number but by
	  another
	  argument.
	  
	  Btw, the situation is the same with a lot of other settings.

2004-01-13 12:36  bouvigne

	* libmp3lame/version.h:
	  3.96a1

2004-01-12 09:03  bouvigne

	* doc/html/history.html:
	  history

2004-01-12 08:59  bouvigne

	* libmp3lame/gain_analysis.c:
	  changed RG level to 89dB

2004-01-12 08:42  bouvigne

	* libmp3lame/version.h:
	  back into 3.95 because of the fixed crash

2004-01-12 08:41  bouvigne

	* libmp3lame/takehiro.c:
	  fixed a potential crash with fast vbr (introduced by myself)

2004-01-11 15:01  bouvigne

	* libmp3lame/lame.c:
	  small cleanup in the lowpass/samplefreq selection code

2004-01-11 11:22  aleidinger

	* configure:
	  regen

2004-01-11 11:20  aleidinger

	* configure.in:
	  Welcome to 3.96.

2004-01-11 10:21  bouvigne

	* libmp3lame/version.h:
	  We are now in 3.96

2004-01-10 21:20  olcios

	* doc/html/history.html:
	  cosmetic

2004-01-10 17:28  olcios

	* USAGE:
	  cosmetic change

2004-01-10 16:14  takehiro

	* libmp3lame/lame.c:
	  fix the display of lowpass fileter config.

2004-01-10 12:29  aleidinger

	* testcase.mp3:
	  testcase for 3.95

2004-01-10 12:29  takehiro

	* doc/html/history.html:
	  added mpglib fix

2004-01-10 12:18  aleidinger

	* ChangeLog:
	  update CVS history

2004-01-10 12:13  aleidinger

	* libmp3lame/version.h:
	  Preperation for 3.95 RELEASE.
	  
	  Please notify me and Gabriel in case you change something from
	  this
	  point in time upto the 3.96 version bump.

2004-01-10 12:10  aleidinger

	* doc/html/history.html:
	  faster on PPC

2004-01-10 12:03  aleidinger

	* libmp3lame/util.h:
	  do not use FastLog implicitely

2004-01-10 12:02  aleidinger

	* configMS.h:
	  explicit use of FastLog

2004-01-10 11:42  aleidinger

	* configure:
	  regen

2004-01-10 11:35  aleidinger

	* configure.in:
	  - use the IEEE 754 hack and FastLog on x86 and PPC only
	  - make the use of FastLog explicit
	  
	  PPC performance:
	  1GHz G4
	  "Standard" : 396 s (100%)
	  FastLog : 371 s ( 93%)
	  FastLog, 754 : 365 s ( 92%)
	  754 : 388 s ( 98%)
	  
	  2GHz G5 (dual), cbr, vbr
	  "Standard" : 166 s (100%), 328 s (100%)
	  FastLog : 169 s (102%), 338 s (102%)
	  FastLog, 754 : 148 s ( 90%), 288 s ( 88%)
	  754 : 145 s ( 87%), 279 s ( 85%)
	  
	  On a G5 we shouldn't use FastLog (I assume highly optimized for
	  G5 log*()
	  routines in the standard math library), but configure doesn't
	  know ATM how
	  do check for a G5.

2004-01-10 10:38  takehiro

	* frontend/get_audio.c:
	  backport from my experimental branch to fix the crash when
	  invalid input.

2004-01-10 10:27  takehiro

	* mpglib/common.c, mpglib/interface.c, mpglib/layer1.c,
	  mpglib/layer3.c:
	  backport from my experimenhtal branch, to fix the crash when
	  invalid input.

2004-01-10 09:55  takehiro

	* doc/html/history.html:
	  effetc->affect reported at sf.net (BugID #867502)

2004-01-08 19:55  olcios

	* libmp3lame/gain_analysis.c:
	  comment update (clarification)

2004-01-08 19:14  olcios

	* USAGE:
	  Removed --replaygain from USAGE, because it had been defaulted
	  and the option removed from LAME. Updated information on
	  --replaygain-accurate.

2004-01-08 15:58  bouvigne

	* configMS.h, doc/html/history.html, doc/html/index.html:
	  predicted release date: January 11th

2004-01-07 17:58  aleidinger

	* frontend/Makefile.in:
	  regen

2004-01-07 17:57  aleidinger

	* frontend/Makefile.am:
	  Add mp3x_vc6.dsp.
	  
	  Requested by: Gabriel

2004-01-07 17:35  bouvigne

	* README.WINGTK, frontend/lame_vc6.dsp, frontend/mp3x_vc6.dsp,
	  lame_vc6.dsw, libmp3lame/libmp3lame_vc6.dsp,
	  mpglib/mpglib_vc6.dsp:
	  added back mp3x to the VC6 projects

2004-01-07 15:57  aleidinger

	* configure:
	  regen

2004-01-07 15:51  aleidinger

	* configure.in:
	  - bump version to 3.95
	  - disable layer1 decoding by default
	  - reword the layer options and some option descriptions
	  - enable the use of the fast log function on every CPU (for x86
	  this
	  is implicit by the ieee hack)

2004-01-07 15:26  aleidinger

	* config.h.in:
	  add the fast log option

2004-01-07 13:26  bouvigne

	* mpglib/mpglib_vc6.dsp:
	  disable layer I

2004-01-06 14:38  bouvigne

	* libmp3lame/version.h:
	  bump alpha

2004-01-06 13:53  bouvigne

	* libmp3lame/VbrTag.c:
	  do not include the vbr header into the vbr seek table as it
	  breaks a player

2004-01-05 16:25  bouvigne

	* doc/html/history.html:
	  history

2004-01-05 16:19  bouvigne

	* ACM/ACMStream.cpp:
	  fixed a buffer requirement error

2004-01-05 09:45  bouvigne

	* libmp3lame/l3side.h, libmp3lame/quantize.c,
	  libmp3lame/takehiro.c:
	  faster count_bits

2004-01-04 11:04  bouvigne

	* libmp3lame/lame.c:
	  corrected -k

2004-01-03 16:47  bouvigne

	* doc/html/history.html:
	  history

2004-01-02 19:53  bouvigne

	* libmp3lame/lame.c:
	  changed quality mapping

2004-01-02 19:51  bouvigne

	* libmp3lame/takehiro.c:
	  faster count_bits

2004-01-02 19:49  bouvigne

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c, libmp3lame/vbrquantize.c:
	  Faster quantizations
	  Only 5% speedup when using Takehiro's ieee754 hack, but about 60%
	  speedup otherwise. (Should please Mac users)

2003-12-22 16:36  bouvigne

	* doc/html/history.html:
	  history

2003-12-22 09:44  bouvigne

	* libmp3lame/quantize_pvt.c:
	  reduced warnings

2003-12-21 18:41  bouvigne

	* frontend/get_audio.c, frontend/main.c, frontend/parse.c:
	  reduced warnings

2003-12-21 18:14  bouvigne

	* libmp3lame/encoder.c, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h:
	  FLOAT8 -> FLOAT

2003-12-21 12:46  bouvigne

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h:
	  FLOAT8 -> FLOAT

2003-12-21 10:16  bouvigne

	* libmp3lame/VbrTag.c, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/util.c, libmp3lame/util.h:
	  FLOAT8 -> FLOAT

2003-12-20 14:48  bouvigne

	* libmp3lame/psymodel.c:
	  fixed calc_interchannel_masking when encoding to mono

2003-12-20 14:33  bouvigne

	* libmp3lame/lame.c:
	  fixed lowpass value for high bitrate vbr profiles

2003-12-20 14:31  bouvigne

	* libmp3lame/version.h:
	  Using 3.95 as version number.
	  We will decide latter what to choose about version numbers

2003-12-19 13:55  bouvigne

	* libmp3lame/version.h:
	  Moved back to 3.94b in order to not have some 3.95b spreading on
	  the web.
	  We now have to decide if we should go in the 3.94release
	  direction or in the 3.95b direction

2003-12-18 18:47  bouvigne

	* libmp3lame/version.h:
	  welcome in 3.95

2003-12-18 17:13  aleidinger

	* ChangeLog:
	  ChangeLog for 3.94b.

2003-12-18 15:46  bouvigne

	* libmp3lame/version.h:
	  3.94b

2003-12-18 12:54  aleidinger

	* frontend/get_audio.c:
	  Takehiro noticed a bug in the pipe fix, it results in a click
	  sound at
	  the end of the encoding of BlackBirds.wav.
	  
	  This fix is a merge from the experimental branch and doesn't seem
	  to be
	  harmful while encoding over a pipe, but I'm not sure if I was
	  able to
	  reproduce the situation where the original fix was needed.
	  
	  I've tested:
	  mkfifo pipetest
	  lame -preset 96 pipetest pipetest.mp3
	  other shell: cat wav >pipetest

2003-12-17 18:45  bouvigne

	* frontend/parse.c:
	  help

2003-12-17 11:52  bouvigne

	* libmp3lame/psymodel.c:
	  updated comment regarding spreading function

2003-12-15 17:07  bouvigne

	* mpglib/layer3.c:
	  fixed 8kHz decoding

2003-12-09 17:22  olcios

	* frontend/parse.c:
	  Fixed typo.

2003-12-09 17:19  olcios

	* frontend/parse.c:
	  Fixed an error message.

2003-12-09 16:08  aleidinger

	* ACM/Makefile.in:
	  regen

2003-12-09 16:01  aleidinger

	* configure:
	  regen

2003-12-09 15:51  aleidinger

	* ACM/ddk/Makefile.in:
	  Add the autoconf Makefile.

2003-12-09 15:49  aleidinger

	* configure.in:
	  Add the ACM/ddk subdirectory.

2003-12-09 15:49  aleidinger

	* ACM/Makefile.am:
	  Add the ddk subdirectory.

2003-12-09 15:48  aleidinger

	* ACM/ddk/Makefile.am:
	  Add an automake Makefile.

2003-12-09 15:06  bouvigne

	* libmp3lame/version.h:
	  Bump alpha

2003-12-09 15:05  bouvigne

	* frontend/parse.c:
	  do not display --r3mix in help

2003-12-09 15:03  bouvigne

	* doc/html/history.html:
	  3.94 release date?

2003-12-09 14:59  bouvigne

	* doc/html/basic.html, doc/html/examples.html,
	  doc/html/history.html, doc/html/index.html, doc/html/modes.html,
	  doc/html/presets.html, doc/html/switchs.html:
	  3.94 documentation

2003-12-08 16:52  bouvigne

	* libmp3lame/VbrTag.c:
	  Use integer format for ReplayGain peak value

2003-12-05 12:49  aleidinger

	* libmp3lame/i386/Makefile.in:
	  add missing "

2003-12-04 12:26  aleidinger

	* libmp3lame/i386/Makefile.in:
	  nasm -> NASM fix

2003-12-03 03:09  olcios

	* include/lame.h:
	  comments / doc

2003-12-03 02:18  olcios

	* frontend/get_audio.c, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c:
	  Brought back the possibility to run lame_decode_init() more than
	  once

2003-12-03 01:44  olcios

	* frontend/get_audio.c, frontend/parse.c, include/lame.h,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/set_get.c:
	  Made it impossible to simultaneously use mp3 input and decoding
	  on the fly.

2003-11-29 12:14  aleidinger

	* ChangeLog:
	  update

2003-11-29 12:10  aleidinger

	* test/cvscheck.sh:
	  add the automated test script which mt used to run every day

2003-11-29 12:09  aleidinger

	* configure:
	  regen with autoconf 2.57

2003-11-29 12:06  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/tinyxml/Makefile.in,
	  Dll/Makefile.in, Makefile.in, debian/Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  mac/Makefile.in, misc/Makefile.in, mpglib/Makefile.in:
	  regen with automake 1.7

2003-11-29 12:05  aleidinger

	* configure.in:
	  - Check for some features a little bit earlier.
	  - Fix a AC_DEFINE so the newer tools don't moan.

2003-11-29 12:03  aleidinger

	* libmp3lame/i386/Makefile.am:
	  HACK HACK HACK, UGLY UGLY HACK!
	  
	  make this work with libtool 1.5
	  
	  If anybody knows how do this in a more sane way, please teach me
	  how to
	  do it.

2003-11-29 12:01  aleidinger

	* ltmain.sh:
	  update with libtool 1.5 version

2003-11-29 12:00  aleidinger

	* aclocal.m4:
	  update with newer autotools

2003-11-29 11:59  aleidinger

	* config.h.in:
	  add some missing undefs

2003-11-29 11:16  aleidinger

	* frontend/rtp.c:
	  fix some errors which are new with gcc 3.x

2003-11-29 09:49  aleidinger

	* depcomp, frontend/depcomp, install-sh, libmp3lame/depcomp,
	  missing, mkinstalldirs, mpglib/depcomp:
	  use the ones provided by automake 1.7

2003-11-29 09:44  aleidinger

	* config.guess, config.sub:
	  use the ones provided with libtool 1.5

2003-11-28 14:27  bouvigne

	* libmp3lame/set_get.c:
	  Disabled replaygain decode because it crashes when using mp3
	  input

2003-11-27 08:24  bouvigne

	* ACM/ddk, ACM/ddk/msacmdrv.h, ACM/readme.txt:
	  Allow compilation of ACM codec without Microsoft platform DDK

2003-11-26 13:06  bouvigne

	* lame_vc6.dsw:
	  Updated ACM dependencies

2003-11-24 17:05  bouvigne

	* libmp3lame/lame.c:
	  Do not compute ReplayGain when not using vbr header

2003-11-24 13:52  bouvigne

	* libmp3lame/lame.c:
	  Reduced bandwith when encoding in mono

2003-11-24 13:39  bouvigne

	* libmp3lame/lame.c:
	  Fixed low bitrates lowpass and sampling rate values
	  (I probably broke it recently)

2003-11-21 17:07  bouvigne

	* Dll/BladeMP3EncDLL.c:
	  Display issues

2003-11-21 16:50  bouvigne

	* Dll/BladeMP3EncDLL.c:
	  Minor corrections

2003-11-20 16:04  bouvigne

	* Dll/MP3export.pas:
	  Fixed calling convention

2003-11-20 13:30  bouvigne

	* Dll/Example_vc6.dsw:
	  Corrected the dll example project

2003-11-20 11:16  bouvigne

	* frontend/get_audio.c, frontend/main.c, frontend/parse.c,
	  include/lame.h, libmp3lame/set_get.c:
	  Changed some Vorbis messages

2003-11-20 11:08  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  Removed mode_automs

2003-11-20 10:57  bouvigne

	* libmp3lame/fft.c, libmp3lame/lame.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  Removed CPU_features.i387

2003-11-20 10:46  bouvigne

	* libmp3lame/presets.c:
	  Aliased r3mix to vbr-mtrh V3

2003-11-19 13:00  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  anihilated streaming/radio/portable named presets, leaving only
	  the -Vx

2003-11-18 13:20  bouvigne

	* libmp3lame/presets.c:
	  low vbr-mtrh presets

2003-11-15 11:21  takehiro

	* configure:
	  regen

2003-11-15 11:14  takehiro

	* configure.in:
	  fix migw build(as is done in the experimental branch)

2003-11-15 10:50  takehiro

	* frontend/get_audio.c:
	  fix ambiguous message (Bug ID #830035)

2003-11-14 16:13  bouvigne

	* libmp3lame/lame.c:
	  use presets by default when using vbr-mtrh

2003-11-14 13:24  bouvigne

	* libmp3lame/presets.c:
	  vbr-mtrh high vbr presets

2003-11-10 15:41  bouvigne

	* libmp3lame/fft.c, libmp3lame/i386/cpu_feat.nas,
	  libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  SIMD->SSE

2003-11-10 15:31  bouvigne

	* libmp3lame/lame.c, libmp3lame/presets.c:
	  Add missing values for vbr_rh -Vx scale

2003-11-10 15:27  takehiro

	* libmp3lame/i386/fft3dn.nas:
	  fix cygwin build (with nasm)

2003-11-10 14:58  bouvigne

	* libmp3lame/lame.c, libmp3lame/presets.c:
	  fix lowpass values for low vbr presets

2003-11-10 12:06  bouvigne

	* libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/set_get.c:
	  Use presets by default when using vbr_rh

2003-11-05 14:03  bouvigne

	* libmp3lame/presets.c, libmp3lame/psymodel.c:
	  Use Vx scale for presets

2003-11-03 15:00  bouvigne

	* libmp3lame/lame.c, libmp3lame/presets.c:
	  allow vbr presets to not be enforced (preparative work before
	  defaulting)

2003-10-30 13:09  bouvigne

	* libmp3lame/lame.c:
	  Moved replaygain analysis after resampling. It was causing
	  problems when
	  used with non standard sampling frequencies.

2003-10-23 12:09  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  presets

2003-10-22 08:26  takehiro

	* libmp3lame/takehiro.c:
	  fix assertion failure reported by Gabriel.

2003-10-22 07:33  takehiro

	* libmp3lame/bitstream.c:
	  better debug message

2003-10-21 09:05  bouvigne

	* libmp3lame/set_get.c, libmp3lame/set_get.h,
	  libmp3lame/vbrquantize.c:
	  cleanup

2003-10-20 16:10  olcios

	* doc/html/history.html:
	  Added the word "Radio" to the description of the ReplayGain
	  contribution, as it is actually the "Radio" ReplayGain that is
	  computed.

2003-10-20 12:30  bouvigne

	* doc/html/history.html:
	  history

2003-10-20 11:08  bouvigne

	* libmp3lame/version.h:
	  bump alpha

2003-10-20 07:20  bouvigne

	* libmp3lame/quantize_pvt.c:
	  fixed build problem

2003-10-17 14:39  bouvigne

	* libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h:
	  One more speedup in calc_noise, but the function is now a little
	  messy

2003-10-17 13:50  bouvigne

	* configMS.h:
	  updated version number

2003-10-17 11:49  bouvigne

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  Speedup calc_noise (about 5-10%) by re-using noise values per sfb
	  between
	  successive calls

2003-10-16 13:26  bouvigne

	* libmp3lame/VbrTag.c, libmp3lame/presets.c:
	  Put preset used in tag even when just using -b or --abr

2003-10-16 12:55  bouvigne

	* libmp3lame/VbrTag.c:
	  fixed a Lame header bug:
	  -b xxx --preset yyy was putting xxx as minimum bitrate value in
	  the
	  tag instead of the value defined in the preset yyy

2003-10-02 09:25  takehiro

	* frontend/parse.c:
	  fix build problem.(I made a mistake when back porting ...)

2003-09-27 06:16  takehiro

	* frontend/parse.c:
	  added priority control on WIN32, backport from experimental
	  branch.

2003-09-12 13:00  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/presets.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/set_get.c, libmp3lame/set_get.h, libmp3lame/util.h:
	  --psymodel 1/2 (gpsycho is 1 and nspsytune is 2)

2003-08-20 16:45  bouvigne

	* libmp3lame/lame.c:
	  fixed CBR settings bug
	  (introduced by me)

2003-08-20 16:16  bouvigne

	* libmp3lame/set_get.c:
	  fixed an assertion failure on lame_get_interChRatio
	  (introduced by me)

2003-08-16 10:36  aleidinger

	* doc/html/history.html:
	  add entry for the named pipe bug

2003-08-16 10:26  aleidinger

	* frontend/get_audio.c:
	  Work around a bug which shows up if the input is a FIFO.
	  
	  Tested by: Ville Herva <vherva@niksula.hut.fi>

2003-08-10 12:48  bouvigne

	* include/lame.h, libmp3lame/lame.c, libmp3lame/presets.c,
	  libmp3lame/psymodel.h, libmp3lame/set_get.c,
	  libmp3lame/set_get.h:
	  use presets by default for abr/cbr

2003-07-16 01:29  rbrito

	* debian/changelog, debian/control, debian/copyright,
	  debian/lame.docs, debian/lame.files, debian/rules:
	  Updated debian packaging.

2003-07-11 16:35  aleidinger

	* doc/html/id3.html, doc/man/lame.1, frontend/parse.c:
	  Add --ignore-tag-errors, mostly submitted by Adam Luter
	  <Adam.Luter@matchwave.com>.

2003-06-04 17:42  takehiro

	* Dll/BladeMP3EncDLL.c:
	  fix quality setting API(patch for BugID 746439 at sf.net,
	  reported by mikokong).

2003-06-01 18:22  aleidinger

	* libmp3lame/Makefile.in:
	  regen

2003-06-01 18:22  aleidinger

	* libmp3lame/Makefile.am:
	  add new private header

2003-06-01 17:20  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/libmp3lame_vc6.dsp,
	  libmp3lame/presets.c, libmp3lame/set_get.c, libmp3lame/set_get.h:
	  private header for internal set/get interfaces

2003-06-01 16:12  bouvigne

	* libmp3lame/id3tag.c:
	  let's use snprintf, with a define for msvc

2003-06-01 09:14  bouvigne

	* libmp3lame/VbrTag.c:
	  fix seek table for freeformat
	  
	  There is still something strange because with
	  -b128 and
	  -b128 --freeformat
	  I do not get the same seek table

2003-05-31 14:05  bouvigne

	* libmp3lame/id3tag.c:
	  ID3 TAG of "Encoder name", backported from experimental branch.
	  Btw snprintf does not exist under msvc, so I used sprintf.

2003-05-25 15:34  bouvigne

	* libmp3lame/VbrTag.c:
	  corrected the bitrate index of LAME tag in case of freeformat,
	  but the seek table is still wrong

2003-05-20 12:21  aleidinger

	* configure, debian/Makefile.in:
	  regen

2003-05-20 12:17  aleidinger

	* aclocal.m4:
	  remove gtk12-config

2003-05-20 12:14  aleidinger

	* Makefile.am, debian/Makefile.am:
	  move the dist-hook into the directory where the action happens

2003-05-20 12:13  aleidinger

	* configure.in:
	  some enhancements for compiler problems and detection of the
	  right size of some types

2003-05-18 15:05  takehiro

	* Makefile.am:
	  added chmod +x debian/rules when make dist.

2003-05-18 14:15  takehiro

	* Dll/BladeMP3EncDLL.h:
	  patch from sf.net, Dll compile with gcc.

2003-05-11 15:03  bouvigne

	* libmp3lame/presets.c:
	  updated low bitrate presets.
	  Considering quality, we have:
	  ns1 comp1 > ns1 comp3 >ns2 comp3 >ns2 comp1

2003-05-11 13:15  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  --sfscale, inhibited -Z

2003-05-11 09:49  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/quantize.c, libmp3lame/set_get.c, libmp3lame/util.h:
	  --sbgain

2003-05-09 14:06  aleidinger

	* frontend/lametime.c, frontend/main.c:
	  Fix on SunOs 4, Submitted by: Ti Kan <ti@amb.org>

2003-05-09 13:44  aleidinger

	* frontend/brhist.c:
	  fix on BSD/OS 3.x, Submitted by: Ti Kan <ti@amb.org>

2003-05-09 13:14  bouvigne

	* libmp3lame/quantize.c:
	  re-enabling the psfb12 analog silence detection. This time, the
	  coeffs should
	  be correct.
	  However, the coding gain is now quite small

2003-05-08 12:58  bouvigne

	* libmp3lame/quantize.c:
	  I am disabling the code for psfb12 analog
	  silence detection, as it produces some kind of
	  dropouts. It probably means that there is a bug
	  somewhere inside.

2003-05-04 15:36  bouvigne

	* libmp3lame/version.h:
	  bump alpha

2003-05-04 15:35  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  working on presets

2003-05-03 14:34  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  ability to set vbr smooth (for vbr-mtrh)

2003-05-02 17:38  bouvigne

	* frontend/parse.c:
	  internal command line options only available in debug or in
	  alpha.
	  
	  This way we should not be worried again by users using non
	  docummented options:
	  they will not be available for them

2003-05-02 16:18  bouvigne

	* libmp3lame/vbrquantize.c:
	  comments based on some mails exchanged with Robert

2003-05-01 15:23  bouvigne

	* libmp3lame/lame.c:
	  maskingadjust/maskingadjust_short available for vbr-mtrh

2003-04-27 15:18  bouvigne

	* libmp3lame/encoder.h, libmp3lame/l3side.h, libmp3lame/lame.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.h:
	  Analog silence detection in partitionned sfb21 or sfb12 for short
	  blocks

2003-04-20 17:32  bouvigne

	* libmp3lame/presets.c:
	  changed short block quantization comparison to method 0
	  for abr/cbr on medium bitrates (because of spahm, fatboy and the
	  likes) ,
	  as suggested by Takehiro.

2003-04-20 17:03  bouvigne

	* libmp3lame/version.h:
	  bump alpha

2003-04-20 16:56  bouvigne

	* libmp3lame/quantize.c:
	  trivial coding change in psfb21_analogsilence

2003-04-20 15:57  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  presets (testing)

2003-04-20 13:28  bouvigne

	* libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  simplified prototype of FindNearestBitrate

2003-04-20 12:58  bouvigne

	* libmp3lame/presets.c:
	  preset insane is now aliased to preset 320

2003-04-20 12:53  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  removed some old experimental presets

2003-04-20 12:33  bouvigne

	* frontend/parse.c, libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/presets.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/set_get.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  removed presetTune struct

2003-04-19 16:34  bouvigne

	* libmp3lame/presets.c:
	  fixed medium bitrates cbr/abr presets

2003-04-19 10:19  bouvigne

	* TODO:
	  updated the TODO file

2003-04-12 13:47  bouvigne

	* doc/html/contributors.html, doc/html/history.html:
	  history

2003-04-12 10:20  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c:
	  enable fast replaygain analysis by default
	  define DECODE_ON_THE_FLY if compiled with mpglib

2003-03-30 15:11  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  presets portable/portable1

2003-03-24 21:21  bouvigne

	* libmp3lame/version.h:
	  bump alpha

2003-03-24 21:19  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  presets

2003-03-24 20:32  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/quantize.c,
	  libmp3lame/set_get.c, libmp3lame/util.h:
	  maskingadjust/maskingadjustshort for vbr-rh

2003-02-27 17:57  bouvigne

	* libmp3lame/libmp3lame_vc6.dsp:
	  fix MSVC build (replay gain)

2003-02-27 17:46  bouvigne

	* libmp3lame/encoder.h, libmp3lame/l3side.h, libmp3lame/lame.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/util.c, libmp3lame/util.h:
	  analog silence detection in partitionned sfb21

2003-02-23 16:30  olcios

	* libmp3lame/lame.c:
	  fixed bug with --resample and --replaygain (input data analysis)

2003-02-22 02:21  olcios

	* Makefile.B32, Makefile.DJGPP, Makefile.MSVC, Makefile.unix,
	  USAGE, configMS.h, frontend/parse.c, include/lame.h,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/gain_analysis.c, libmp3lame/gain_analysis.h,
	  libmp3lame/lame.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/set_get.c, libmp3lame/util.h:
	  ReplayGain analysis of a single file (RadioGain)

2003-02-21 15:54  cisc

	* config.guess, config.sub:
	  Update to latest config files to support more hosts/targets.

2003-02-19 19:46  olcios

	* libmp3lame/mpglib_interface.c, libmp3lame/util.h,
	  mpglib/decode_i386.c, mpglib/decode_i386.h, mpglib/interface.c,
	  mpglib/interface.h, mpglib/layer3.c, mpglib/layer3.h:
	  Extending mpglib by the option to return unclipped floating-point
	  values of samples.

2003-02-19 12:14  takehiro

	* libmp3lame/presets.c:
	  fix comment style. ( // -> /* */ )

2003-02-18 01:11  olcios

	* Makefile.DJGPP:
	  Workaround for a problem with make clean, regarding long
	  filenames on some systems.

2003-02-16 18:46  bouvigne

	* libmp3lame/presets.c:
	  cbr/abr presets (still not finished?)

2003-02-16 08:38  takehiro

	* libmp3lame/psymodel.c:
	  PE calculation improvement (backported from experimental branch).
	  Better pe calculation for nspsytune. This brings
	  - better MS/LR switching
	  - better reservoir handling (brings better ABR/CBR result)
	  
	  I hope the trouble on "SeriousTrouble.wav" is partially fixed
	  (not perfect ?).

2003-02-16 08:25  takehiro

	* libmp3lame/quantize_pvt.h:
	  fix build problem (calc_xmin() prototype fix)

2003-02-16 08:22  takehiro

	* libmp3lame/quantize.c:
	  fix one more subblock_gain handling code.

2003-02-15 14:07  bouvigne

	* libmp3lame/quantize.c:
	  init max_nonzero_coeff to 575 instead of 0

2003-02-14 17:41  bouvigne

	* libmp3lame/l3side.h, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/takehiro.c:
	  do not calc noise in upper 0 part

2003-02-09 12:36  bouvigne

	* libmp3lame/quantize_pvt.c:
	  revert speedup

2003-02-09 01:25  markt

	* include/lame.h:
	  added a comment about reading the file API

2003-02-08 17:46  bouvigne

	* libmp3lame/presets.c, libmp3lame/version.h:
	  presets

2003-02-08 13:53  bouvigne

	* libmp3lame/quantize_pvt.c:
	  small speedup

2003-02-08 09:51  takehiro

	* frontend/main.c:
	  trivial simplification

2003-02-07 18:17  bouvigne

	* DEFINES:
	  KLEMM_44

2003-02-06 20:04  markt

	* include/lame.h:
	  updated comments about lame_init() from Kendrick Hamilton

2003-02-02 18:51  bouvigne

	* libmp3lame/version.h:
	  bump alpha

2003-02-02 18:50  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/presets.c,
	  libmp3lame/set_get.c:
	  preset medium

2003-02-02 17:10  bouvigne

	* libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  removed KLEMM_44

2003-02-02 15:50  bouvigne

	* libmp3lame/lame.c, libmp3lame/util.h:
	  removed useless structure (coding_t)

2003-02-02 14:45  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  presets: standard should be fine, working on medium

2003-02-02 10:15  bouvigne

	* libmp3lame/psymodel.c:
	  comments regarding TMN and NMT (thank you Takehiro)

2003-01-30 12:13  takehiro

	* frontend/get_audio.c:
	  bit better mp3 input file sync. but we need one more for
	  Chris.mp3....

2003-01-30 08:48  markt

	* libmp3lame/encoder.c, libmp3lame/lame.c:
	  Formula for the amount of data that had to be in the buffer
	  before it could be encoded was incorrect. Old formula
	  (for the polyphase filterbank) was:
	  
	  old formula: 286+576*(1+gfc->mode_gr) = 2014 (MPEG1)
	  
	  Correct formula is: 512+framesize -32 = 1632
	  
	  Best way to see this is to go through window_subband() in
	  newmdct.c
	  It access samples x[32*i] through x[510+32*i] for i=0..17.
	  
	  So our old condition was too conservative.
	  
	  However, the FFT still requires 1904 samples, so the
	  net change is just to lower the requirement from 2014 to 1904.

2003-01-28 00:29  markt

	* libmp3lame/lame.c:
	  short blocks are coupled, even in regular stereo mode.
	  This is to fix bugs reported in several different hardware
	  decoders.
	  see comments in lame.c

2003-01-27 11:39  takehiro

	* libmp3lame/VbrTag.c:
	  fix warning(added return value)

2003-01-26 17:37  takehiro

	* frontend/get_audio.c:
	  added ID3tag v2 skipping function.

2003-01-26 12:09  bouvigne

	* libmp3lame/version.h:
	  bumped alpha

2003-01-26 12:08  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  presets (under testing)

2003-01-26 11:31  bouvigne

	* libmp3lame/lame.c:
	  verbose

2003-01-25 11:02  bouvigne

	* libmp3lame/lame.c, libmp3lame/quantize.c:
	  changed -q mapping

2003-01-24 19:26  bouvigne

	* doc/html/history.html:
	  history

2003-01-21 02:46  takehiro

	* frontend/parse.c:
	  added \n in --preset help

2003-01-20 18:09  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  presets (under testing)

2003-01-19 15:55  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/presets.c:
	  presets (under testing)

2003-01-13 00:49  glessard

	* frontend/get_audio.c, frontend/main.c, frontend/main.h,
	  frontend/parse.c:
	  NOGAP-related changes.
	  - added '-T' option to force VBR/INFO tag inclusion.
	  - moved a VBR/INFO tag inclusion decision to parse.c, eliminating
	  the use
	  of a global variable across modules.
	  - fixed reading of multiple little-endian AIFF-C files in
	  get_audio.c. For
	  those files, 'swapbytes' was getting toggled for each file opened
	  in the
	  sequence, so only every other output file was correct.

2003-01-11 18:40  bouvigne

	* libmp3lame/presets.c:
	  low bitrates presets

2003-01-11 17:56  markt

	* libmp3lame/VbrTag.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/util.c:
	  reorganized VBR tag initialization code so that it
	  could be reset/initialized several times
	  (for use when encoding in nogap mode)

2003-01-11 17:42  markt

	* libmp3lame/lame_global_flags.h:
	  removed comment within comment delimiter

2003-01-11 17:41  markt

	* include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/lame_global_flags.h:
	  cleaned up some comments about VBR tags being written
	  to the mp3 file after internal data structures are free
	  (this has not been true for a long time)
	  
	  na

2003-01-11 15:43  bouvigne

	* libmp3lame/presets.c:
	  low bitrates presets

2003-01-11 12:46  takehiro

	* libmp3lame/lame.c:
	  fix ath curve initialization.

2003-01-11 04:16  markt

	* frontend/main.c, libmp3lame/VbrTag.c:
	  in nogap mode, with VBR tags, we were calling init_bitstream()
	  (which resets VBR tag data) before writing the VBR tag data.
	  This is now fixed, but there is still a bug when initializing the
	  VBR data for the 2nd (or more) encoding in the nogap series.

2003-01-10 21:22  markt

	* frontend/parse.c:
	  forgot a \n in the --longhelp documentation

2003-01-10 20:18  markt

	* USAGE:
	  note that --voice option is obsolete

2003-01-10 20:12  markt

	* frontend/main.c, frontend/main.h, frontend/parse.c:
	  added the "--nogaptags" option.
	  
	  this will allow VBR tags even for gapless encoding.
	  The default behavoir is to disable tags for gapless encoding.

2003-01-10 20:07  markt

	* frontend/parse.c, libmp3lame/VbrTag.c:
	  Updated the TOC calculation in the VBR tag to also count
	  the size of the VBR tag itself.
	  
	  This is to make the VBR tag consistent with the recent change
	  in the VBR frame count (changed so that it also counts the
	  VBR tag as a frame)

2003-01-02 18:42  bouvigne

	* libmp3lame/psymodel.c:
	  In psymodel_init set blocktype_old to NORM_TYPE because the VBR
	  header is long blocks. This might be the problem encountered in
	  some WMP 6.4

2003-01-02 18:37  bouvigne

	* libmp3lame/VbrTag.c:
	  Add the Info frame to the number of frames reported by the
	  Info/vbr tag

2003-01-02 18:30  bouvigne

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/vbrquantize.c:
	  reduced 63 warnings to 55 under visual c++. Those were mainly
	  differences between function prototypes in declarations and
	  implementations.

2003-01-02 18:27  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/set_get.c, libmp3lame/util.c, libmp3lame/util.h:
	  Ability to change ath4 shape independantly from vbr_q

2003-01-01 16:14  aleidinger

	* libmp3lame/machine.h:
	  Removed u_char typedefs, u_char is not used in the tree
	  (according to grep -r u_char).

2003-01-01 16:12  aleidinger

	* frontend/rtp.c:
	  Include alloca.h if neccessary.
	  
	  Noticed by: Tomi Vainio <Tomi.Vainio@Sun.COM>

2002-12-27 04:42  takehiro

	* libmp3lame/psymodel.c:
	  use ATH value as short block masking lower limit.

2002-12-27 04:21  takehiro

	* frontend/parse.c, include/lame.h, libmp3lame/set_get.c,
	  libmp3lame/version.h:
	  new option --shortthreshold and bump up version number(alpha 8)
	  and new psymodel version(0.90) for ATHadjust aware psymodel

2002-12-27 04:17  takehiro

	* libmp3lame/psymodel.c:
	  fix assertion failure (oops, my mistake in ATH handling change)

2002-12-26 05:43  takehiro

	* libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/psymodel.h:
	  cleanup psymodel initialization preparing for --shortthreshold
	  option.

2002-12-25 12:13  takehiro

	* libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/set_get.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  ATH handling fix and simplification.
	  - use ATH adjust in ns_msfix(), mask_add(), and masking lower
	  limit calculation.
	  - all ATH adjust related thresholding(ns_msfix, quantcompare and
	  scalefac_scale) is removed.

2002-12-18 16:29  takehiro

	* libmp3lame/libmp3lame_vc6.dsp:
	  removed tools.h

2002-12-18 16:27  takehiro

	* libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/debugscalefac.c, libmp3lame/tools.c,
	  libmp3lame/tools.h:
	  remove unused files and regen Makefile

2002-12-16 18:46  takehiro

	* libmp3lame/quantize.c:
	  oops, I checked in the wrong version...

2002-12-16 18:45  takehiro

	* libmp3lame/quantize.c:
	  fix inc_subblock_gain() to properly handle sfb12 (16kHz over
	  region) MDCT coefs.

2002-12-15 12:56  bouvigne

	* frontend/parse.c, libmp3lame/lame.c:
	  warning about alpha versions

2002-12-15 11:59  takehiro

	* ACM/ACMStream.cpp:
	  fix access violation when LAME failed to initilize.
	  I hope this fixes BUG ID 646894 (at sf.net).

2002-12-15 11:11  takehiro

	* libmp3lame/lame.c:
	  fix warning (reported at sf.net by krick.)

2002-12-15 11:10  takehiro

	* libmp3lame/takehiro.c, libmp3lame/util.h:
	  fix warning (reported at sf.net by krick.

2002-12-15 09:11  takehiro

	* libmp3lame/lame.c:
	  fix -q0 convined with --substep x option.

2002-12-15 08:56  takehiro

	* Makefile.B32, Makefile.DJGPP, Makefile.MSVC, Makefile.unix:
	  fix for vorbis support remove (I hope)

2002-12-14 19:19  robert

	* ChangeLog, Makefile.MSVC, lame.bat, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/util.c,
	  misc/lame4dos.bat, misc/mlame:
	  removed my email address

2002-12-13 17:48  takehiro

	* libmp3lame/quantize.c:
	  off by one fix for global_gain.

2002-12-13 16:58  takehiro

	* configure, frontend/main.h, libmp3lame/Makefile.in:
	  remove vorbis support and regen.

2002-12-13 16:55  takehiro

	* config.h.in, configure.in, frontend/get_audio.c, frontend/main.c,
	  frontend/parse.c, libmp3lame/Makefile.am, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c,
	  libmp3lame/vorbis_interface.c:
	  remove vorbis support.

2002-12-13 14:34  takehiro

	* libmp3lame/lame.c:
	  fix for 64bit environment, reported by Tomi Vainio
	  <Tomi.Vainio@Sun.COM>

2002-12-09 19:00  bouvigne

	* libmp3lame/quantize.c, libmp3lame/version.h:
	  fixed the quant_comp problem (probably introduced by me)
	  so bump the alpha version number

2002-12-08 17:40  takehiro

	* libmp3lame/version.h:
	  now 3.94 alpha 6
	  - changed default block switching threshold
	  - fixed mpglib crash.

2002-12-08 17:35  takehiro

	* libmp3lame/psymodel.c:
	  changed long/short block switching threshold.

2002-12-08 17:23  takehiro

	* testcase.mp3:
	  new testcase for the new default cutoff frequency.

2002-12-08 17:07  takehiro

	* mpglib/interface.c:
	  null pointer check patch from Peter Pawlowski <peter at blorp dot
	  com>, to prevent the random crash.

2002-12-08 17:02  takehiro

	* Makefile.B32, Makefile.DJGPP, Makefile.MSVC, Makefile.am,
	  Makefile.in, Makefile.unix, configMS.h, configure, configure.in,
	  frontend/brhist.c, frontend/brhist.h, frontend/console.c,
	  frontend/console.h, frontend/get_audio.c, frontend/get_audio.h,
	  frontend/gpkplotting.h, frontend/gtkanal.c, frontend/gtkanal.h,
	  frontend/lametime.c, frontend/lametime.h, frontend/main.c,
	  frontend/main.h, frontend/parse.c, frontend/parse.h,
	  frontend/portableio.h, frontend/rtp.c, frontend/rtp.h,
	  frontend/timestatus.c, frontend/timestatus.h, include/lame.h,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/VbrTag.c, libmp3lame/VbrTag.h, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/fft.c, libmp3lame/fft.h,
	  libmp3lame/gain_analysis.c, libmp3lame/gain_analysis.h,
	  libmp3lame/i386/Makefile.am, libmp3lame/i386/Makefile.in,
	  libmp3lame/id3tag.h, libmp3lame/l3side.h,
	  libmp3lame/lame-analysis.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/libmp3lame.dsp,
	  libmp3lame/machine.h, libmp3lame/mpglib_interface.c,
	  libmp3lame/newmdct.c, libmp3lame/newmdct.h, libmp3lame/presets.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/set_get.c, libmp3lame/tables.c, libmp3lame/tables.h,
	  libmp3lame/takehiro.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c, libmp3lame/vbrquantize.h,
	  libmp3lame/version.c, libmp3lame/version.h,
	  libmp3lame/vorbis_interface.c, mpglib/common.c, mpglib/common.h,
	  mpglib/dct64_i386.h, mpglib/decode_i386.h, mpglib/huffman.h,
	  mpglib/interface.c, mpglib/interface.h, mpglib/l2tables.h,
	  mpglib/layer1.c, mpglib/layer1.h, mpglib/layer2.c,
	  mpglib/layer2.h, mpglib/layer3.c, mpglib/layer3.h,
	  mpglib/mpg123.h, mpglib/mpglib.h, mpglib/tabinit.h, testcase.mp3:
	  overwrite by takehiro_stable branch.

2002-12-08 16:59  takehiro

	* mac:
	  added

2002-12-01 16:32  aleidinger

	* ChangeLog:
	  update for 3.93.1

2002-12-01 16:18  aleidinger

	* configure:
	  regen

2002-12-01 16:15  aleidinger

	* configure.in:
	  change version to 3.93.1

2002-12-01 16:12  aleidinger

	* configure:
	  regen

2002-12-01 16:10  aleidinger

	* doc/html/history.html:
	  sync with webpages

2002-12-01 16:06  aleidinger

	* configure.in:
	  Check for socket() in libsocket if it isn't in libc so rtp can
	  get compiled
	  on Solaris.
	  
	  Noticed by: "Niki W. Waibel" <niki.waibel@gmx.net>
	  
	  While I'm here, bail out if some other functions aren't there.

2002-11-30 18:18  bouvigne

	* libmp3lame/version.h:
	  version

2002-11-29 17:07  bouvigne

	* Dll/BladeMP3EncDLL.def:
	  dll exports

2002-11-28 21:05  bouvigne

	* doc/html/history.html, doc/html/index.html:
	  docs

2002-11-28 18:06  bouvigne

	* libmp3lame/version.h:
	  go back to 3.93

2002-11-28 18:04  bouvigne

	* libmp3lame/lame.c:
	  revert back q0 to q1 state (due to reports about quality
	  problems)

2002-11-25 18:00  bouvigne

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  sync presets with the executable front-end and libmp3lame

2002-11-23 18:54  bouvigne

	* libmp3lame/psymodel.c:
	  quick fix of nspsytune

2002-11-17 11:11  aleidinger

	* configure.in, libmp3lame/version.h:
	  We are now at 3.94alpha...

2002-11-17 10:41  aleidinger

	* doc/html/history.html:
	  Copy history from webpage.

2002-11-17 10:29  aleidinger

	* ChangeLog:
	  ChangeLog for 3.93.

2002-11-17 10:25  aleidinger

	* Dll/Makefile.in, Makefile.in:
	  regen

2002-11-17 10:24  aleidinger

	* Makefile.am:
	  Add lame.dsw.

2002-11-17 10:23  aleidinger

	* Dll/Makefile.am:
	  Oops... add missing backslash.

2002-11-17 10:18  aleidinger

	* Dll/Makefile.in, libmp3lame/Makefile.in:
	  regen

2002-11-17 10:16  aleidinger

	* libmp3lame/Makefile.am:
	  Add gain_analysis.[ch] to the distributed files.

2002-11-17 10:11  aleidinger

	* Dll/Makefile.am:
	  Add some missing files.

2002-11-17 10:09  aleidinger

	* configure:
	  gtk12-config -> gtk-config

2002-11-17 10:05  aleidinger

	* testcase.mp3:
	  "Official" testcase.mp3 for this release.

2002-11-17 09:51  aleidinger

	* ChangeLog.header:
	  Add some information about branches.

2002-11-17 09:46  aleidinger

	* libmp3lame/version.h:
	  We are at 3.93-RELEASE now...

2002-11-17 09:45  aleidinger

	* doc/man/lame.1, frontend/parse.c:
	  Warn about fast presets.

2002-11-16 10:07  bouvigne

	* doc/html/presets.html:
	  warning about fast presets

2002-11-13 20:24  bouvigne

	* libmp3lame/lame.c, libmp3lame/presets.c:
	  rollback of lowpass change before release

2002-11-07 18:47  bouvigne

	* libmp3lame/lame.c, libmp3lame/presets.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  change of the default lowpass values (will break testcases)

2002-10-27 14:37  aleidinger

	* configure:
	  regen

2002-10-27 14:36  aleidinger

	* configure.in:
	  Forced commit: last commit was
	  - remove -funroll-loops as suggested in the nearby comment
	  Noticed by: takehiro

2002-10-27 14:32  aleidinger

	* configure.in:
	  remove options as suggested in the nearby comment; Noticed by:
	  takehiro

2002-10-24 00:05  robert

	* libmp3lame/vbrquantize.c:
	  bug fix in vbr-new
	  possible division by zero

2002-10-21 01:27  markt

	* frontend/parse.c:
	  truely silent when input is stdin patch committed

2002-10-19 02:11  robert

	* frontend/brhist.c:
	  little bit prettier printing of block types stats

2002-10-18 22:54  robert

	* frontend/brhist.c, frontend/parse.c, include/lame.h,
	  libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  added some simple statistics about block types used:
	  added lame_block_type_hist() and lame_bitrate_block_type_hist()
	  to API
	  using new switch --brief will show one line of percentage types
	  used
	  using switch --verbose will show in more detail used block types

2002-10-17 17:05  bouvigne

	* frontend/parse.c:
	  removed some unused functions

2002-10-17 13:03  aleidinger

	* include/lame.h:
	  correct comment about *_bWriteVbrTag()

2002-10-16 21:13  robert

	* libmp3lame/vbrquantize.c:
	  fixing compiler warnings (unused code removed), some clean up

2002-10-16 18:01  bouvigne

	* frontend/get_audio.c, frontend/get_audio.h, frontend/parse.c,
	  frontend/parse.h, frontend/timestatus.h, include/lame.h,
	  libmp3lame/set_get.c, libmp3lame/takehiro.c:
	  minor code cleanups in order to reduce warnings

2002-10-15 19:16  robert

	* Makefile.unix, libmp3lame/psymodel.c, libmp3lame/util.h:
	  reenentrant fix for loudness approx.

2002-10-15 15:24  aleidinger

	* configure.in:
	  different way of specifying the gcc version in the sparc section,
	  the did not worked because autoconf does some nasty things with [
	  and ]

2002-10-15 11:30  aleidinger

	* doc/man/lame.1:
	  catch up with HTML docs

2002-10-15 09:54  aleidinger

	* configure, doc/html/Makefile.in:
	  regen

2002-10-15 09:54  aleidinger

	* doc/html/Makefile.am:
	  add presets.html

2002-10-15 09:53  aleidinger

	* doc/html/presets.html:
	  typo fix

2002-10-15 09:49  aleidinger

	* configure.in:
	  Use sparc optimization on gcc 3.[123456].
	  
	  Submitted by: "Spud McDuck" <dontspamonme@excite.com>

2002-10-15 09:33  aleidinger

	* frontend/brhist.c, frontend/main.c, frontend/main.h,
	  frontend/mp3rtp.c, frontend/mp3x.c, frontend/parse.c:
	  PATH_MAX is in POSIX.1 (does not include the trailing NUL, so add
	  +1), not MAXPATHLEN

2002-10-14 20:06  bouvigne

	* doc/html/contributors.html, doc/html/index.html,
	  doc/html/presets.html:
	  html docs

2002-10-08 19:30  bouvigne

	* libmp3lame/VbrTag.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  preset info stored into Info tag

2002-10-08 17:43  bouvigne

	* frontend/main.h:
	  win32 build fix

2002-10-08 12:58  takehiro

	* frontend/main.c:
	  sync my branch: fix warning of ReportLameTagProgress() definition

2002-10-08 09:09  aleidinger

	* configure:
	  regen

2002-10-08 08:59  aleidinger

	* ChangeLog:
	  update with uptodate data

2002-10-08 08:57  aleidinger

	* ChangeLog.header:
	  add info how to show the branches in the ChangeLog

2002-10-08 08:53  aleidinger

	* libmp3lame/version.h:
	  bump alpha version

2002-10-08 08:51  aleidinger

	* configure.in:
	  replace the -pipe hack for mips-sgi-irix with a test for GNU ld
	  (sorry, I forgot who submitted it)

2002-10-08 08:50  aleidinger

	* acinclude.m4:
	  actually commit the generated file with the alignment fix for 64
	  bit architectures

2002-10-08 08:50  aleidinger

	* ltconfig:
	  improved "deplibs_check_method" for freebsd-elf

2002-10-08 08:44  aleidinger

	* frontend/main.c, frontend/main.h, frontend/mp3rtp.c,
	  frontend/mp3x.c, frontend/parse.c:
	  - spell MAX_NAME_LEN the POSIX way (MAXPATHLEN) and make sure we
	  have
	  a fallback if it doesn't get defined (please test this on
	  Windows)
	  - fix for sizeof(short int) != sizeof(int) in mp3rtp.c
	  
	  Patches based upon bugreport from Greg Lehey <grog@freebsd.org>
	  and fixes
	  in the FreeBSD ports collection from Mikhail Teterin
	  <mi@freebsd.org>.

2002-09-30 16:49  bouvigne

	* libmp3lame/encoder.c, libmp3lame/quantize.c, libmp3lame/util.c:
	  some code cleanups

2002-09-28 09:30  takehiro

	* libmp3lame/VbrTag.c:
	  remove some #include (these header files are included from
	  VbrTag.h.)

2002-09-23 07:00  bouvigne

	* include/lame.h:
	  presets (sorry, this part was missing)

2002-09-22 16:53  bouvigne

	* frontend/parse.c, libmp3lame/presets.c:
	  presets

2002-09-21 00:31  markt

	* libmp3lame/quantize_pvt.c:
	  fixed bug which was causing the calc_noise assert failure.

2002-09-16 16:28  takehiro

	* configure, configure.in:
	  we don't want to run the X :)

2002-09-15 17:47  bouvigne

	* libmp3lame/reservoir.c, libmp3lame/set_get.c:
	  Relax of freeformat reservoir after discussion with Robert

2002-09-13 20:04  markt

	* frontend/brhist.c:
	  This code:
	  
	  
	  #if defined(HAVE_TERMCAP_H)
	  # include <termcap.h>
	  #elif defined(HAVE_NCURSES_TERMCAP_H)
	  # include <ncurses/termcap.h>
	  #endif
	  
	  changed to:
	  
	  #if defined(HAVE_NCURSES_TERMCAP_H)
	  # include <ncurses/termcap.h>
	  #elif defined(HAVE_TERMCAP_H)
	  # include <termcap.h>
	  #endif
	  
	  because on my system, both TERMCAP and NCURSES_TERMCAP were
	  defined by configure, but term.h and termcap.h have conflicts,
	  whileterm.h and ncurses/termcap.h works fine.

2002-09-09 21:43  robert

	* frontend/parse.c, libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/psymodel.c:
	  lame.c, encoder.c: fix padding oddities
	  psymodel.c: fix nspsy code for all/no short blocks
	  parse.c: -d now obsolete

2002-09-06 11:11  aleidinger

	* configure:
	  regen

2002-09-06 10:53  aleidinger

	* configure.in:
	  - no opts if non empty cflags and no expopt
	  - accumulate-outgoing-args is x86 only
	  Noticed by: Atro Tossavainen <atossava@cc.helsinki.fi>
	  - IRIX-gcc doesn't like '-pipe'
	  Noticed by: Atro Tossavainen <atossava@cc.helsinki.fi>

2002-09-06 09:24  aleidinger

	* frontend/brhist.c, frontend/console.c:
	  try to include the right headers for the tgetstr prototype;
	  should fix problems on 64bit arches
	  Noticed by: Atro Tossavainen <atossava@cc.helsinki.fi>
	  Ti Kan <ti@amb.org>
	  Tested by: Ti Kan <ti@amb.org>

2002-09-06 09:22  aleidinger

	* aclocal.m4:
	  prevent unaligned access on some 64 bit arches
	  Noticed by: Ti Kan <ti@amb.org>
	  Tested by: Ti Kan <ti@amb.org>

2002-09-05 16:53  bouvigne

	* libmp3lame/reservoir.c, libmp3lame/set_get.c:
	  free format fix

2002-09-03 11:34  aleidinger

	* frontend/parse.c:
	  1 argument too much for fprintf

2002-09-03 11:18  aleidinger

	* doc/man/lame.1:
	  add --noasm

2002-09-02 12:02  takehiro

	* doc/html/history.html:
	  added my psymodel change, and so on.

2002-09-02 11:56  takehiro

	* testcase.mp3:
	  regen with the latest LAME after psymodel was fixed.
	  This is produced by the LAME with ./configure --enable-nasm
	  --enable-debug --enable-mp3x, compiled gcc 3.1.1, on linux2.4.19,
	  glibc2.2.2, Duron 1GHz(3DNow! is enabled).

2002-09-02 11:51  takehiro

	* libmp3lame/lame.c:
	  fix for Tru64 UNIX build, reported by Ti Kan.

2002-09-01 16:08  bouvigne

	* doc/html/contributors.html, doc/html/history.html,
	  doc/html/index.html, doc/html/switchs.html, frontend/parse.c:
	  docs

2002-09-01 16:03  takehiro

	* libmp3lame/psymodel.c:
	  fix small psymodel bugs since long time ago(some of them are
	  since 3.70 era...)
	  - highpass filter of MS energy calculation to neglect the DC.
	  - always do the Robert's short block pre-echo prevention and
	  fix its initialization (the fix is only for nspsy).
	  - fix energy and threshold of DC and the highest freq. element
	  are halven.
	  - fix nspsytune FIR filter length

2002-08-30 20:52  robert

	* Makefile.MSVC:
	  MSVC: fix "-GX" typo
	  ICL: allow special compiles for PII and PIII

2002-08-27 22:43  rbrito

	* debian/changelog, debian/rules:
	  Update Debian packaging. More changes to come.

2002-08-26 20:38  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/util.h:
	  undo of VBR bitres min fill patch

2002-08-25 20:17  bouvigne

	* frontend/parse.c:
	  merge of alt presets

2002-08-20 16:28  takehiro

	* libmp3lame/encoder.c:
	  fix mp3x spectrum display

2002-08-08 09:03  aleidinger

	* Makefile.in:
	  regen

2002-08-08 09:00  aleidinger

	* Makefile.am:
	  Add lame.spec to the distribution tarball, as suggested in bug
	  #580586

2002-08-08 08:49  aleidinger

	* Makefile.DJGPP, README.DJGPP:
	  Fix from Peder Hedlund, see bug #590981.

2002-08-05 18:04  afaber

	* Dll/BladeMP3EncDLL.c:
	  Fixed crashing problem when the INFO TAG was written while it
	  should not

2002-08-02 20:53  afaber

	* Dll/BladeMP3EncDLL.c:
	  Some code clarifications (remarks from Martin Ruckert)

2002-08-02 20:52  afaber

	* Dll/LameDLLInterface.htm:
	  Small textual changes (remarks from Martin Ruckert)

2002-07-30 20:24  aleidinger

	* frontend/get_audio.c, libmp3lame/VbrTag.c, libmp3lame/lame.c:
	  Portability fixes for SunOS 4.x.

2002-07-28 09:18  takehiro

	* include/lame.h:
	  fix unclosed comment
	  fix typo

2002-07-28 06:59  rbrito

	* debian/changelog:
	  Updating debian/changelog's reported version.

2002-07-27 10:35  afaber

	* Dll/Example.cpp, Dll/Example_vc6.dsp, Dll/Example_vc6.dsw:
	  Added example VC6 project files, updated Example.cpp file

2002-07-26 22:27  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.def,
	  Dll/BladeMP3EncDLL.h:
	  changes to support encoding of multiple streams

2002-07-26 18:56  markt

	* frontend/get_audio.c, include/lame.h,
	  libmp3lame/mpglib_interface.c:
	  memory leak in multiple decodes fixed.

2002-07-21 15:15  aleidinger

	* include/lame.h:
	  ANSIfy comments. Requested by Ti Kan <ti@amb.org> for xmcd

2002-07-19 17:20  afaber

	* Dll/LameDll_vc6.dsp:
	  added ,HAVE_NASM,MMX_choose_table defines to project file,
	  since beGetBersion did not report that the DLL was compiled with
	  MMX enabled

2002-07-15 21:22  markt

	* libmp3lame/util.c, libmp3lame/util.h:
	  enabled domd's fast_log code

2002-07-13 08:11  aleidinger

	* configure, libmp3lame/Makefile.in:
	  regen

2002-07-13 08:09  aleidinger

	* configure.in:
	  Some fixes in the compiler options, removed superfluous BeOS
	  section.

2002-07-13 08:08  aleidinger

	* libmp3lame/Makefile.am:
	  Fixes for BeOS (and perhaps other arches) to allow the build of a
	  shared lib.
	  
	  Submitted by: Andrew Bachmann <shatty@myrealbox.com>

2002-07-05 15:17  afaber

	* Dll/LameDLLInterface.htm:
	  fixe a couple of typos'

2002-07-03 22:28  afaber

	* Dll/BladeMP3EncDLL.c:
	  fixed minor problems with the ALT-PRESET CBR/ABR modes

2002-07-03 21:14  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  avoid crash in beWriteVbrTag when gfp is not allocated, fixed
	  return value of beWriteVbrTag in *.h file

2002-06-29 12:29  aleidinger

	* libmp3lame/lame.c, libmp3lame/util.c:
	  Patch from Simon Blandford <ome@simonb.org.uk>:
	  ---snip---
	  The problem:
	  When in MONO mode, LAME merges the two channels before the
	  left/right
	  scaling. The left scale acts on the remaining signal and the
	  right scale
	  is ignored.
	  
	  The solution:
	  The scaling operation is simply moved before the channel merge
	  operation. Also, all these operations are moved before the
	  re-sample
	  operation otherwise the signal would have be re-sampled on two
	  channels
	  even in MONO mode which would be a waste of CPU time.
	  ---snip---

2002-06-17 17:19  domd

	* DEFINES:
	  added USE_FAST_LOG define

2002-06-16 20:16  domd

	* libmp3lame/psymodel.c:
	  optimizations on mask_add and L3psycho_anal_ns

2002-06-16 20:14  domd

	* libmp3lame/quantize_pvt.c:
	  optimization : use of FAST_LOG10 macro in calc_noise

2002-06-16 20:13  domd

	* libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  added the FAST_LOG10 and FAST_LOG macros

2002-06-15 17:56  markt

	* Dll/BladeMP3EncDLL.c:
	  added missing ")"

2002-06-12 18:41  markt

	* dshow/Encoder.cpp:
	  directshow, like ACM, cannot support INFO tag.
	  INFO tag disabled in dshow code.

2002-06-12 18:36  markt

	* ACM/ACMStream.cpp:
	  added note about INFO tag not possible when using ACM codec,
	  since ACM has no mechanism to rewind output stream when
	  encoding is finished.

2002-06-12 18:27  markt

	* API, Dll/BladeMP3EncDLL.c, Dll/Example.cpp,
	  Dll/LameDLLInterface.htm:
	  updated documentation (API and DLL)
	  to clarify that INFO tag must be written
	  unless explicitly disabled.

2002-06-11 09:21  aleidinger

	* configure:
	  gtk-config fix (oops... sorry)

2002-06-09 16:01  takehiro

	* include/lame.h, libmp3lame/set_get.c:
	  fix return type of lame_get_interChRatio()

2002-06-04 19:16  bouvigne

	* libmp3lame/psymodel.c:
	  merged little part of Dom's speed optimization

2002-06-04 15:04  aleidinger

	* configure:
	  regen

2002-06-04 15:03  aleidinger

	* configure.in:
	  -maccumulate-outgoing-args is x86 only
	  
	  Reported by: "Spud McDuck" <dontspamonme@excite.com>

2002-06-02 19:09  aleidinger

	* configure:
	  regen

2002-06-02 19:08  aleidinger

	* config.h.in, configure.in:
	  damn autoconf, hopefully this version works as intended

2002-06-02 17:02  aleidinger

	* configure:
	  regen

2002-06-02 17:00  aleidinger

	* configure.in:
	  use "else if ... fi" instead of "elif ..."

2002-06-02 17:00  aleidinger

	* libmp3lame/Makefile.in:
	  regen (removal of pcm.[ch])

2002-06-02 14:50  robert

	* libmp3lame/Makefile.in, libmp3lame/libmp3lame_vc6.dsp:
	  moved unused code pcm.[ch] into the Attic
	  interested parties will find it there

2002-06-02 14:45  robert

	* libmp3lame/Makefile.am, libmp3lame/pcm.c, libmp3lame/pcm.h:
	  moved unused code pcm.[ch] into the Attic
	  interested parties will find it there

2002-06-02 14:20  aleidinger

	* configure:
	  regen + perl -pi -e 's:--no-verify::' configure
	  
	  Does anyone know how to do this with autoconf directly?

2002-06-02 14:17  aleidinger

	* aclocal.m4:
	  undo last commit, wrong file and wrong commit message

2002-06-02 14:16  aleidinger

	* acconfig.h:
	  not needed anymore

2002-06-02 14:15  aleidinger

	* Makefile.in:
	  regen after removal of aclocal.m4

2002-06-02 14:13  aleidinger

	* config.h.in:
	  regen

2002-06-02 14:13  aleidinger

	* configure.in:
	  - redo {,u}int{8,16,32,64}_t stuff with new autoconf 2.53 macros
	  - fix code which sets experimental options
	  Thanks to: Akos Maroy <darkeye@tyrell.hu>
	  - fix GCC_version for 3.1
	  Thanks to: Akos Maroy <darkeye@tyrell.hu>
	  - further autoconf 2.53 updates
	  - minor fixes

2002-06-02 14:10  aleidinger

	* aclocal.m4:
	  is not needed anymore

2002-06-02 12:21  aleidinger

	* aclocal.m4:
	  update to autoconf 2.53

2002-06-02 12:17  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/VbrTag.h, libmp3lame/pcm.c,
	  libmp3lame/pcm.h, libmp3lame/util.h:
	  add some includes

2002-05-28 08:28  aleidinger

	* ACM/ADbg/Makefile.in, ACM/Makefile.in, ACM/tinyxml/Makefile.in,
	  Dll/Makefile.in, configure, debian/Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  regen

2002-05-28 08:07  aleidinger

	* configure.in:
	  - update to autoconf 2.53
	  - add ultrasparc detection and compiler switches in case of gcc
	  3.1

2002-05-28 08:04  aleidinger

	* Makefile.in:
	  regen

2002-05-28 08:04  aleidinger

	* Makefile.am:
	  add mac subdir

2002-05-28 08:03  aleidinger

	* mac/Makefile.in:
	  generated Makefile.in

2002-05-28 07:52  aleidinger

	* mac/Makefile.am:
	  add automake clue

2002-05-28 07:48  aleidinger

	* mac/LAME Carbon Debug.pch, mac/LAME Carbon Final.pch, mac/LAME
	  Classic Debug.pch, mac/LAME Classic Final.pch,
	  mac/LAME_Carbon_Debug.pch, mac/LAME_Carbon_Final.pch,
	  mac/LAME_Classic_Debug.pch, mac/LAME_Classic_Final.pch:
	  rename files because of automake (space to underscore
	  conversation)

2002-05-25 23:30  markt

	* libmp3lame/lame.c:
	  fixed a mono encoding bug found by Justin Schoeman
	  
	  For single channel input, the 2nd channel was still required
	  to exit and have as much data as the 1st channel becaus of
	  memcpy's (even though this data was never used)

2002-05-24 17:28  markt

	* Dll/BladeMP3EncDLL.c, libmp3lame/lame.c:
	  removed all references to "sample_t" from the DLL.
	  
	  someone put in a call to lame_encode_buffer_sample_t
	  (not a public interface to libmp3lame)
	  when they should have used lame_encode_buffer_float(),
	  since they were assuming 'sample_t' was a float.

2002-05-20 22:21  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/util.h:
	  VBR bitreservoir minimum filling depending on back pointer from
	  largest frame allowed

2002-05-20 18:22  robert

	* Makefile.unix, libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/vbrquantize.c:
	  Makefile.unix:
	  -malign-double enabled, helps memcpy
	  
	  lame.c:
	  new experimental -X{5,9} settings for --vbr-new
	  
	  quantize.c:
	  build up bit reservoir for VBR, allows to limit max frame size
	  with -Bxxx
	  without a too large quality impact
	  
	  vbrquantize.c:
	  fixing sfb21 issues: -Y did not show the wanted effect
	  xxx_block_xr34: replaced copy loop by memcpy, not faster but
	  makes things
	  more obvious
	  some very minor speed ups

2002-05-19 22:11  markt

	* frontend/timestatus.c:
	  initialize ellapsed time counter to 0.

2002-05-19 22:03  robert

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c:
	  avoid calling best_huffman_divide twice for vbr-new

2002-05-14 21:32  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c, libmp3lame/vbrquantize.h:
	  removed warning when using -Y
	  removed unused bitpressure strategie 2 (VBR)

2002-05-14 21:02  robert

	* libmp3lame/.indent.pro:
	  changing max line length from 80 to 100 chars

2002-05-14 16:23  markt

	* frontend/timestatus.c:
	  added comment about clock rollover check

2002-05-14 16:19  markt

	* frontend/lametime.c, frontend/timestatus.c:
	  put in rollover check in frontend cpu/wall clock timing routines.
	  
	  didn't we used to have this?
	  
	  user reported (on a 35min encode) that the output of clock()
	  (of type clock_t) was rolling over and becoming negative.
	  On his system, clock_t is a signed 32 bit int.

2002-05-12 18:47  markt

	* INSTALL:
	  removed info about vorbis encoding from INSTALL

2002-05-12 18:01  markt

	* ltmain.sh:
	  applied patch to ltmain.sh
	  from sourceforge.net bug report. it passes LDFLAGS on to the
	  linker?
	  
	  *** ltmain.sh.org Sun Oct 22 14:06:32 2000
	  --- ltmain.sh Sun Apr 28 11:56:54 2002
	  ***************
	  *** 818,824 ****
	  convenience=
	  old_convenience=
	  deplibs=
	  ! linkopts=
	  
	  if test -n "$shlibpath_var"; then
	  # get the directories listed in $shlibpath_var
	  --- 818,824 ----
	  convenience=
	  old_convenience=
	  deplibs=
	  ! linkopts="$LDFLAGS"
	  
	  if test -n "$shlibpath_var"; then
	  # get the directories listed in $shlibpath_var

2002-05-12 17:20  markt

	* include/lame.h, libmp3lame/lame-analysis.h:
	  added documentation to lame.h for lame_decode* functions
	  
	  uped READ_AHEAD from 10 to 40 for the frame analyzer

2002-05-12 09:30  takehiro

	* libmp3lame/quantize.c:
	  fix inc_scalefac_scale() ignoring preflag.

2002-05-11 18:49  markt

	* lame.spec.in:
	  new lame.spec.in file from Yosi.

2002-05-11 18:29  markt

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  applied Takehiro's bug fix for single channel bit reservoir usage
	  for pre-echo control.

2002-05-09 23:25  robert

	* libmp3lame/vbrquantize.c:
	  vbr-new:
	  more time consuming strategy to handle the case when more bits
	  used than available

2002-05-07 23:08  robert

	* libmp3lame/quantize_pvt.h, libmp3lame/vbrquantize.c:
	  allow best huffman divide to be used in vbr-new

2002-05-07 21:24  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  moved the tiny function ms_convert from quantize_pvt.c to
	  quantize.c--the
	  only client for it--giving a little speed increase

2002-05-07 21:11  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/vbrquantize.c:
	  better decoupling of vbrquantize.c and quantice.c

2002-05-07 20:21  robert

	* libmp3lame/bitstream.h, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.h, libmp3lame/util.h:
	  sorry Takehiro, we are back to the last known working version
	  from Sunday
	  13:35 UTC now

2002-05-07 20:15  robert

	* libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/reservoir.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c:
	  sorry Takehiro, we are back to the last known working version
	  from Sunday
	  13:35 UTC now

2002-05-06 16:03  takehiro

	* libmp3lame/quantize.c:
	  fix for new reservoir handling code with my ABR patch.
	  we don't need the totbits, because the reservoir size is reduced
	  in
	  ResvAdjust() in gr/ch loop and reservoir handler uses it.

2002-05-06 15:49  takehiro

	* ACM, ACM/ADbg, ACM/tinyxml:
	  added cvsignore to ignore Makefile, etc.

2002-05-06 15:30  takehiro

	* include/lame.h, libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/quantize.c, libmp3lame/reservoir.c:
	  cleanup. remove a compiling problem and some warnings.

2002-05-05 18:34  takehiro

	* libmp3lame/quantize.c:
	  one more outer_loop() fix for VBR.

2002-05-05 18:18  takehiro

	* libmp3lame/quantize.c:
	  outer_loop() fix for vbr_rh and vbr_mtrh

2002-05-05 17:37  takehiro

	* libmp3lame/quantize_pvt.c:
	  update comment

2002-05-05 17:17  takehiro

	* libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/encoder.c, libmp3lame/quantize.c,
	  libmp3lame/reservoir.c, libmp3lame/util.c, libmp3lame/util.h:
	  cleanup arround getframebits().

2002-05-05 16:50  takehiro

	* libmp3lame/quantize.c, libmp3lame/reservoir.c:
	  fix NORES_TEST. and works not only for CBR (don't tell me that's
	  useful :p)

2002-05-05 16:40  takehiro

	* libmp3lame/quantize.c, libmp3lame/reservoir.c,
	  libmp3lame/reservoir.h:
	  oops, I mis-checked in the old files. this is real one.

2002-05-05 16:29  takehiro

	* libmp3lame/quantize_pvt.c, libmp3lame/reservoir.c,
	  libmp3lame/reservoir.h:
	  clean up arround reservoir.

2002-05-05 15:47  takehiro

	* libmp3lame/quantize.c, libmp3lame/reservoir.c,
	  libmp3lame/reservoir.h:
	  trivial cleanup of arround resvoir.

2002-05-05 15:34  takehiro

	* libmp3lame/quantize.c:
	  cleanup arround inc_scalefac_scale() with new cod_info strcture.

2002-05-05 14:35  takehiro

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  cleanup arround calc_noise() with new cod_info strcture.

2002-05-05 13:59  takehiro

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c, libmp3lame/vbrquantize.c:
	  more cleanup of new cod_info structure, arround count_bits()

2002-05-05 13:48  takehiro

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.h:
	  inner_loop() is now static function.

2002-05-05 13:43  takehiro

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c:
	  changed outer_loop() algorithm. it needs lesser copy and runs
	  faster.
	  - try with new scalefactor combination on work area, and if we
	  get the better result, copy it to the returning area.
	  - make bin_search_StepSize() return the global_gain which assures
	  targ_bits > found_bits.

2002-05-05 13:33  robert

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h:
	  some clean-up in parameter passings due to recent change in
	  cod_info structure

2002-05-05 13:01  robert

	* libmp3lame/encoder.c, libmp3lame/quantize.c,
	  libmp3lame/quantize.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vbrquantize.h:
	  removed again some unused code I thought that was removed some
	  long time ago

2002-05-05 11:58  takehiro

	* libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/l3side.h, libmp3lame/quantize.c:
	  bit cleanup. removed window_switching_flag in l3side.

2002-05-05 11:53  takehiro

	* libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/vbrquantize.c:
	  To fix it does not call the "freorder" and to support for mixed
	  block,
	  the vbrquantize.c uses init_outer_loop()/init_xrpow() in
	  quantize.c to
	  initializatoin. this may fix the MPEG2 problem (but not tested).

2002-05-04 22:49  markt

	* configure, testcase.mp3:
	  updated testcase.mp3

2002-05-04 15:48  takehiro

	* libmp3lame/encoder.c, libmp3lame/quantize.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  clean up and speed up (optimize for new data structure).
	  - devided init_outer_loop() into init_outer_loop() and
	  init_xrpow(). This makes it easy to use GOGO-no-coda's xrpow
	  calculation routine.
	  # thanx Robert, your fix makes me aware my mistake in VBR mode.
	  
	  - restoring duplicated data of VBR loop are removed.
	  
	  more mixed_block prepare. freorder() in util.c and its caller in
	  encoder.c are moved into init_outer_loop() in quantize_pvt.c

2002-05-04 15:36  takehiro

	* libmp3lame/quantize.c:
	  trivial optimization

2002-05-04 14:54  takehiro

	* libmp3lame/bitstream.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/vbrquantize.c:
	  the signess process of l3_enc[] are moved from quantize*.c to
	  bistream.c. bit faster, much simpler.

2002-05-04 13:58  takehiro

	* libmp3lame/quantize_pvt.c:
	  removed unused variables.

2002-05-04 13:55  takehiro

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  removed log calculation of klemm_noise in main loop

2002-05-04 09:30  afaber

	* ChangeLog, Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  Added quality setting to lame_enc.dll interface

2002-04-30 22:20  robert

	* libmp3lame/version.h:
	  3.93 alpha 1 -> 3.93 alpha 2

2002-04-30 22:00  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  fixing latest floating point exceptions I got for some days
	  within the calc noise routine

2002-04-28 08:15  takehiro

	* libmp3lame/quantize_pvt.c:
	  Mark, because max_noise is initialized with 1E-20 and refreshes
	  Max(max_noise, sfb_noise), so it will never lesser than 1e-20.

2002-04-27 21:15  takehiro

	* libmp3lame/takehiro.c:
	  bit faster best_huffman_divide().
	  - use the pointer to the structure, not structure itself.

2002-04-27 20:58  takehiro

	* configure, libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/encoder.c, libmp3lame/l3side.h, libmp3lame/newmdct.c,
	  libmp3lame/newmdct.h, libmp3lame/quantize.c,
	  libmp3lame/quantize.h, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c, libmp3lame/vbrquantize.h:
	  large data structure update (1)
	  - mdct values (xr), encoded values (l3_enc), and scalefactors
	  (scalefac) are in the gr_info structure.
	  - it runs faster.
	  - reduce stack size but require large III_side_info_t and
	  lame_global_flags.

2002-04-27 20:41  takehiro

	* libmp3lame/machine.h:
	  fix --enable-all-float. when it enabled, FLOAT8_MAX should be
	  FLT_MAX about 1e37, not 1e99.

2002-04-27 18:24  takehiro

	* libmp3lame/bitstream.c:
	  trivial

2002-04-24 11:10  aleidinger

	* Makefile.unix:
	  add a mingw comment from Hubert Hanghofer

2002-04-22 17:51  markt

	* libmp3lame/quantize_pvt.c:
	  Takehiro, when you changed this code:
	  
	  res->max_noise = 10.*log10(Max(1e-20,max_noise ));
	  res->klemm_noise = 10.*log10(Max(1e-20,klemm_noise));
	  
	  to:
	  
	  res->max_noise = 10.*log10(max_noise);
	  res->klemm_noise = 10.*log10(klemm_noise);
	  
	  I started getting FPE (I always compile with --enable-debug
	  which aborts on FPE). I went ahead and changed it back- it should
	  have little effect.

2002-04-21 20:31  afaber

	* Dll/BladeMP3EncDLL.c:
	  Added additional debug flags

2002-04-21 18:10  takehiro

	* libmp3lame/l3side.h:
	  remove unused definition

2002-04-21 18:05  takehiro

	* libmp3lame/lame.c:
	  fix typo

2002-04-21 17:33  takehiro

	* libmp3lame/lame.c:
	  trivial(reduce magic number)

2002-04-21 17:03  afaber

	* Dll/BladeMP3EncDLL.c:
	  Avoid the VBR Tag space was written when VBR method was disabled
	  (caused empty frame in fornt of MPEG stream for certain CBR
	  settings)

2002-04-21 16:56  takehiro

	* frontend/parse.c, include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/quantize.c, libmp3lame/set_get.c,
	  libmp3lame/takehiro.c, libmp3lame/util.h:
	  new option, --substep, pseudo substep noise shaping now works.
	  - new set_get function, lame_set_substep(), lame_get_substep()
	  - final fix of "do substep" region of selected scalefactor.
	  - noise_shaping_amp == 3 is removed and the method of how to use
	  substep is selecte by the new member "substep" in structure
	  "lame_internal_flags".

2002-04-21 13:57  takehiro

	* libmp3lame/bitstream.c:
	  trivial optimization: gfp/gfc cleanup

2002-04-20 22:31  robert

	* Makefile.unix:
	  *** empty log message ***

2002-04-20 20:40  takehiro

	* libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/l3side.h, libmp3lame/newmdct.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c:
	  simplify the structure.
	  l3_side->gr[gr].ch[ch].tt is now known as l3_side->tt[gr][ch].

2002-04-20 20:31  takehiro

	* libmp3lame/l3side.h:
	  fix misleading indent

2002-04-20 20:28  takehiro

	* libmp3lame/takehiro.c:
	  prepare for mixed block support: count_bits() and
	  best_scalefac_store().
	  fix bug of noise_shaping_amp==3 on a long block.

2002-04-20 20:17  takehiro

	* libmp3lame/quantize.c:
	  more mixed_block preparation: inc_scalefac_scale(),
	  amp_scalefac_bands()

2002-04-20 20:16  takehiro

	* libmp3lame/quantize_pvt.c:
	  for better mixed_blcok support, make calc_xmin() use
	  cod_info->sfb_lmax, etc.

2002-04-20 19:54  takehiro

	* libmp3lame/quantize.c:
	  fix mixed mode (long block max sfb)

2002-04-20 19:50  takehiro

	* libmp3lame/l3side.h, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c:
	  more generalized mixed_block support and considering in MPEG2/2.5
	  - added new members to cod_info. psy_lmax and psy_smax.

2002-04-20 19:39  takehiro

	* libmp3lame/quantize.c:
	  we don't need two elements of l3_xmin.

2002-04-20 19:34  takehiro

	* libmp3lame/quantize_pvt.c:
	  - calc_xmin() supports mixed_block noise calculation.
	  - mixed_block parameter(long block end point and short block
	  start point) fix for calc_noise()

2002-04-20 18:51  takehiro

	* libmp3lame/quantize_pvt.c:
	  calc_noise() supports mixed_block noise calculation.

2002-04-20 18:40  takehiro

	* libmp3lame/quantize_pvt.c:
	  remove out-of-date comments

2002-04-20 18:37  takehiro

	* libmp3lame/machine.h, libmp3lame/quantize_pvt.c:
	  first step of clean up calc_noise() for preparing mixed block
	  support.
	  - removed unused variables
	  - make the two loops (long and short) same structure.

2002-04-20 11:49  robux4

	* ACM/lameACM.dsp:
	  Added the preset to compile the project (not used... yet)

2002-04-19 14:51  bouvigne

	* Dll/BladeMP3EncDLL.c:
	  Now the dll should properly applying mono and resampling settings
	  even if presets are used

2002-04-19 11:23  bouvigne

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  alt-presets now used from inside libmp3lame

2002-04-17 22:13  robert

	* Makefile.unix:
	  missing presets.c added

2002-04-17 08:11  aleidinger

	* configure:
	  regen

2002-04-17 08:08  aleidinger

	* configure.in:
	  bump version

2002-04-16 19:47  markt

	* testcase.mp3:
	  updated testcases (just because CVS version bumped up to
	  3.93alpha)

2002-04-15 18:36  markt

	* libmp3lame/version.h:
	  updated version to 3.93 alpha 1

2002-04-15 16:42  bouvigne

	* frontend/parse.c, libmp3lame/presets.c:
	  abr presets now into libmp3lame

2002-04-15 11:51  bouvigne

	* frontend/parse.c:
	  --cbr

2002-04-15 11:42  bouvigne

	* libmp3lame/lame.c:
	  3dnow identification

2002-04-15 08:18  aleidinger

	* Makefile.in, configure:
	  regen

2002-04-15 06:58  markt

	* Makefile.in, configure, debian/Makefile.in:
	  got automake 1.5 and ran it:

2002-04-15 06:57  markt

	* debian/Makefile.am:
	  updated debian/Makefile.am but my version of automake
	  is too old:
	  
	  debian/Makefile.am:5: require version 1.5, but have 1.4-p5

2002-04-15 06:47  markt

	* doc/html/history.html, libmp3lame/version.h, testcase.mp3:
	  bumped up version number to 3.92 for release

2002-04-15 06:46  markt

	* ChangeLog:
	  ran cvs2cl.pl

2002-04-14 12:21  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.def,
	  Dll/BladeMP3EncDLL.h:
	  Added strict-iso option to the interface, added output samplerate
	  parameter to the debug log file

2002-04-12 19:44  markt

	* libmp3lame/bitstream.c:
	  added message regarding "internal buffer inconsistency error" to
	  printout:
	  
	  ERRORF(gfc,"This is a fatal error. It has several possible
	  causes:");
	  ERRORF(gfc,"90% LAME compiled with buggy version of gcc using
	  advanced optimizations");
	  ERRORF(gfc," 9% Your system is overclocked");
	  ERRORF(gfc," 1% bug in LAME encoding library");

2002-04-10 12:55  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  --noasm option

2002-04-10 06:37  bouvigne

	* Makefile.MSVC:
	  presets.c

2002-04-09 15:43  bouvigne

	* Dll/BladeMP3EncDLL.c:
	  dm presets in the dll are now used from initialized from inside
	  libmp3lame

2002-04-09 15:28  aleidinger

	* libmp3lame/Makefile.in:
	  regen

2002-04-09 14:46  bouvigne

	* frontend/parse.c, include/lame.h, libmp3lame/Makefile.am,
	  libmp3lame/libmp3lame_vc6.dsp, libmp3lame/presets.c,
	  libmp3lame/set_get.c:
	  vbr presets are now part of libmp3lame

2002-04-09 11:56  bouvigne

	* frontend/parse.c:
	  --cbr switch

2002-04-08 20:30  robux4

	* ACM/acm.rc:
	  minor change on the DLL versioning

2002-04-08 18:43  robux4

	* ACM/ACM.cpp, ACM/AEncodeProperties.cpp:
	  Added some new defines that might not be in all platform DDKs
	  (just in case)

2002-04-08 18:41  aleidinger

	* ACM/Makefile.in, Makefile.in:
	  regen

2002-04-08 18:41  aleidinger

	* ACM/Makefile.am, Makefile.am:
	  add new vc6 file

2002-04-08 12:43  bouvigne

	* ACM/lameACM_vc6.dsp, lame_vc6.dsw:
	  added ACM codec to vc6 workspace

2002-04-08 09:56  bouvigne

	* lame_projects_vc6.dsp, lame_vc6.dsw:
	  dummy project to generate all lame projects (select lame_projects
	  in the workspace)

2002-04-07 14:09  robux4

	* ACM/TODO:
	  Removed the things already done

2002-04-07 14:07  robux4

	* ACM/ACM.cpp, ACM/AEncodeProperties.cpp, ACM/lameACM.dsp:
	  Clicking on the LAME URL will open the default browser

2002-04-07 13:31  robux4

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h:
	  Added tool-tips to the configuration dialog

2002-04-07 12:28  robux4

	* ACM/ACM.cpp, ACM/acm.rc:
	  Improved LAME version display

2002-04-07 10:17  robux4

	* ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h, ACM/acm.rc,
	  ACM/lame_acm.xml, ACM/resource.h:
	  ABR support now complete

2002-04-06 14:12  aleidinger

	* lame.bat:
	  bugfix, sorry, lost the mail for it

2002-04-06 14:11  aleidinger

	* ACM/Makefile.in, ACM/tinyxml/Makefile.in, Dll/Makefile.in,
	  Makefile.in, configure, frontend/Makefile.in,
	  libmp3lame/Makefile.in, mpglib/Makefile.in:
	  regen

2002-04-06 14:09  aleidinger

	* configure.in:
	  add tinyxml directory

2002-04-06 14:08  aleidinger

	* ACM/Makefile.am, ACM/tinyxml/Makefile.am, Dll/Makefile.am,
	  Makefile.am, frontend/Makefile.am, libmp3lame/Makefile.am,
	  mpglib/Makefile.am:
	  add new files

2002-04-06 14:01  robux4

	* ACM/ACM.cpp, ACM/ACM.h, ACM/ACMStream.cpp, ACM/ACMStream.h:
	  Encoding in ABR is starting to work
	  the different format strings are sorted correctly

2002-04-06 11:34  robux4

	* ACM/ACM.cpp, ACM/ACM.h, ACM/AEncodeProperties.cpp,
	  ACM/AEncodeProperties.h, ACM/acm.rc, ACM/resource.h:
	  First part of the ABR support
	  (missing use ABR settings on encoding when applied, configuration
	  of ABR params)

2002-04-06 11:32  robux4

	* ACM/tinyxml/readme.txt, ACM/tinyxml/tinyxml.cpp,
	  ACM/tinyxml/tinyxml.dsp, ACM/tinyxml/tinyxml.h,
	  ACM/tinyxml/tinyxmlparser.cpp, ACM/tinyxml/xmltest.cpp:
	  Updated with version 2.0 beta

2002-04-04 13:35  bouvigne

	* Dll/LameDll_vc6.dsp, frontend/lame_vc6.dsp, lame_vc6.dsw,
	  libmp3lame/libmp3lame_vc6.dsp, mpglib/mpglib_vc6.dsp:
	  I was annoyed to not be able to open the VC projects in vc6, so I
	  added vc6 ones

2002-03-22 18:29  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  Added lame presets to DLL interface

2002-03-16 00:26  robert

	* Makefile.MSVC:
	  no fast float to int conversion (Intel compiler)

2002-03-16 00:21  robert

	* Makefile.MSVC, frontend/parse.c:
	  fix for Win2k path seperator issue
	  found by Peter Hartley

2002-03-15 23:16  robert

	* libmp3lame/quantize.c:
	  bugfix, confusion about sideinfo length
	  found by "Kei Ishida"

2002-03-06 00:02  robert

	* lame.dsw:
	  VC Workspace for all(?) targets

2002-03-03 18:57  aleidinger

	* libmp3lame/psymodel.c:
	  remove /* withhin comment

2002-03-03 18:38  markt

	* TODO:
	  msvc6 notes

2002-02-22 11:53  shibatch

	* libmp3lame/quantize_pvt.c:
	  mp3x now displays quantization noise correctly when using
	  nspsytune.

2002-02-21 00:54  rbrito

	* debian/changelog, debian/control, debian/copyright,
	  debian/lame-dev.docs, debian/lame-dev.files,
	  debian/lame-extras.files, debian/lame.docs,
	  debian/liblame0-dev.docs, debian/liblame0-dev.files,
	  debian/liblame0.files, debian/rules:
	  Reorganization of the lame/debian directory for generating
	  packages closer
	  to what the Debian Policy specifies. Still not there (the
	  liblame0 package
	  still has a dreaded -rpath and thus, is not lintian clean).

2002-02-17 15:24  takehiro

	* doc/html/history.html, doc/html/switchs.html:
	  fix option description, and changed & to &amp;

2002-02-17 14:22  takehiro

	* libmp3lame/psymodel.c:
	  now inter-channel masking works even when without --nspsytune or
	  with -m s

2002-02-17 13:59  takehiro

	* libmp3lame/psymodel.c:
	  fix for --enable-all-float.

2002-02-17 13:55  takehiro

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/psymodel.c,
	  libmp3lame/set_get.c, libmp3lame/util.h:
	  added inter channel masking effect.
	  gfp->interChRatio is the ratio of inter-channel masking effect,
	  which could
	  be indicated by --interch X option.
	  The default value is 0.0, which means there's no inter-channel
	  masking effect.
	  
	  I think ratio = 0.01 is OK for lower bitrates (like under
	  128kbps),
	  for like old days songs by the Beatles.

2002-02-17 13:30  bouvigne

	* frontend/parse.c:
	  presets down to 8kbps

2002-02-17 08:12  takehiro

	* libmp3lame/util.h:
	  prepare for the interchannel masking and substep noise shaping.

2002-02-16 17:19  robux4

	* ACM/tinyxml/Makefile.tinyxml, ACM/tinyxml/changes.txt,
	  ACM/tinyxml/dox, ACM/tinyxml/makedistlinux,
	  ACM/tinyxml/makedistwin.bat, ACM/tinyxml/readme.txt,
	  ACM/tinyxml/tinyxml.cpp, ACM/tinyxml/tinyxml.dsp,
	  ACM/tinyxml/tinyxml.h, ACM/tinyxml/tinyxmlerror.cpp,
	  ACM/tinyxml/tinyxmlparser.cpp, ACM/tinyxml/xmltest.cpp:
	  Update with newer version of TinyXML

2002-02-15 11:42  shibatch

	* frontend/parse.c, libmp3lame/psymodel.c:
	  Now, nspsytune2 should work, maybe...

2002-02-14 15:38  shibatch

	* libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/util.h:
	  Preparation work for nspsytune2.

2002-02-13 16:30  aleidinger

	* frontend/rtp.h:
	  compile fix for gcc 3.x from Len Walter <len@unsw.edu.au>

2002-02-09 18:51  markt

	* frontend/gtkanal.c:
	  added preflag count to statics window in mp3x

2002-02-07 11:58  shibatch

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  preparation work for nspsytune2

2002-02-04 17:10  markt

	* TODO, libmp3lame/util.c:
	  temporary fix for resampling code. downsampling from 44101 to
	  44100 causes
	  a seg fault. Workaround in place for now: resampling disabled
	  if input/output samplerates agree to 4 digits.

2002-02-01 18:58  markt

	* INSTALL, TODO:
	  minor edits

2002-02-01 17:41  markt

	* TODO:
	  added notes about building a working, all encompasing MSVC6
	  workspace to TODO list

2002-02-01 17:37  markt

	* INSTALL:
	  added nots about the ACM codec and directshow filter
	  to INSTALL

2002-02-01 17:08  robux4

	* ACM/ACM.cpp:
	  Correct a problem when more than 2 channels are involved

2002-01-31 21:17  robux4

	* libmp3lame/lame.c:
	  Less memory copying on interleaved buffer

2002-01-31 21:03  uid32249

	* ACM/tinyxml/Makefile, ACM/tinyxml/Makefile.tinyxml:
	  replace Makefile with Makefile.tinyxml

2002-01-30 23:25  robux4

	* ACM/ACM.cpp, ACM/acm.rc:
	  resolved FhG incompatibility, first beta version (can go public),
	  no decoding yet

2002-01-29 21:20  robux4

	* ACM/ACM.cpp, ACM/TODO, ACM/acm.rc:
	  update version and removed unused buffers

2002-01-29 20:51  robux4

	* ACM/lameACM.dsp:
	  build with special dynamic memory handling for ACM

2002-01-29 20:37  robux4

	* ACM/ACM.cpp, ACM/ACMStream.cpp, ACM/main.cpp, configMS.h:
	  More debugging, special dynamic memory handling for ACM

2002-01-28 22:33  ivanski

	* mac/LAME.mcp:
	  CW6 Project File

2002-01-28 22:13  ivanski

	* mac, mac/.DS_Store, mac/LAME Carbon Debug.pch, mac/LAME Carbon
	  Final.pch, mac/LAME Classic Debug.pch, mac/LAME Classic
	  Final.pch, mac/MacDLLMain.c, mac/Precompile_Common.h:
	  Mac support

2002-01-28 20:12  robux4

	* ACM/TODO, ACM/acm.rc:
	  more TODO, added icon credit

2002-01-28 18:18  markt

	* libmp3lame/lame.c, test/CBRABR.op:
	  set devault qval to 2 for CBR and ABR.

2002-01-25 18:18  robux4

	* ACM/TODO, ACM/lame.ico:
	  new icon, more to do

2002-01-25 17:51  robux4

	* ACM/ACM.cpp, ACM/ACM.h, ACM/AEncodeProperties.cpp,
	  ACM/DecodeStream.cpp, ACM/acm.rc:
	  version 0.7.6, improved the smart output mode, corrected a config
	  saving bug

2002-01-25 17:47  robux4

	* ACM/ADbg/ADbg.cpp, ACM/ADbg/ADbg.h:
	  more explicit cast

2002-01-25 11:27  aleidinger

	* misc/Makefile.in:
	  regen

2002-01-25 11:23  aleidinger

	* misc/Makefile.am:
	  add lameid3.pl

2002-01-25 11:22  aleidinger

	* doc/man/lame.1:
	  some fixes

2002-01-24 19:38  robux4

	* ACM/ACM.cpp, ACM/ACM.h, ACM/ACMStream.cpp,
	  ACM/AEncodeProperties.cpp, ACM/AEncodeProperties.h, ACM/TODO,
	  ACM/acm.rc, ACM/lameACM.dsp, ACM/lame_acm.xml, ACM/main.cpp,
	  ACM/resource.h:
	  introducing smart output mode, clean dialog boxes, config
	  saving/loading now works

2002-01-23 20:51  robux4

	* ACM/DecodeStream.cpp:
	  correct a build problem when ENABLE_DECODING is not defined

2002-01-23 20:45  robux4

	* ACM/readme.txt:
	  More doc on the release files

2002-01-23 20:40  robux4

	* ACM/ADbg/ADbg.dsp:
	  removed a minor build problem

2002-01-23 20:39  robux4

	* ACM/ACM.cpp, ACM/ACM.dsw, ACM/ACM.h, ACM/ACMStream.cpp,
	  ACM/ACMStream.h, ACM/AEncodeProperties.cpp,
	  ACM/AEncodeProperties.h, ACM/DecodeStream.cpp,
	  ACM/DecodeStream.h, ACM/LameACM.inf, ACM/TODO, ACM/acm.rc,
	  ACM/lameACM.dsp, ACM/lame_acm.xml, ACM/readme.txt,
	  ACM/resource.h, ACM/tinyxml, ACM/tinyxml/Makefile,
	  ACM/tinyxml/changes.txt, ACM/tinyxml/readme.txt,
	  ACM/tinyxml/test.dsp, ACM/tinyxml/test.dsw,
	  ACM/tinyxml/tinyxml.cpp, ACM/tinyxml/tinyxml.dsp,
	  ACM/tinyxml/tinyxml.h, ACM/tinyxml/tinyxmlerror.cpp,
	  ACM/tinyxml/tinyxmlparser.cpp, ACM/tinyxml/xmltest.cpp:
	  Initial configuration dialog and XML config saving, initial
	  decoding support (not working yet)

2002-01-23 20:32  robux4

	* mpglib/interface.h:
	  better handling of C++ include

2002-01-23 01:23  markt

	* lame.spec.in:
	  commiting Yosi's new lame.spec.in file

2002-01-22 19:45  robux4

	* ACM/ACM.cpp, ACM/ACM.dsw, ACM/ACM.h, ACM/ACMStream.cpp,
	  ACM/ACMStream.h, ACM/LameACM.inf, ACM/TODO, ACM/acm.rc,
	  ACM/adebug.h, ACM/lameACM.def, ACM/lameACM.dsp, ACM/main.cpp,
	  ACM/readme.txt:
	  clean the code, put LGPL license, add mpglib to the build

2002-01-21 21:33  markt

	* TODO, configure, configure.in:
	  added MAX OSX patch from sf bugtrack page

2002-01-21 20:24  markt

	* USAGE, misc/lameid3.pl:
	  added lameid3.pl script to misc directory.
	  (use to reencode mp3's and preserve id3 tags)

2002-01-20 19:29  robux4

	* Dll/BladeMP3EncDLL.h, libmp3lame/util.h:
	  Better C++ handling

2002-01-20 15:05  aleidinger

	* configure:
	  regen (ACM)

2002-01-20 15:05  aleidinger

	* configure.in:
	  Add ACM directory

2002-01-20 15:04  aleidinger

	* ACM/Makefile.in:
	  Add generated autoconf glue

2002-01-20 14:59  aleidinger

	* ACM/ADbg/Makefile.in:
	  Add generated autoconf glueACM/Makefile.in

2002-01-20 14:58  aleidinger

	* ACM/ADbg/Makefile.am, ACM/Makefile.am:
	  Add automake clue (for make dist)

2002-01-20 14:56  robux4

	* ACM/lameACM.dsp:
	  Eanble correct nasm support

2002-01-20 14:03  robux4

	* ACM, ACM/ACM.cpp, ACM/ACM.dsw, ACM/ACM.h, ACM/ACMStream.cpp,
	  ACM/ACMStream.h, ACM/ADbg, ACM/ADbg/ADbg.cpp, ACM/ADbg/ADbg.dsp,
	  ACM/ADbg/ADbg.h, ACM/LameACM.inf, ACM/acm.rc, ACM/adebug.h,
	  ACM/lame.ico, ACM/lameACM.def, ACM/lameACM.dsp, ACM/main.cpp,
	  ACM/readme.txt, ACM/resource.h:
	  Initial ACM support

2002-01-19 11:52  bouvigne

	* README, TODO:
	  updated TODO

2002-01-17 16:43  aleidinger

	* Makefile.unix:
	  NASM for Windows fixes (Hubert Hanghofer)

2002-01-17 13:42  aleidinger

	* configure:
	  regen

2002-01-17 13:42  aleidinger

	* configure.in:
	  - Update gcc 3.x and --enable-expopt=full flags based upon
	  suggestions
	  and benchmarks from Hubert Hanghofer <hhanghof@netbeer.co.at> and
	  Craig Okruhlica <craig@velio.com>.
	  - Remove gcc flags which are marked 'experimental' in the gcc
	  3.0.3
	  documentation (noticed by Hubert Hanghofer).
	  - Add comment what "expopt" means, and what it doesn't mean.

2002-01-17 13:42  aleidinger

	* Makefile.unix:
	  - Update MINGW part pased upon patches and comments from
	  Hubert Hanghofer <Hubert.Hanghofer@m-real.com>.
	  - Some minor changes for consistency.
	  
	  Note: this file is superseded by configure, except for systems
	  where
	  configure didn't work, so if something is broken you have to send
	  patches.

2002-01-17 05:08  takehiro

	* libmp3lame/i386:
	  added *.lo

2002-01-09 12:09  takehiro

	* debian, dshow:
	  added autogen Makefile to ignore

2002-01-09 12:09  takehiro

	* frontend, libmp3lame, mpglib:
	  added *.il, for the intel compiler with ipo option

2002-01-07 15:51  aleidinger

	* configure:
	  oops, forgot to s/gtk12-config/gtk-config/

2002-01-07 15:48  aleidinger

	* configure:
	  regen

2002-01-07 15:48  aleidinger

	* configure.in:
	  add gcc 3 specific optimizations, based upon suggestions by
	  Sébastien FORESTIER <forestie@caramail.com>

2002-01-07 10:37  aleidinger

	* configure, frontend/Makefile.in:
	  regen: Fix Cygwin and HPUX problems.

2002-01-07 10:36  aleidinger

	* configure.in, frontend/Makefile.am:
	  Fix Cygwin and HPUX problems.

2002-01-01 16:55  robert

	* misc/lame4dos.bat:
	  modified some remark

2001-12-30 23:54  markt

	* doc/html/history.html:
	  updated hostory.html from master copy in webpages project

2001-12-29 17:36  takehiro

	* doc/html/history.html:
	  fix option for pseudo-half step quantization

2001-12-29 16:23  aleidinger

	* configure:
	  regen: Welcome to 3.92

2001-12-29 16:21  aleidinger

	* configMS.h, configure.in, libmp3lame/version.h:
	  Welcome to 3.92

2001-12-29 15:45  aleidinger

	* doc/html/history.html:
	  sync with webpages/history.html

2001-12-28 18:55  takehiro

	* .:
	  added stamp-h1

2001-12-28 12:47  aleidinger

	* Dll/Example.dsp, Dll/MP3EncDll.dsp, dshow/dshow.dsp,
	  dshow/dshow.dsw, frontend/lame.dsp, libmp3lame/libmp3lame.dsp,
	  mpglib/mpglib.dsp:
	  unix->dos

2001-12-27 16:12  aleidinger

	* doc/man/lame.1:
	  add --alt-preset

2001-12-27 12:50  aleidinger

	* misc/Makefile.in:
	  regen: add missing files to the distribution

2001-12-27 12:49  aleidinger

	* misc/Makefile.am:
	  add missing files to the distribution

2001-12-23 20:19  dibrom

	* libmp3lame/set_get.c:
	  disable noise shaping 2 for "fast" extreme mode also

2001-12-23 20:11  dibrom

	* libmp3lame/set_get.c, libmp3lame/vbrquantize.c:
	  fix small bug and completely disable noise_shaping 2 with "fast"
	  mode until a better solution can be found.

2001-12-23 12:25  aleidinger

	* misc/Makefile.in:
	  regen: add some files to the distribution

2001-12-23 12:22  aleidinger

	* misc/Makefile.am:
	  add auenc, mugeco,sh and mlame to distrib

2001-12-22 13:32  aleidinger

	* DEFINES:
	  document NON_LINEAR_PSYMODEL, remove KLEMM_43

2001-12-22 13:29  aleidinger

	* libmp3lame/psymodel.c:
	  non linear psymodel (disabled by default) for GPSYCHO (nspsytune
	  is untouched)

2001-12-22 13:27  aleidinger

	* configure:
	  regen: Welcome to v3.91

2001-12-22 13:26  aleidinger

	* configMS.h, configure.in, libmp3lame/version.h:
	  Welcome to v3.91

2001-12-21 15:09  aleidinger

	* libmp3lame/machine.h:
	  add icc check

2001-12-21 15:05  aleidinger

	* libmp3lame/Makefile.in:
	  regen: (add vbrquantize.h)

2001-12-21 15:05  aleidinger

	* libmp3lame/Makefile.am:
	  add vbrquantize.h

2001-12-21 11:58  jd-

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.def,
	  Dll/BladeMP3EncDLL.h:
	  Added DLL interface function beEncodeChunkFloatS16NI for floating
	  point audio
	  Samples are input within the range +/- 32768, in non-interleaved
	  channels

2001-12-21 03:29  markt

	* testcase.mp3:
	  updated testcase

2001-12-21 03:29  markt

	* libmp3lame/version.c, libmp3lame/version.h:
	  updated version to lame 3.90

2001-12-20 23:11  markt

	* ChangeLog:
	  update changelog

2001-12-20 08:44  dibrom

	* libmp3lame/encoder.c:
	  no message

2001-12-20 08:33  dibrom

	* libmp3lame/encoder.c, libmp3lame/set_get.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  improved quality of "fast" modes in certain situations

2001-12-18 17:44  dibrom

	* frontend/parse.c:
	  removed bitrate compensation stuff for the abr bitrates since it
	  appears no longer necessary

2001-12-18 17:28  dibrom

	* doc/html/switchs.html:
	  added --alt-preset to the list of switches

2001-12-18 17:13  dibrom

	* doc/html/history.html:
	  documented many changes which have been made over 3.90

2001-12-18 16:27  dibrom

	* frontend/parse.c:
	  small tweak to "standard" preset

2001-12-18 16:14  dibrom

	* libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/set_get.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  small quality improvement for "extreme" and "insane", more
	  bitrate tunings for "extreme" -- should average to 256kbps more
	  often

2001-12-18 11:16  aleidinger

	* configure:
	  regen: vorbis changes

2001-12-18 11:14  aleidinger

	* configure.in:
	  - Change vorbis default to disabled
	  - Added -logg in the vorbis case (seems to be needed, I haven't
	  tested it)

2001-12-18 11:13  aleidinger

	* libmp3lame/vorbis_interface.c:
	  Try to unbreak vorbis support (it can't get worser).
	  You may need the vorbis source and add
	  -I/path/to/vorbis_src/includes/ to
	  CFLAGS prior to running configure
	  
	  Tested by: Jonathan Ryshpan <jonrysh@pacbell.net> on RedHat 7.2
	  Submitted by: Cyrille Lefevre <clefevre@citeweb.net> via FreeBSD
	  problem report

2001-12-17 04:40  markt

	* libmp3lame/quantize.c:
	  Minor changes to Gabriel's adjustment for ABR bitrate.
	  Updated testcases.
	  
	  Added the following comments:
	  
	  /*
	  res_factor is the percentage of the target bitrate that should
	  be used on average. the remaining bits are added to the
	  bitreservoir and used for difficult to encode frames.
	  
	  Since we are tracking the average bitrate, we should adjust
	  res_factor "on the fly", increasing it if the average bitrate
	  is greater than the requested bitrate, and decreasing it
	  otherwise. Reasonable ranges are from .9 to 1.0
	  
	  Until we get the above suggestion working, we use the following
	  tuning:
	  compression ratio res_factor
	  5.5 (256kbps) 1.0 no need for bitreservoir
	  11 (128kbps) .93 7% held for reservoir
	  
	  with linear interpolation for other values.
	  
	  */
	  res_factor = .93 + .07 * (11.0 - gfp->compression_ratio) / (11.0
	  - 5.5);

2001-12-16 16:00  bouvigne

	* libmp3lame/quantize.c:
	  better abr bitrate approximation (will break abr testcases)

2001-12-15 07:31  dibrom

	* libmp3lame/psymodel.c, libmp3lame/set_get.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  More tunings for "fast" options for "standard" and "extreme"
	  modes.

2001-12-14 16:27  dibrom

	* libmp3lame/set_get.c:
	  small tweak to "extreme" mode

2001-12-14 07:00  dibrom

	* frontend/parse.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/set_get.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  - changed --dm-presets to --alt-presets
	  - modified preset system
	  - applied tunings to vbr and psymodel, especially with
	  --alt-preset standard, improves many difficult clips over the
	  default vbr modes/presets (even those such as --r3mix) such as
	  impulse cases (fatboy, spahm, them, gbtinc, ravebase, etc),
	  pre-echo cases (castanets, death2, florida_seq, etc), tonality/js
	  cases (2nd_vent, serioustrouble, bloodline, etc), low volume
	  clips (piano, rach_original), many clips with problems with noise
	  shaping 2, and much much more. Practically all cases should be
	  equal or improved in quality.
	  - added --no-preset-tune to disable tunings for --alt-preset
	  "modes"

2001-12-03 18:15  markt

	* USAGE, frontend/parse.c:
	  updated descriptions of some command line options, as per Dmitry

2001-11-30 18:35  markt

	* lame.spec.in:
	  patch from lassauge@mail.dotcom.fr for lame.spec.in file.

2001-11-30 18:05  markt

	* include/lame.h, libmp3lame/lame.c:
	  added lame_encode_buffer_long2(), since lame_encode_buffer_long()
	  had
	  a scaling which restricted the precision of the input to the same
	  as if the input was 'short int'
	  
	  decided to leave the original routine, on the off chance there
	  is some program out there which uses it.

2001-11-28 12:30  aleidinger

	* Makefile.in:
	  regen

2001-11-28 12:29  aleidinger

	* Makefile.am:
	  change description of make test

2001-11-26 15:21  takehiro

	* libmp3lame/takehiro.c:
	  tuning of "pseudo half step noise shaping" with listening test.
	  and it works now even when scalefac_scale == 1

2001-11-26 15:20  takehiro

	* libmp3lame/VbrTag.c, libmp3lame/psymodel.c:
	  fix warning

2001-11-26 11:00  aleidinger

	* libmp3lame/version.h:
	  Bump alpha version.

2001-11-25 15:24  takehiro

	* libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/takehiro.c,
	  libmp3lame/util.h:
	  new noise shaping algorithm(pseudo half step noise shaping)
	  it brings slow but more accurate noise shaping.
	  enables with -q0 when CBR

2001-11-19 21:26  markt

	* frontend/parse.c, libmp3lame/encoder.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  Reverted CVS back to before "gglower" option was added.
	  
	  --scale is a better way to do this, and contrary to Hans' claims,
	  --scale is also lossless.

2001-11-18 22:44  robert

	* frontend/parse.c, libmp3lame/encoder.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  patch by Hans van der Heijden, original email:
	  
	  Datum: Sun, 18 Nov 2001 20:13:11 +0100
	  Von: "Hans van der Heijden" <H.Heijden@inter.NL.net>
	  An: <Robert.Hegemann@gmx.de>
	  
	  Hi Robert,
	  
	  For the first time, I compiled Lame, being a non-programmer
	  (well, my work
	  involves programming, but that's a world apart with logic
	  symbols, not C
	  language).
	  Because sometimes I have the urge to try something with the
	  sourcecode.
	  
	  I'd like to add a switch to avoid the use of --scale for
	  preventing
	  clipping. It manipulates global_gain, similar as the mp3gain tool
	  does, but
	  this would be much more convenient of course (and better
	  quality-wise, being
	  'lossless').
	  
	  Since I have no programming experience, can you take a look if I
	  implemented
	  it correctly? The switch is called --gglower x, where each x-step
	  stands for
	  1,5 dB. gg is short for global gain. So for example, --gglower 1
	  will
	  decrease the output volume with 1,5 dB (0,841). I would probably
	  use this by
	  default, it will be enough to prevent clipping for all but the
	  oddest
	  samples/settings.
	  
	  I did a lot of copy/pasting from the --athlower switch, otherwise
	  I probably
	  couldn't figure it all out!
	  Files involved are parse.c, lame_global_flags.h, set_get.c,
	  lame.h and
	  bitstream.c.
	  I attached these files, if you search in each for 'gglower'
	  you'll find the
	  added code.
	  Files came from the 11 Nov CVS.
	  
	  Also attached is a cygwin lame.exe if you like to play with it.
	  
	  Let's add it to CVS if it's OK (don't know how to do that myself,
	  have to
	  tell Mark?).
	  
	  Bye,
	  Hans

2001-11-17 11:17  aleidinger

	* Dll/Example.cpp, Dll/Makefile.mingw32, Makefile.unix:
	  some mingw fixes from Vladislav Naumov <vnaum@echo.ru>

2001-11-14 16:25  aleidinger

	* include/lame.h:
	  fixes in comments

2001-11-12 21:58  robert

	* libmp3lame/lame.c:
	  *** empty log message ***

2001-11-12 17:36  aleidinger

	* include/lame.h:
	  spelling fixes in comments

2001-11-11 22:13  afaber

	* Dll/BladeMP3EncDLL.c:
	  fixed bugs: 451879 and 438835

2001-11-11 15:09  aleidinger

	* configure, doc/man/Makefile.in:
	  regen

2001-11-11 15:06  aleidinger

	* libmp3lame/bitstream.c:
	  add CVS Id

2001-11-11 14:57  aleidinger

	* Dll/Makefile.in, Makefile.in, debian/Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  regen

2001-11-11 14:53  aleidinger

	* configure.in, frontend/Makefile.am:
	  bugfix: --disable-{mp3x,mp3rtp} had build them
	  add: --disable-fontend

2001-11-11 14:48  aleidinger

	* include/lame.h, libmp3lame/version.c, libmp3lame/version.h:
	  remove mp3x version, it doesn't belong into the lib
	  (this breaks every app which uses the function, but mp3x should
	  be the only app which uses it, so this is ok because I moved it
	  into gtkanal.c)

2001-11-11 14:47  aleidinger

	* frontend/gtkanal.c:
	  move mp3x version here (from libmp3lame/version.[ch])

2001-11-11 11:31  aleidinger

	* libmp3lame/Makefile.in:
	  regen

2001-11-11 11:30  aleidinger

	* libmp3lame/Makefile.am:
	  add missing headers

2001-11-10 14:25  aleidinger

	* libmp3lame/VbrTag.c:
	  at least fix the bus error on SPARC until Roel or someone else
	  decides if we should fix it in another way

2001-11-06 14:43  aleidinger

	* doc/html/switchs.html:
	  resample only for encoding

2001-11-06 12:41  aleidinger

	* doc/man/lame.1:
	  athaa fixes, resample only for encoding

2001-11-06 01:31  jd-

	* frontend/parse.c:
	  Cleaned up preset help; send explicitly requested help to stdout,
	  not stderr.
	  Modularized original presets, with aliases to preserve the
	  original behavior.

2001-11-05 15:26  dibrom

	* frontend/parse.c:
	  modify lowpass values for dm-preset abr mode

2001-11-05 05:46  dibrom

	* frontend/parse.c:
	  Updated dm-presets

2001-11-05 05:20  markt

	* libmp3lame/encoder.h, testcase.mp3:
	  updated padding at end of MP3 to be 1152 samples.
	  So the last frame containing real data will always be followed
	  by a frame containing silence. (since most decoders will only
	  decode the second to last frame)

2001-11-04 23:42  robert

	* Makefile.MSVC:
	  lost i added

2001-11-04 23:39  robert

	* Makefile.MSVC:
	  optimizing some intel compile flags

2001-11-04 18:52  robert

	* DEFINES, libmp3lame/quantize.h:
	  removed MAXQUANTERROR precompiler define

2001-11-04 18:51  robert

	* configMS.h:
	  moved MSVC specific #pragma into MSVC section

2001-11-04 18:48  robert

	* frontend/parse.c, include/lame.h:
	  we don't want to make Naoki's msfix part of the API at this time,
	  removed
	  from lame.h again.
	  parse.c compilation was broken by JD's presets_alias function.

2001-11-04 18:31  robert

	* libmp3lame/encoder.c, libmp3lame/quantize.c,
	  libmp3lame/vbrquantize.c, libmp3lame/vbrquantize.h:
	  removed MAXQUANTERROR precompiler define

2001-11-04 18:20  dibrom

	* frontend/parse.c:
	  update dm-presets.. more to come shortly

2001-11-04 18:19  dibrom

	* include/lame.h:
	  added msfix

2001-11-04 17:28  robert

	* include/lame.h, libmp3lame/lame.c:
	  --vbr-mtrh is now known as --vbr-new

2001-11-04 17:22  robert

	* Dll/BladeMP3EncDLL.c:
	  --vbr-mtrh is now known as --vbr-new

2001-11-04 13:39  aleidinger

	* libmp3lame/set_get.c:
	  some fixes in the comments

2001-11-04 13:39  aleidinger

	* frontend/parse.c:
	  mark --{voice,radio,cd,studio,phone} as deprecated

2001-11-04 13:33  aleidinger

	* doc/man/lame.1:
	  remove --voice (deprecated)

2001-11-04 13:01  jd-

	* frontend/parse.c:
	  Moved "--dm-preset *" preset profiles to "--preset dm-*".
	  Added code for aliasing presets to multiple broad or narrow
	  purpose labels.

2001-11-04 08:37  jd-

	* frontend/parse.c:
	  Allow specifying the r3mix preset with "--preset r3mix" for
	  consistency

2001-11-03 10:51  jd-

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c:
	  Renamed experimental switch "--ath-adjust" to "--athaa-type" for
	  consistency.
	  All switches directly affecting ATH auto-adjustment begin with
	  "--athaa-*".

2001-11-03 09:31  jd-

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/psymodel.c, libmp3lame/set_get.c, libmp3lame/util.h:
	  Added a bit of documentation to the long help.
	  To better reflect purpose, renamed adapt-thres-type and
	  adapt-thres-level
	  to athaa-loudapprox and athaa-sensitivity, respectively.

2001-10-30 20:31  markt

	* libmp3lame/VbrTag.c:
	  When determining enc_delay and enc_padding from VBR header, check
	  for reasonable values.

2001-10-29 22:00  markt

	* frontend/get_audio.c, frontend/main.c, frontend/main.h,
	  frontend/parse.c, include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/VbrTag.h, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, mpglib/interface.c,
	  mpglib/mpglib.h:
	  VbrTag.c: GetVBRTag() updated to parse enc_delay, enc_padding
	  fields
	  enc_delay is used by 'lame --decode' to determine how many
	  samples
	  to skip at the beginning during decoding.

2001-10-29 20:01  markt

	* frontend/main.c:
	  for --nogap option, do not allow Xing VBR tag.
	  
	  We should also not allow id3 tags and --nogap, but the are not
	  turned on by
	  default, and it is more than a 3 line fix :-)

2001-10-29 08:41  bouvigne

	* libmp3lame/gain_analysis.c, libmp3lame/gain_analysis.h:
	  adding gain analysis files

2001-10-29 06:08  markt

	* frontend/main.c, libmp3lame/VbrTag.c,
	  libmp3lame/lame_global_flags.h, mpglib/interface.c:
	  --nogap and --nogapout fixed in frontend. didn't check
	  if output is really gapless - hopefully no bugs have crept into
	  that while --nogap was broken.

2001-10-28 22:24  markt

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/tables.c,
	  libmp3lame/util.c, mpglib/interface.c:
	  Fixed bug in VbrTag.c which is why Xing tag was broken for MPEG2
	  files.
	  
	  This code:
	  
	  if( h_id == 0 )
	  pTagData->samprate >>= 1;
	  
	  was, for unknown reasons, cutting the samplrate down by
	  a factor of two for MPEG2, causing the framesize of the Xing
	  header to be too large, causing it to overwrite the first
	  valid MP3 frame.

2001-10-25 11:54  roelvdb

	* libmp3lame/VbrTag.c:
	  undo previous edit, misunderstanding

2001-10-24 23:59  roelvdb

	* libmp3lame/VbrTag.c:
	  if ABR, {store bitrate <=255} else { use 4 low bits to store
	  bitrateindex so "-b" for VBR(/CBR) is known. 0F for freeformat}

2001-10-24 20:45  markt

	* testcase.mp3:
	  updated testcases

2001-10-24 16:33  aleidinger

	* doc/man/lame.1:
	  sync with HTML

2001-10-24 12:44  bouvigne

	* doc/html/basic.html, doc/html/examples.html,
	  doc/html/history.html, doc/html/node6.html,
	  doc/html/switchs.html:
	  doc fixes

2001-10-24 12:32  bouvigne

	* doc/html/switchs.html:
	  a small athtype doc modif

2001-10-23 16:35  bouvigne

	* doc/html/switchs.html:
	  fixes from Dmitry

2001-10-23 16:17  bouvigne

	* doc/html/switchs.html:
	  fixes from Alexander

2001-10-23 14:06  aleidinger

	* doc/man/lame.1:
	  add line breaks, expand examples section, add ID3 section, fixes

2001-10-23 14:06  aleidinger

	* frontend/main.c:
	  fix for cygwin/mingw from egor duda <deo@logos-m.ru>

2001-10-22 21:31  roelvdb

	* libmp3lame/VbrTag.c:
	  no message

2001-10-22 13:25  aleidinger

	* doc/man/lame.1:
	  update based on HTML version

2001-10-21 17:00  bouvigne

	* doc/html/basic.html, doc/html/switchs.html:
	  doc

2001-10-20 16:32  aleidinger

	* Makefile.unix:
	  Submitted by: "Vladislav Naumov" <vnaum@inbox.ru>
	  
	  I canged number 2 of his fixes to be equal to configure.in
	  (-O3 is the highest official supportes optimization option, some
	  of
	  the options are superfluous (because they are included in -O3),
	  some may
	  perhaps change the ABI (ABI, not API!), e.g. -malign-*).
	  
	  Quote:
	  1. NOUNIXCMD = YES (don't use shell & sed)
	  2. CC_OPTS are set to produce optimized code. Somewhat faster.
	  3. frontend_sources & lib_sources are now fixed.
	  GNU Make version 3.79.1 croaked about recursive variable.
	  4. ASFLAGS are correctly set, 'make -fMakefile.unix UNAME=MSDOS
	  HAVE_NASM=YES'
	  works smoothly (if you have nasm.exe, of course :))

2001-10-20 16:28  aleidinger

	* Dll/Makefile.mingw32:
	  Submitted by: "Vladislav Naumov" <vnaum@inbox.ru>
	  
	  Quote:
	  This makefile compiles lame_enc.dll with mingw32 (and possibly
	  cygwin)
	  Of course, you must first build ../libmp3lame/libmp3lame.a.
	  liblame_enc.a can be used to link the lame_enc.dll to your
	  programs.
	  Tested with EAC 0.9pb9 (my own favorite,
	  http://www.exactaudiocopy.de/).

2001-10-19 17:39  bouvigne

	* libmp3lame/VbrTag.c:
	  headering...

2001-10-18 19:55  bouvigne

	* doc/html/switchs.html, frontend/parse.c:
	  documenting...hopefully nearly finished

2001-10-17 19:29  robert

	* frontend/main.c:
	  patch by Gian-Carlo Pascutto
	  
	  
	  I saw someone reported a problem with my --nogapout
	  switch on the mailinglist. I have done a quick attempt
	  to fix the problems (and address another potential one).
	  
	  I haven't actually tested the fixes, but I'm reasonably
	  sure they are correct, or at least they show how to
	  address the problem.

2001-10-17 13:23  roelvdb

	* libmp3lame/VbrTag.c:
	  after discussing with Gabriel added some more AQ flag refinements
	  (--athonly,--noath,--scale-*, free format)
	  
	  and
	  
	  made #define MAXFRAMESIZE 2880, because LAME used to fill the
	  freeformat 640 LAME Tag with junk at the end.

2001-10-17 11:31  aleidinger

	* configMS.h:
	  fix VERSION after last commit

2001-10-17 01:41  roelvdb

	* libmp3lame/VbrTag.c:
	  added overflow protection to that lowpass field in the LAME Tag,
	  just in case someone enters a very high lowpass value.

2001-10-17 01:18  roelvdb

	* libmp3lame/lame.c:
	  44.1kHz:
	  instead of "transition band: 22228 Hz - 22761 Hz" lowpass filter,
	  "-b320 -a" (mono) will now give "transition band: 21517 Hz -
	  22050 Hz"
	  (as in the stereo case)
	  
	  likewise on other samplerates. I can imagine this can be done in
	  a much more intelligent fashion.
	  
	  also commented out a bit of similar code that seems obsolete now?

2001-10-16 18:49  bouvigne

	* doc/html/switchs.html:
	  documenting - not finished

2001-10-16 18:07  afaber

	* configMS.h, include/lame.h, libmp3lame/VbrTag.h,
	  libmp3lame/bitstream.c, libmp3lame/fft.c, libmp3lame/fft.h,
	  libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h:
	  Fixed some compiler warnings (mainly differences between function
	  declaration and definition)

2001-10-16 15:00  roelvdb

	* libmp3lame/VbrTag.c:
	  added the needed "else", fixed lowpass, added -V0/V1, --nores,
	  -k, ...

2001-10-16 10:36  bouvigne

	* doc/html/history.html:
	  history

2001-10-15 13:44  roelvdb

	* libmp3lame/VbrTag.c:
	  fixed a wrong bracket causing >=256 CBR not to be AQ
	  added ABR>=256 S as AQ, consistent with CBR >=256S

2001-10-13 19:00  markt

	* testcase.mp3:
	  updated testcases

2001-10-13 14:02  aleidinger

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/set_get.c,
	  libmp3lame/util.c:
	  add --scale-l and --scale-r options
	  
	  Requested by: fasttimes@mochamail.com
	  Written by: mp3gain@hotmail.com

2001-10-13 14:00  aleidinger

	* doc/man/lame.1:
	  update from "lame --longhelp"

2001-10-13 14:00  aleidinger

	* USAGE, doc/html/switchs.html:
	  add --scale-l, --scale-r

2001-10-12 18:39  bouvigne

	* doc/html/index.html, doc/html/switchs.html:
	  a few doc updates - not yet finished

2001-10-11 20:06  bouvigne

	* USAGE, doc/html/basic.html, doc/html/examples.html,
	  doc/html/history.html, doc/html/id3.html, doc/html/modes.html,
	  doc/html/node6.html, doc/html/switchs.html:
	  *** empty log message ***

2001-10-11 12:21  aleidinger

	* libmp3lame/encoder.c, libmp3lame/util.h:
	  fix misunderstanding of gfc->padding

2001-10-10 02:17  roelvdb

	* libmp3lame/VbrTag.c, libmp3lame/lame.c:
	  consequence, it's "LAME Tag", longhelp said Xing header

2001-10-10 02:15  roelvdb

	* frontend/parse.c:
	  consequence, it's "LAME Tag"

2001-10-10 02:00  roelvdb

	* frontend/main.c:
	  added 13 " if (silent < 10) "s to make "lame --decode --quiet"
	  really silent. there's still a LF I can't locate though...

2001-10-10 01:42  roelvdb

	* Dll/BladeMP3EncDLL.c:
	  removed experimental X from --r3mix

2001-10-09 23:09  roelvdb

	* libmp3lame/VbrTag.c:
	  adds delays/padding to tag, no range check and tested exclusively
	  on windows

2001-10-09 22:34  roelvdb

	* frontend/parse.c:
	  test commit CVS to fix --r3mix size increase (robert identified
	  is as -X being used)

2001-10-09 14:30  aleidinger

	* libmp3lame/version.h:
	  bump alpha version

2001-10-09 14:26  aleidinger

	* Dll/BladeMP3EncDLL.c, frontend/parse.c, libmp3lame/VbrTag.c,
	  libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/util.h:
	  Use the official enum names from lame.h instead of numbers or
	  defines.

2001-10-08 12:54  bouvigne

	* libmp3lame/util.c:
	  athtype fix

2001-10-06 19:25  markt

	* include/lame.h:
	  added notes about scaling. there are some inconsistencies in
	  the API:
	  
	  lame_encode_buffer() takes short int as input, range +/- 32768
	  lame_encode_float() takes float, but data must be in the range
	  +/- 32768
	  lame_encode_double() takes double, but data must be in the range
	  +/- 32768
	  
	  lame_encode_int() takes int, but data must be in the range of
	  'int'
	  for 32bit ints, this is +/- 2147483684

2001-10-06 19:17  markt

	* frontend/get_audio.c:
	  updated error messages we now support 8,16,24 and 32 bit input

2001-10-06 19:05  markt

	* frontend/get_audio.c, frontend/gtkanal.c, frontend/parse.c,
	  include/lame.h, libmp3lame/lame.c:
	  added support for 24 and 32 bit when not using libsndfile.
	  jd had already added the code to do this, just had to make
	  the libsndfile --bitwidth option available when not using
	  libsndfile. Also updated unpack_read_samples to support 32bit.
	  
	  Could not test this since I dont have any 24 bit files, and
	  sox cant seem to create them.

2001-10-05 22:43  robert

	* libmp3lame/lame.c, libmp3lame/machine.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/vbrquantize.c:
	  using -Xn similar to the old code to select different noise
	  measurings with
	  the newer VBR code. the actual mapping is not final yet.

2001-10-05 13:00  bouvigne

	* libmp3lame/util.c:
	  ath 4 switches to ath 2 in cbr mode

2001-10-05 12:59  bouvigne

	* libmp3lame/lame.c:
	  made NsPsytune stopping using ath0 as default, as everyone is
	  overriding the athwhen using nspsytune

2001-10-05 12:20  bouvigne

	* Makefile.MSVC, libmp3lame/libmp3lame.dsp:
	  makefile.msvc and projects on par?

2001-10-04 10:01  aleidinger

	* libmp3lame/VbrTag.c, libmp3lame/VbrTag.h:
	  fixes for Solaris 8 / add missing include

2001-10-02 16:02  markt

	* frontend/main.c, frontend/parse.c:
	  added nogap patch from Gian-Carlo Pascutto.
	  
	  lame --nogapout <outputdir> --nogap file1.wav file2.wav file3.wav
	  ...
	  
	  untested.

2001-10-02 03:54  markt

	* configure:
	  ran autoconf to make a new configure script

2001-10-02 03:53  markt

	* INSTALL, configure.in, frontend/main.c, frontend/parse.c,
	  libmp3lame/lame.c, libmp3lame/vorbis_interface.c:
	  split Vorbis support into two pieces:
	  
	  1.
	  HAVE_VORBIS (decoding support). This code still works!
	  (at least it compiles - I have't actually run it).
	  Can be enabled with ./configure --with-vorbis. Changed the
	  default in configure.in to define HAVE_VORBIS
	  
	  2.
	  HAVE_VORBIS_ENCODER
	  This is the code which hasn't compiled since Vorbis beta1.
	  Mainly because how you specify varous encoding "modes" has been
	  changed. It's probably not to much work to get this working,
	  but there is no reason for this to be in LAME.

2001-10-02 00:00  markt

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c, libmp3lame/lame.c:
	  Music CRC computed on the fly. Code stubs in place for
	  decoding on the fly.

2001-10-01 23:02  markt

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/util.h:
	  started work to put in the ability to compute MusicCRC on
	  the fly, as well as decoding on the fly.

2001-09-30 20:36  robert

	* frontend/parse.c, libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/set_get.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  removed some unused variable

2001-09-28 21:29  markt

	* frontend/main.c:
	  cpu affinity back to #define _WIN32

2001-09-28 21:13  markt

	* frontend/main.c:
	  process affinity patch is back

2001-09-28 20:46  markt

	* USAGE:
	  more small updated to USAGE

2001-09-28 20:38  markt

	* USAGE:
	  updated USAGE to reflect Dmitry's -B comments

2001-09-28 12:04  bouvigne

	* Makefile.MSVC:
	  harmonization between makefile and project

2001-09-27 13:33  bouvigne

	* libmp3lame/libmp3lame.dsp:
	  3dnow fft assembly

2001-09-21 04:50  markt

	* frontend/main.c, testcase.mp3:
	  updated testcases (lowpass added to Info tag?)

2001-09-21 04:03  markt

	* frontend/main.c:
	  disabled CPU affinity code.
	  should only be in a #ifdef NT-system, but I dont know what
	  the ifdef should be.

2001-09-20 08:38  bouvigne

	* libmp3lame/libmp3lame.dsp, mpglib/mpglib.dsp:
	  corrected dsp files for VC++. Should work also under VC5, but not
	  tested. Please report success/error

2001-09-20 07:15  bouvigne

	* libmp3lame/VbrTag.c, libmp3lame/lame.c:
	  default lowpass stored in info tag

2001-09-19 14:32  markt

	* testcase.mp3:
	  updated testcase.mp3

2001-09-18 13:37  aleidinger

	* libmp3lame/util.c:
	  Fix memory hole (allocation from psymodel_init()).
	  Found by: Michael Fink <finkm@asamnet.de>

2001-09-17 23:54  markt

	* libmp3lame/VbrTag.c, libmp3lame/util.c:
	  moved a 1M array from stack to heap in VbrTag.c
	  
	  Some OS, when launching LAME on a seperate thread,
	  allocate a tine (128K?) stack. moving this to the heap
	  is an ugly solution (requires a malloc() and free())
	  but lets see if it fixes the reported segfaults.

2001-09-17 21:17  markt

	* frontend/main.c, include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/lame.c, libmp3lame/util.c:
	  more code related to putting encoder padding into VBR tag

2001-09-17 20:05  markt

	* include/lame.h, libmp3lame/VbrTag.c, libmp3lame/encoder.h,
	  libmp3lame/lame.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/set_get.c:
	  added new call to API,
	  lame_get_encoder_padding()
	  which returns the amount of padding appended to the input
	  during encoding.

2001-09-17 19:15  markt

	* libmp3lame/util.c:
	  moved MSVC cpu affinity code to the top of disable_FPE()

2001-09-15 09:57  aleidinger

	* configure:
	  regen

2001-09-15 09:49  aleidinger

	* configure.in:
	  Bugfix for libffm (Alpha architecture).
	  Submitted by: Simon Burge <simonb@wasabisystems.com>

2001-09-14 17:37  markt

	* HACKING, README:
	  added a note about LAME being threadsave to the HACKING file.

2001-09-14 17:18  markt

	* libmp3lame/util.c:
	  put Todd Richmond's EAC/lame SMP fix in util.c,
	  with all the messy code for setting floating point exceptions.
	  I put this in the
	  #if defined(_MSC_VER)
	  block, so it should only affect people using MSVC.

2001-09-11 07:47  potsticker

	* libmp3lame/VbrTag.c:
	  Changed id string from 'Xing' to 'Info' for CBR files only

2001-09-08 18:40  potsticker

	* frontend/main.c, libmp3lame/VbrTag.c, libmp3lame/VbrTag.h:
	  Don't write Lame Tag status message if --quiet

2001-09-08 18:27  potsticker

	* libmp3lame/VbrTag.c, libmp3lame/VbrTag.h:
	  Don't write Lame Tag status message if --quiet

2001-09-07 23:45  potsticker

	* frontend/parse.c, libmp3lame/VbrTag.c, libmp3lame/lame.c:
	  Write Lame Tag for all files (unless -t switch used)

2001-09-04 00:06  potsticker

	* libmp3lame/VbrTag.c:
	  Implemented most of Lame Tag rev 0 (but no 'replay gain' yet)

2001-09-03 23:45  potsticker

	* libmp3lame/VbrTag.c, libmp3lame/VbrTag.h:
	  Implemented most of Lame Tag rev 0 (but no 'replay gain' yet)

2001-09-03 23:43  potsticker

	* libmp3lame/id3tag.h:
	  Moved some macro definitions into id3tag.h (referenced in
	  VbrTag.c)

2001-09-03 20:56  robert

	* libmp3lame/id3tag.c:
	  cr/lf patch

2001-09-02 20:08  potsticker

	* libmp3lame/id3tag.c, libmp3lame/id3tag.h:
	  Moved some macro definitions into id3tag.h (referenced in
	  VbrTag.c)

2001-08-31 18:41  markt

	* libmp3lame/util.c:
	  lowered tolerence in an assert() in the lowpass filtering
	  routine.

2001-08-31 08:17  bouvigne

	* libmp3lame/VbrTag.c, libmp3lame/util.h:
	  I hope it's fixed now

2001-08-29 20:08  bouvigne

	* Dll/BladeMP3EncDLL.c, include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/VbrTag.h, libmp3lame/version.c:
	  first implementation of lame Xvbr tag from J Dee

2001-08-28 18:51  markt

	* API, HACKING, frontend/parse.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/psymodel.c, libmp3lame/set_get.c:
	  msfix changes. forgot to commit them yesterday

2001-08-28 17:06  aleidinger

	* Makefile.unix:
	  OS/2 mods from slaughter@malaconet.org (in principle, I had to
	  apply them by hand and modified it a little bit)

2001-08-28 15:41  aleidinger

	* configure:
	  oops... wrong gtk-config

2001-08-27 21:06  aleidinger

	* configure:
	  regen (+update to automake 1.5)

2001-08-27 21:06  aleidinger

	* Dll/Makefile.in, Makefile.in, aclocal.m4, debian/Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.am, mpglib/Makefile.in:
	  update to automake 1.5

2001-08-27 21:04  aleidinger

	* Makefile.unix:
	  USE_FFT3DN died some time ago

2001-08-27 21:03  aleidinger

	* Makefile.am:
	  fixes for the update and diff target if you build in a seperate
	  directory

2001-08-27 21:01  aleidinger

	* frontend/main.c:
	  Bump MAX_NOGAP from 20 to 40, I wanted to make 34 gapless mp3s.

2001-08-27 20:59  aleidinger

	* Makefile.am.global, config.guess, config.sub, depcomp,
	  misc/depcomp, missing, mkinstalldirs, mpglib/depcomp:
	  update to automake 1.5

2001-08-27 20:56  aleidinger

	* DEFINES, Dll/MP3EncDll.dsp, Makefile.B32, Makefile.MSVC,
	  libmp3lame/libmp3lame.dsp, libmp3lame/vorbis_interface.c:
	  remove USE_FFT3DN, this define died some time ago

2001-08-27 20:55  aleidinger

	* configure.in:
	  remove --malign-double completely

2001-08-27 20:52  aleidinger

	* libmp3lame/fft.c:
	  USE_FFT3DN -> HAVE_NASM

2001-08-27 20:50  aleidinger

	* lame.spec.in:
	  - Build and include docs and libs correctly
	  - Build extra programs
	  
	  Submitted by: Markus Linnala <maage@cs.tut.fi>
	  
	  Note:
	  I didn't have a rpm based system, if this breaks feel free
	  to revert my commit, I'm at a "commit what isn't obviously
	  broken and seems to be a good idea" run at the moment.

2001-08-27 20:49  aleidinger

	* USAGE, doc/html/switchs.html, frontend/parse.c:
	  - OS/2 priority control
	  Submitted by: "DoC" <slaughter@malaconet.org> (with minor changes
	  by me)

2001-08-27 16:41  markt

	* doc/man/lame.1, misc/Lame.vbs:
	  typo in man page fixed, new Lame.vbs from
	  "Ralf Kempkens" <Ralf.Kempkens@epost.de>

2001-08-26 19:01  markt

	* configure.in:
	  removed -malign-double, based on Klemm's email.

2001-08-26 18:59  markt

	* configure.in:
	  some missing quotes for AmigaOS section of configure.in,
	  submitted by Fredrik Mellstrom <traal@altavista.net>
	  
	  (I cant beieve there are still people out there using
	  an amiga :-)

2001-08-26 18:38  markt

	* mpglib/interface.c, mpglib/mpglib.h:
	  replaced "look_for_xing" variable with more descriptive name
	  "sync_bitstream".

2001-08-21 19:48  bouvigne

	* frontend/parse.c:
	  switching --vbr-new to mtrh

2001-08-21 08:57  viral

	* debian/changelog, debian/control, debian/lame-dev.docs,
	  debian/lame.docs, debian/rules:
	  Debian packaging modifications.

2001-08-20 22:26  afaber

	* Dll/BladeMP3EncDLL.c, Dll/MP3EncDll.dsp:
	  Updated new R3Mix settings, use lame get/set function and version
	  info
	  so I got rid of the version.h and lame_global_flag.h include
	  files

2001-08-20 20:54  robert

	* doc/html/contributors.html:
	  adding Darin Morrison (author of dm presets)

2001-08-16 08:21  robert

	* misc/abx.c:
	  update of abx.c by Frank Klemm

2001-08-15 21:58  robert

	* frontend/parse.c, libmp3lame/lame.c:
	  making -Z a truely toggling switch
	  reordering the longhelp options list
	  dropping --extrahelp

2001-08-15 00:37  robert

	* libmp3lame/lame.c:
	  dropped a nagging message for --vbr-mtrh (nspsytune tuned for
	  ...)

2001-08-15 00:12  robert

	* frontend/parse.c, libmp3lame/util.c:
	  update for --r3mix by Roel

2001-08-14 21:46  robert

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.h, libmp3lame/version.h:
	  some good presets tuned by Dibrom after die hard listening tests
	  --dm-preset {standard,xtreme,insane,metal}
	  and a switch to modify the --ns-treble setting for sfb21 by
	  --ns-sfb21 x

2001-08-08 10:52  aleidinger

	* configure:
	  regen

2001-08-08 10:51  aleidinger

	* configure.in:
	  - remove optimizations which get already switched on by -O2 or
	  better
	  - new experimental optimizations
	  - bump version to 3.90

2001-08-08 10:50  aleidinger

	* configMS.h:
	  bump version to 3.90

2001-08-07 20:41  afaber

	* libmp3lame/libmp3lame.dsp:
	  Added NASM release build option

2001-08-02 18:42  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h, Dll/Example.cpp,
	  Dll/MP3EncDll.dsp:
	  Added NASM build option, added Very High Quality option

2001-08-01 21:45  robert

	* frontend/parse.c, libmp3lame/lame.c,
	  libmp3lame/lame_global_flags.h, libmp3lame/quantize_pvt.c,
	  libmp3lame/set_get.c, libmp3lame/version.h:
	  BUG fix in "on_pe", was allocating often more than the absolute
	  limit of 4095 bits per granule
	  attention: testcases are most likely broken now!

2001-07-26 23:05  robert

	* libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/vbrquantize.c, libmp3lame/version.h:
	  VBR mtrh related, bring --remix near to --r3mix bitrate wise

2001-07-25 20:19  robert

	* frontend/main.c:
	  Michel SUCH, wildcard expansion for nogap using the OS/2 shell

2001-07-24 12:13  aleidinger

	* configure:
	  regen

2001-07-24 12:12  aleidinger

	* configure.in:
	  gcc 3.0 fix from Daniel T. Chen

2001-07-23 08:13  aleidinger

	* Dll/Makefile.in, Makefile.am.global, Makefile.in, aclocal.m4,
	  config.guess, config.sub, configure, debian/Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  update to automake 1.4-p5

2001-07-22 19:24  robert

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c:
	  small fixes

2001-07-22 06:37  jd-

	* libmp3lame/encoder.c:
	  Pushed the maximum auto-level adjustment of the ATH from 20 dB to
	  32 dB

2001-07-20 23:56  robert

	* libmp3lame/quantize.c:
	  small speed up for --vbr-old

2001-07-20 01:01  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/vbrquantize.c,
	  libmp3lame/version.h:
	  1) defaulted the old -Y code for --vbr-old
	  2) new behaviour of -Y: turns off sfb21 noise shaping (VBR)
	  3) some more tunings for --vbr-old

2001-07-19 11:36  aleidinger

	* configure:
	  regen

2001-07-19 11:35  aleidinger

	* configure.in:
	  - removed useless (because already handled by Cygwin) MS-DOS case
	  - bug fixes for alpha*-dec-osf*, *sgi-irix* and sparc*-sunos4*
	  - print a "please contact us" message for sparc*-sunos4* and
	  *beos
	  (theses parts of configure.in are perhaps not neccessary)

2001-07-18 20:16  robert

	* frontend/parse.c, libmp3lame/vbrquantize.c:
	  vbr-mtrh related: encode sfb21, may bloat on heavy metal music

2001-07-18 20:14  robert

	* misc/abx.c:
	  update by Frank

2001-07-16 19:35  robert

	* Dll/BladeMP3EncDLL.c, frontend/parse.c, include/lame.h,
	  libmp3lame/lame.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/psymodel.c, libmp3lame/set_get.c, libmp3lame/util.h,
	  libmp3lame/version.h:
	  adding a new switch --allshort
	  this one forces LAME to use short blocks on every granule
	  useful for testing the short block masking abilities

2001-07-13 22:37  robert

	* libmp3lame/psymodel.c, libmp3lame/quantize.c, libmp3lame/util.h,
	  libmp3lame/version.h:
	  first try to fix the short block threshold calculation BUG
	  this patch tries to resolve the shortcommings for the short block
	  threshold calculation using VBR. Applying this patch for CBR/ABR
	  would break the testcases. It should not affect the vanilla
	  --nspsy...
	  too, maybe later.

2001-07-07 19:54  bouvigne

	* libmp3lame/reservoir.c:
	  maxmp3buf was beeing ignored

2001-07-05 21:25  robert

	* libmp3lame/quantize.c:
	  fixing the assertion failures mentioned by Albert Faber

2001-07-05 15:56  markt

	* configure:
	  regenerated configure

2001-07-05 15:55  markt

	* configure.in:
	  updated gcc 2.96 check to use -O instead of -O1

2001-07-05 15:55  markt

	* configure:
	  *** empty log message ***

2001-07-05 15:54  markt

	* configure.in:
	  gcc 2.96 check

2001-07-05 15:53  markt

	* libmp3lame/version.h:
	  updated repository to 3.90alpha

2001-07-05 15:35  markt

	* doc/html/history.html:
	  updateded history file from web page cvs repository

2001-07-05 15:32  markt

	* ChangeLog:
	  update

2001-07-05 15:27  markt

	* libmp3lame/version.h:
	  updated to 3.89beta

2001-07-02 21:45  robert

	* libmp3lame/quantize.c:
	  bug fix for --nspsytune
	  minimum bits allocated was too low, resulting in assertion
	  failures

2001-07-02 21:08  robert

	* libmp3lame/lame.c:
	  changing --vbr-new noise measuring back to its usual one

2001-07-01 20:36  markt

	* libmp3lame/psymodel.c:
	  unbreak CBR testcases.

2001-07-01 20:04  markt

	* frontend/get_audio.c, frontend/main.h, frontend/parse.c:
	  patch from Nick Duffek <nsd@bosbc.com>
	  for signed/unsigned and bitwidth options.

2001-07-01 16:48  robert

	* libmp3lame/lame.c:
	  defaulting --vbr-mtrh to -q2 = -h

2001-07-01 13:11  robert

	* libmp3lame/lame.c:
	  allow --vbr-{old,new} to select between different ATH autoadjust
	  types too

2001-07-01 11:50  robert

	* libmp3lame/lame.c, libmp3lame/set_get.c:
	  nspsytune related:
	  now it is possible to turn the scalefac_feature off again giving
	  --nspsytune -Z
	  the change was necessary, because of the hard coded -Z

2001-07-01 05:24  jd-

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/util.h:
	  Restructured ATH auto-leveling code to fix bug for
	  "--adapt-thres-type 1";
	  relaxed tuning for "--adapt-thres-type 2" for compromise with
	  "--athtype 3"
	  (the previous bit rate is still available with
	  "--adapt-thres-level -1.25")

2001-06-30 16:45  robert

	* Makefile.MSVC, frontend/parse.c:
	  fix for compiling mp3x on windows using Makefile.MSVC

2001-06-30 00:26  robert

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c:
	  2 new modes for --vbr-mtrh available by different -q n
	  selections:
	  -q0 truely max noise
	  -q1 almost max noise

2001-06-29 20:58  robert

	* libmp3lame/set_get.c:
	  revive --nssafejoint --ns-{alto,treble,bass}
	  lame_set_exp_nspsytune is *not an on/off switch*

2001-06-29 15:23  markt

	* configure, configure.in, libmp3lame/lame.c:
	  patch for configure.in, replaced a bash specific loop,
	  from Oswald Buddenhagen <ob6@inf.tu-dresden.de>

2001-06-27 15:47  aleidinger

	* configure:
	  regen

2001-06-27 15:46  aleidinger

	* configure.in:
	  some hardware specific optimization options for gcc

2001-06-27 14:53  aleidinger

	* libmp3lame/i386/Makefile.in:
	  regen

2001-06-27 14:53  aleidinger

	* libmp3lame/i386/Makefile.am:
	  this is nasty, I want native nasm support in libtool, at least it
	  works for me (sort of)

2001-06-27 14:11  aleidinger

	* libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in:
	  regen

2001-06-27 14:11  aleidinger

	* libmp3lame/Makefile.am, libmp3lame/i386/Makefile.am:
	  go back to let libtool compile liblameasmroutines, the static
	  case was broken, now the libtool v1.4 case may be broken

2001-06-26 11:38  aleidinger

	* configure:
	  regen

2001-06-26 11:36  aleidinger

	* configure.in:
	  make the use of nasm a configure option

2001-06-25 02:14  cisc

	* frontend/main.c:
	  Include portableio.h for the WriteBytes() and WriteBytesSwapped()
	  functions.

2001-06-24 21:01  robert

	* lame.bat:
	  some comments updated

2001-06-24 16:47  robert

	* doc/html/contributors.html, libmp3lame/version.h:
	  adding two more contributors

2001-06-24 15:58  robert

	* Makefile.MSVC:
	  FLOAT8=float for IC

2001-06-24 15:57  robert

	* Makefile.B32:
	  let it work again

2001-06-24 02:04  robert

	* libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c:
	  BUG fix in inc_subblock gain, was amplifying at the wrong place
	  this fix may break the testcases
	  some unused parameters removed

2001-06-23 18:26  robert

	* Dll/MP3EncDll.dsp:
	  patch for DLL Project file by Dmitry

2001-06-23 17:56  robert

	* Makefile.MSVC, frontend/lametime.c, frontend/parse.c,
	  mpglib/common.c:
	  some fixes needed by the Borland Compiler BCC, first attempt to
	  support BCC with Makefile.MSVC, DLL not supported yet, MP1 MP2
	  reencoding does not work with BCCcd lame (why?)

2001-06-23 12:19  robert

	* libmp3lame/quantize_pvt.c:
	  fix letting Borland C++ compiler getting thru' again,
	  Makefile.B32 is *not uptodate*

2001-06-23 12:15  robert

	* frontend/get_audio.c:
	  BUG fix for mp1 and mp2 input, fixes problem introduced in
	  Revision 1.63

2001-06-22 08:55  aleidinger

	* ChangeLog:
	  update changelog

2001-06-22 08:49  aleidinger

	* Makefile.am:
	  regen

2001-06-22 08:48  aleidinger

	* Makefile.in:
	  we have a test subdir in cvs, force the test target to be out of
	  date at every invocation of make

2001-06-21 20:40  bouvigne

	* libmp3lame/reservoir.c:
	  change of the max frame size

2001-06-19 04:46  markt

	* misc/Lame.vbs, misc/lameGUI.html:
	  updates from "Ralf Kempkens" <Ralf.Kempkens@epost.de>

2001-06-18 18:51  markt

	* TODO, frontend/get_audio.c, frontend/get_audio.h,
	  frontend/main.c, frontend/main.h, frontend/parse.c,
	  include/lame.h, libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/set_get.c, mpglib/interface.c, mpglib/mpglib.h:
	  added two new functions to return number of PCM samples buffererd
	  but not encoded, and the number of bytes of mp3 data buffered but
	  not output. also added some comments to mpglib/interface.c to
	  make it easier to tell it to look for Xing headers in the middle
	  of a bitstream.

2001-06-16 00:40  robert

	* Makefile.MSVC, Makefile.unix:
	  CPU feature detection for fft routines

2001-06-16 00:39  robert

	* libmp3lame/fft.c, libmp3lame/util.h:
	  CPU feature autodetection for fft routines

2001-06-15 22:23  markt

	* TODO, frontend/main.c, include/lame.h, libmp3lame/lame.c:
	  Robert's fix to init histgram data with --nogap.
	  
	  lame_reinit_bitstream() renamed lame_init_bitstream(), and
	  it is called by lame_init_params(). It can also be called
	  after a call to lame_encode_flush_nogap().

2001-06-15 21:25  markt

	* TODO, frontend/main.c, include/lame.h, libmp3lame/lame.c:
	  some fixes for VBR tags with --nogap option.

2001-06-14 23:09  robert

	* frontend/parse.c:
	  fixed stdin

2001-06-14 00:58  robert

	* libmp3lame/version.h:
	  bumped version to alpha 7

2001-06-14 00:43  robert

	* libmp3lame/lame.c, libmp3lame/psymodel.c:
	  added JD's loudness approximation to Naoki's duplicated psy
	  routine
	  as I defaulted JD's adapt stuff I was under the impression that
	  he
	  added his code to both ones, so...

2001-06-14 00:19  markt

	* frontend/main.c:
	  fix for nogap truncation call. When processing the last
	  of the .wav files, we need to call lame_encode_flush()
	  instead of lame_encode_flush_nogap().

2001-06-13 22:46  robert

	* frontend/parse.c:
	  LFN support for Win32: realized that it would be a good idea to
	  release the search file handle after searching the filename

2001-06-13 08:44  aleidinger

	* Makefile.DJGPP:
	  nasm fix

2001-06-12 21:58  robert

	* frontend/parse.c:
	  long file name support for Win32 systems
	  works under Win95 using Intel compiler 4.5 / MSVC 4.2
	  not tested under NT, and not much tested anyway ;-)

2001-06-12 21:03  robert

	* Makefile.unix:
	  seems to be forgotten by the last committ

2001-06-12 20:32  markt

	* libmp3lame/bitstream.c:
	  old crc code removed from bitstream.c.
	  I had added this in when looking for the xing/crc bug

2001-06-12 12:47  aleidinger

	* frontend/lametime.c:
	  fix cygwin build, sort some includes

2001-06-12 12:45  aleidinger

	* configure, frontend/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in:
	  regen

2001-06-12 12:45  aleidinger

	* libmp3lame/Makefile.am, libmp3lame/i386/Makefile.am:
	  do not use libtool for the asm lib, libtool did not know about
	  nasm

2001-06-12 12:42  aleidinger

	* configure.in:
	  - enable shared lib
	  - disable asm routines
	  - fix build of asm routines in the win32 case

2001-06-12 12:41  aleidinger

	* frontend/Makefile.am:
	  statically link libmp3lame to the frondends

2001-06-12 05:27  markt

	* libmp3lame/lame.c:
	  removed CRC error message

2001-06-12 05:27  markt

	* libmp3lame/VbrTag.c:
	  fix for Xing/CRC problem.
	  
	  Xing toc data must start directly after sideinfo data, assuming
	  no CRC. If CRC is set, then the Xing toc data must *still* start
	  in the same location (since original Xing demo code does not
	  check for CRC and modify the toc offset).

2001-06-12 04:51  markt

	* libmp3lame/bitstream.c, libmp3lame/lame.c:
	  put old CRC routine back in (but disabled)

2001-06-12 03:35  markt

	* Dll/BladeMP3EncDLL.c:
	  added some comments about the DLL's non-standard use of
	  libmp3lame.a

2001-06-12 03:31  markt

	* libmp3lame/Makefile.am:
	  Robert created lame_global_flags.h,
	  so I'm adding it to Makefile.am, in the "noinst" section.

2001-06-12 00:48  robert

	* Dll/BladeMP3EncDLL.c, libmp3lame/lame_global_flags.h,
	  libmp3lame/quantize.c, libmp3lame/util.h:
	  - fixed Makefile.unix to let mp3rtp compile again (does it still
	  work?)
	  - had to move lame_global_flags structure into a file of its own,
	  because the Dll compiled but did not link anymore.
	  Now it works again, tested on Win95 using Makefile.MSVC

2001-06-11 23:42  markt

	* Dll/BladeMP3EncDLL.c, include/lame.h, libmp3lame/set_get.c,
	  libmp3lame/util.h:
	  committed Robert's patch. lame_global_flags is gone :-)

2001-06-11 22:07  robert

	* frontend/main.c:
	  some more new API things

2001-06-11 18:49  markt

	* frontend/gtkanal.c, frontend/parse.c:
	  converted rest of frontend code over to "new api"

2001-06-11 16:35  markt

	* frontend/get_audio.c, frontend/get_audio.h, frontend/gtkanal.c,
	  libmp3lame/mpglib_interface.c, mpglib/interface.c:
	  Tracked down a bug in lame_decoder. It was not correctly
	  flushing the internal mpglib buffers when it hit EOF.
	  
	  Problem was most severe with low bitrate example. It the frame
	  size
	  was 32 bytes, only the first 1/3 of the file would be decoded
	  and the reamining 2/3 of the entire file would be buffered
	  inside mpglib.

2001-06-10 20:35  robert

	* Dll/BladeMP3EncDLL.c, include/lame.h:
	  making the old VBR default in preparation for the next release

2001-06-10 19:19  markt

	* API, USAGE, misc/Lame.vbs, misc/lameGUI.html:
	  added new version of Lame.vbs from "Ralf Kempkens"
	  <Ralf.Kempkens@epost.de>
	  including a GUI/HTML interface.

2001-06-08 21:40  markt

	* frontend/get_audio.c, frontend/gtkanal.c, frontend/main.c,
	  frontend/main.h:
	  updated some comments related to restructering main.c
	  for the --nogap option

2001-06-08 21:29  markt

	* API, TODO, frontend/get_audio.c, frontend/gtkanal.c,
	  frontend/main.c, frontend/mp3rtp.c, frontend/mp3x.c,
	  frontend/parse.c, frontend/parse.h, frontend/timestatus.c,
	  include/lame.h, libmp3lame/lame.c, libmp3lame/set_get.c:
	  preliminary --nogap option added.
	  
	  --nogap file1 file2 file3
	  
	  and it produces file1.mp3 file2.mp3 file3.mp3
	  
	  problems: no id3tags, no vbr tags.

2001-06-08 18:59  markt

	* test/lametest.py:
	  fix so lametest.py doesn't crash if lame does not create an
	  output file.

2001-06-08 00:00  markt

	* frontend/main.c, libmp3lame/bitstream.c:
	  fixed bug with lame_encode_flush_nogap routine

2001-06-07 22:12  robert

	* DEFINES, TODO:
	  *** empty log message ***

2001-06-07 22:06  robert

	* frontend/parse.c, libmp3lame/encoder.c:
	  --ath-adjust 1 (Gaby's original code) was disabled, any reasons
	  for that JD?

2001-06-07 19:14  markt

	* HACKING, frontend/get_audio.c, frontend/gtkanal.c,
	  frontend/parse.c, frontend/timestatus.c, libmp3lame/lame.c:
	  started updating frontend code to use the new API.

2001-06-07 14:02  aleidinger

	* libmp3lame/fft.c:
	  unbreak (at least it compiles), more work needed

2001-06-07 10:11  jd-

	* libmp3lame/encoder.c:
	  Allow ATH auto-adjust code to begin at the minimum value on the
	  leading frame

2001-06-07 08:40  jd-

	* libmp3lame/encoder.c:
	  Restored some ATH auto-adjust indentation and comments (altered
	  since 3.88
	  in a code clean up)

2001-06-07 06:06  pgubanov

	* libmp3lame/fft.c:
	  Choose asm-optimized FHT version if ARCH_X86 defined in fft.c

2001-06-05 22:13  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  Added R3Mix quality setting

2001-06-05 19:19  aleidinger

	* doc/html/history.html:
	  sync my entry with the one at the webpages

2001-06-05 00:39  markt

	* INSTALL, TODO:
	  added note about mpglib i-stereo bug in TODO list.
	  fixed typo in INSTALL file

2001-06-04 23:57  markt

	* TODO, libmp3lame/reservoir.c:
	  documented some variables in reservoir.c.
	  
	  Current status is:
	  
	  maxframesize = no restrictions
	  
	  if strict_ISO is defined, we use Gaby's less restrictive
	  interpretation of the standard.
	  
	  see TODO version 1.30 for a user who has a hardware player which
	  cannot handle 320kbps frames if the bit reservoir is used.
	  
	  I guess the final verdict is tough luck :-)

2001-06-04 23:29  markt

	* API, HACKING, frontend/parse.c, libmp3lame/lame.c:
	  print error message and fail if user selects error_protection
	  (since this is currently broken)
	  
	  added some comments to HACKING on how to add a new option.
	  
	  updated API to reflect that all lame_set/get functions are now
	  working.

2001-06-04 23:10  markt

	* include/lame.h, libmp3lame/lame.c:
	  updated comments about lame_encode_flush_nogap()

2001-06-04 23:08  markt

	* include/lame.h, libmp3lame/bitstream.c, libmp3lame/lame.c:
	  added lame_encode_flush_nogap().
	  
	  This routine will flush all internal mp3 buffers, and pad the
	  last frame with ancillary data so it is a complete mp3 frame.
	  
	  It will then set ResvSize to 0, so we can continue encoding.
	  
	  after a call to lame_encode_flush_nogap(), if we close
	  the mp3 file and open a new one for the future data,
	  then the two mp3 files should play back 'gapless'
	  if they are first concatenated together.

2001-06-04 19:59  robert

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/version.h:
	  tweaks to VBR

2001-06-04 10:02  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  Extended interface with VBR method selection

2001-06-04 10:01  afaber

	* mpglib/layer3.c:
	  Removed exit call(s) in error conditions, since these errors
	  seems to be good recoverable

2001-06-04 01:54  robert

	* libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/util.h:
	  small tweaks to the VBR codes

2001-06-02 20:22  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/util.h,
	  libmp3lame/version.h:
	  started different bit-pressure strategies (VBR)

2001-06-02 15:26  aleidinger

	* configure:
	  regen

2001-06-02 15:25  aleidinger

	* configure.in:
	  force cygwin users to use "CC=gcc ./configure"

2001-06-01 19:40  glessard

	* frontend/get_audio.c:
	  Added support for non-compressed AIFF-C files in
	  parse_aiff_header().

2001-05-30 01:38  markt

	* configure, configure.in:
	  removed -pedantic gcc option.
	  From the gcc man page: "There is no reason to use this
	  option; it exists only to satisfy pedants".
	  
	  And there is a reason not to use it: It causes people to
	  send me email asking if all the warning messages about c++
	  comments are ok.

2001-05-29 22:14  robert

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/set_get.c:
	  quick fix for a typo that turned off auto adjust code by
	  accident, sorry

2001-05-27 18:58  robert

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/version.h:
	  made JD's ATH adapt stuff default for --vbr-mtrh
	  some typings
	  changed bitpressure strategy for VBR a bit

2001-05-26 21:31  jd-

	* libmp3lame/util.c:
	  use more digestible code in ATHformula_jd for a pre-release
	  compiler bundled
	  with a popular Linux distribution

2001-05-26 16:17  robert

	* libmp3lame/lame.c:
	  let the user select -qx for --vbr-mtrh, just to avoid useless
	  e-mails

2001-05-24 22:25  robert

	* libmp3lame/lame.c, libmp3lame/quantize_pvt.c:
	  just 5 minutes to add some more on the --verbose printing side

2001-05-21 18:55  bouvigne

	* doc/html/history.html:
	  history

2001-05-21 18:19  bouvigne

	* libmp3lame/lame.c:
	  update of the default lowpass value. Will probably break some
	  testcases

2001-05-20 20:45  robert

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/util.h:
	  some more --verbose output

2001-05-19 16:16  robert

	* libmp3lame/psymodel.c:
	  typo fixed

2001-05-19 16:13  robert

	* frontend/get_audio.c, frontend/main.c, frontend/parse.c,
	  frontend/timestatus.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/psymodel.c, libmp3lame/util.h:
	  starting a very verbose print function about all internal
	  settings, aiming more transparency for the experienced users

2001-05-15 01:24  robert

	* libmp3lame/psymodel.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/vbrquantize.c:
	  some tunings
	  still problems with fatboy.wav and ns.wav

2001-05-13 20:59  bouvigne

	* libmp3lame/util.c:
	  comments in the ath functions

2001-05-13 11:33  robert

	* libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  fixing tonality problem for --vbr-mtrh

2001-05-11 23:33  robert

	* libmp3lame/quantize_pvt.c:
	  trying to point out the real difference in l3xmin calculation
	  with nspsytune and without it

2001-05-11 22:14  robert

	* libmp3lame/lame.c:
	  nspsytune & newer vbr

2001-05-11 01:30  robert

	* frontend/get_audio.c, frontend/main.h, include/lame.h,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/version.h, test/VBR.op:
	  compile fixes for Intel Compiler, dropped sfb21 early stopping
	  for vbr-old (found differences)

2001-05-10 19:01  aleidinger

	* include/lame.h, libmp3lame/set_get.c:
	  - Make the last few set/get functions functional.
	  (TODO: use them in the frontend).
	  - Converted padding_type to an enum.
	  - Correct some typos.
	  - Adjust some lines to conform to the style of the majority of
	  the file.
	  
	  TODO:
	  - Search for XXX and do appropriate actions.
	  - Search for typos (because of alot of cut&paste programming).

2001-05-08 15:00  bouvigne

	* frontend/parse.c, libmp3lame/util.c:
	  update of --r3mix

2001-05-06 15:11  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/vbrquantize.c:
	  some desperate values for the fast mtrh VBR mode

2001-05-04 01:07  jd-

	* libmp3lame/tools.h, libmp3lame/util.c:
	  Added "--athtype 5" to better represent higher frequencies (14 -
	  19 kHz).
	  Included supporting functions for quadratic interpolation.

2001-05-03 20:13  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c:
	  some VBR-codeclean-up

2001-05-02 21:02  robert

	* misc/lame4dos.bat:
	  patch by Alexander Stumpf:
	  - now works both with drag & drop (SFN) and from the
	  4DOS-Commandline (LFN)
	  - Using wildcards when prozessing from mp3 to mp3 caused a loop -
	  fixed
	  - removed the mp3-extention test since LAME recognizes the input
	  file fomat
	  correctly for several versions now (i.e. no --mp3input neccessary
	  anymore)

2001-05-01 22:55  robert

	* libmp3lame/version.h:
	  bumped version number

2001-05-01 22:18  robert

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c:
	  new -q2 mode for --vbr-mtrh

2001-05-01 20:37  bouvigne

	* libmp3lame/reservoir.c:
	  change of max frame size

2001-04-28 15:37  robert

	* libmp3lame/encoder.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.c, libmp3lame/vbrquantize.c:
	  --vbr-mtrh uses now -Y by default, some tweaks

2001-04-27 23:19  robert

	* libmp3lame/psymodel.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.c, libmp3lame/util.h:
	  --experimentalY selects a different ATH adjustment
	  - keeps original noise floor
	  - affects high freqs more than low ones

2001-04-26 23:06  cisc

	* config.guess, config.sub:
	  Added MorphOS recognition to configure script.

2001-04-26 04:38  markt

	* README.B32, TODO:
	  removed mmx notes from README.B32

2001-04-25 16:21  aleidinger

	* configure, libmp3lame/i386/Makefile.in:
	  regen

2001-04-25 16:20  aleidinger

	* configure.in:
	  bugfix for gcc version detection

2001-04-25 16:19  aleidinger

	* libmp3lame/i386/Makefile.am:
	  bugfix for non x86 architectures with installed nasm, tested by
	  Karl Pauls <kpauls@eecs.wsu.edu> on Darwin

2001-04-25 06:59  markt

	* TODO, frontend/get_audio.c, frontend/main.h, frontend/parse.c,
	  include/lame.h, libmp3lame/set_get.c:
	  removed 'disable_wave_header' variable from encoding library.
	  It is in the frontend code only.

2001-04-24 01:38  markt

	* TODO, frontend/get_audio.c:
	  added something to the TODO list.
	  
	  changed:
	  
	  "buffer[1]+i" code to "&buffer[1][i]"
	  
	  in get_audio.c. buffer[1]+i is clever, but it confused the
	  maintainer of EAC enough to cause him to submit it as a bug
	  report.

2001-04-23 19:47  bouvigne

	* libmp3lame/lame.c, libmp3lame/util.c:
	  *** empty log message ***

2001-04-22 22:57  bouvigne

	* libmp3lame/util.c:
	  update of the ath functions

2001-04-22 01:34  jd-

	* frontend/get_audio.c:
	  Fixed trailing data bug introduced with the 24 bit input code

2001-04-21 22:23  jd-

	* frontend/get_audio.c, frontend/get_audio.h, frontend/gtkanal.c,
	  frontend/main.c, frontend/mp3rtp.c, include/lame.h,
	  libmp3lame/lame.c:
	  Add 24-bit PCM waveform input capability (with or without
	  libsndfile),
	  and glue code to preserve the functionality of routines that only
	  handle
	  up to 16-bit data. The changes support both little and big endian
	  architectures.

2001-04-17 20:05  afaber

	* Dll/BladeMP3EncDLL.c, frontend/lame.dsp,
	  libmp3lame/libmp3lame.dsp, mpglib/mpglib.dsp:
	  Added MMX compile option

2001-04-17 19:22  bouvigne

	* libmp3lame/util.c:
	  auto adjusting ath shape for vbr on ath type 4

2001-04-16 21:26  bouvigne

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/psymodel.c:
	  wrapped a few things inside of ifdef HAVE_GTK

2001-04-16 11:50  takehiro

	* libmp3lame/psymodel.c, libmp3lame/util.h:
	  spread function is now packed into linear array.
	  it reduces memory assumption and improves cache efficienty.

2001-04-16 08:27  takehiro

	* include/lame.h, libmp3lame/VbrTag.c:
	  now try to calculate the CRC of VBR header.

2001-04-16 07:58  takehiro

	* libmp3lame/bitstream.c, libmp3lame/bitstream.h:
	  cleaned and improved CRC calculation method

2001-04-13 15:13  aleidinger

	* libmp3lame/set_get.c:
	  add some comments about converting int->enum

2001-04-13 15:12  aleidinger

	* frontend/parse.c:
	  use some of Marks set-functions, also without error checking
	  (yet)

2001-04-12 17:40  bouvigne

	* libmp3lame/encoder.c, libmp3lame/lame.c:
	  *** empty log message ***

2001-04-08 07:52  takehiro

	* libmp3lame/machine.h, libmp3lame/quantize_pvt.c:
	  portability enchancement(from sourceforge's LAME open discussion
	  forum).
	  
	  quantize_pvt.c: changed 1e99 into FLOAT8_MAX
	  machine.h: defined FLOAT8_MAX

2001-04-04 21:58  bouvigne

	* libmp3lame/lame.c:
	  update of the lowpass values for vbr rh

2001-04-04 18:11  bouvigne

	* libmp3lame/psymodel.c:
	  update of -d for joint stereo

2001-04-01 03:22  jd-

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/set_get.c, libmp3lame/util.h:
	  added an --adapt-thres-level option to allow adding an offset (in
	  dB)
	  to the point below which adaptive ATH level adjustment occurs

2001-04-01 02:11  jd-

	* include/lame.h:
	  added set/get functions to include/lame.h for option
	  --adapt-thres-type

2001-03-31 23:44  jd-

	* frontend/parse.c, libmp3lame/set_get.c:
	  added set/get functions to libmp3lame/set_get.c for option
	  --adapt-thres-type, and used the set function in frontend

2001-03-30 18:43  markt

	* frontend/get_audio.c, test/lametest.py, testcase.mp3:
	  updated testcase.mp3
	  removed some unused code from get_audio.c

2001-03-30 17:16  markt

	* test/lametest.py:
	  delete output mp3 file before running lame.
	  (before, if lame crashes and does not write any output, the
	  script would use the old output mp3 file, and if it hadn't
	  changed,
	  all tests would pass.)

2001-03-29 18:31  cisc

	* libmp3lame/version.c, libmp3lame/version.h:
	  Moved inclusion of lame.h to version.h for lame_version_t in
	  prototype.

2001-03-27 12:26  jd-

	* libmp3lame/psymodel.c:
	  added --adapt-thres-type 2 (auto-level adjustment of the ATH,
	  with a frequency-based loudness approximation).

2001-03-27 12:04  jd-

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/util.h:
	  added option, --adapt-thres-type 2 (auto-level adjustment of the
	  ATH
	  with a frequency-based loudness approximation).

2001-03-27 10:56  aleidinger

	* configure:
	  regen (do not enable vorbis by default)

2001-03-27 10:56  aleidinger

	* configure.in:
	  do not enable vorbis by default

2001-03-26 23:02  markt

	* libmp3lame/set_get.c:
	  added code to more of the lame_set/lame_get functions-
	  but without any error checking :-(
	  
	  Only 47 more to go!

2001-03-26 20:46  markt

	* TODO, libmp3lame/util.c:
	  dont scale the input if scale=1.0.

2001-03-26 18:38  markt

	* TODO:
	  updated TODO list with some filtering stuff

2001-03-26 17:42  markt

	* libmp3lame/encoder.c:
	  another MPEG2 bug found by: "Jiri Hajek" <jirik@ucw.cz>
	  
	  This code:
	  
	  > int sum_pe_MS = pe_MS[0][0] + pe_MS[0][1] + pe_MS[1][0] +
	  pe_MS[1][1];
	  > int sum_pe_LR = pe [0][0] + pe [0][1] + pe [1][0] + pe [1][1];
	  
	  does not work for MPEG2. now fixed.

2001-03-26 16:05  aleidinger

	* configure:
	  regen (version bump)

2001-03-26 16:04  aleidinger

	* configure.in:
	  Welcome to 3.89

2001-03-26 00:38  markt

	* frontend/brhist.c:
	  bug in brhist_disp_total()
	  
	  for --freeformat bmode, br_frames = 0 (why?)
	  so sum / br_frames would segfault.
	  
	  if br_frames == 0 , routine now just returns without printing
	  anything.

2001-03-26 00:12  markt

	* TODO:
	  minor editing of TODO file

2001-03-25 23:18  markt

	* libmp3lame/version.h:
	  updated version to 3.89 alpha 1

2001-03-25 23:16  markt

	* libmp3lame/version.h:
	  updated version to 3.88beta

2001-03-25 23:14  markt

	* API, README, TODO, libmp3lame/lame.c, libmp3lame/libmp3lame.dsp:
	  made ath type 2 the default.
	  updated test cases
	  added set_get.c to libmp3lame.dsp

2001-03-25 21:37  shibatch

	* libmp3lame/psymodel.c:
	  Fixed nspsytune.

2001-03-25 18:43  afaber

	* Dll/BladeMP3EncDLL.def, Dll/MP3EncDll.dsp,
	  libmp3lame/bitstream.h, libmp3lame/encoder.h,
	  libmp3lame/psymodel.c, libmp3lame/tables.h:
	  Removed the last remainings of KLEMM_43 in order to complet the
	  Windows DLL compilation

2001-03-20 00:42  markt

	* libmp3lame/util.c:
	  added a few more minor comments to resampling code

2001-03-20 00:28  markt

	* libmp3lame/util.c:
	  removed some cruft from blackman() filtering routine

2001-03-19 21:26  markt

	* frontend/get_audio.c, libmp3lame/util.c, test/lametest.py:
	  Tested fixed resampling code. It seemed to work.
	  Added a few comments.
	  
	  Also added documentation for "-w" for the usage function in
	  lametest.py

2001-03-18 14:31  aleidinger

	* libmp3lame/vorbis_interface.c:
	  use the new MSGF API in Vorbis

2001-03-18 14:25  aleidinger

	* frontend/get_audio.c, libmp3lame/vorbis_interface.c:
	  use the new ERRORF API in Vorbis encoding (+fixes for needed API
	  change to lame_decode_ogg_{initfile,fromfile}())

2001-03-15 19:40  markt

	* libmp3lame/util.c:
	  Put in fix for resampling bug in fill_buffer_resample.
	  
	  Code assumed it would be fed at least BLACKSIZE samples.
	  For each input buffer, the code would save the last BLACKSIZE
	  samples for use for the next call to fill_buffer_resample.
	  
	  If the code recieves less than blacksize samples, it will
	  now shift in just as many new samples as are available into
	  the save buffer.
	  
	  Not tested :-(

2001-03-14 14:38  aleidinger

	* Dll/BladeMP3EncDLL.c, dshow/Encoder.cpp, dshow/Mpegac.cpp,
	  dshow/PropPage_adv.cpp:
	  regexp: MPG_MD_(.*) -> $1

2001-03-14 14:38  aleidinger

	* Makefile.B32, Makefile.DJGPP, Makefile.MSVC, Makefile.unix:
	  add set_get.c to the build

2001-03-14 10:57  aleidinger

	* Makefile.in:
	  regen

2001-03-14 10:56  aleidinger

	* Makefile.am:
	  do not forget Makfile.am.global at "make dist"

2001-03-14 10:49  aleidinger

	* ChangeLog:
	  update

2001-03-14 10:40  aleidinger

	* doc/html/Makefile.in, doc/man/Makefile.in,
	  libmp3lame/i386/Makefile.in:
	  regen with automake 1.4d instead of 1.4

2001-03-13 06:15  markt

	* libmp3lame/util.c:
	  added assert() to resampling code

2001-03-12 20:58  markt

	* libmp3lame/psymodel.c:
	  a few more comments added to psymodel.c

2001-03-12 20:40  markt

	* libmp3lame/psymodel.c, libmp3lame/tables.c:
	  removed ISO TABLE reading code from psymodel.c,
	  and ISO psy data from tables.c

2001-03-12 20:31  markt

	* libmp3lame/psymodel.c:
	  removed shortlived MINTHMM experiment

2001-03-12 18:52  markt

	* USAGE, configure:
	  updated USAGE.
	  ran configuer with --enable-maintainer-mode which seems to
	  have touched all this .in files, sorry about that...

2001-03-12 07:26  markt

	* API, libmp3lame/vbrquantize.c:
	  a few updates to API

2001-03-12 04:38  markt

	* frontend/get_audio.c, include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/encoder.h, libmp3lame/lame.c, libmp3lame/set_get.c:
	  fixed a few compiler warnings and the assert in lame_get_mode().
	  added NOT_SET value to MPEG_mode_e

2001-03-11 11:25  aleidinger

	* DEFINES:
	  - remove outdated entries
	  - add OLD_ATH_AUTO_ADJUST

2001-03-11 11:24  aleidinger

	* Makefile.in:
	  - regen from Makefile.am (testupdate target)

2001-03-11 11:24  aleidinger

	* frontend/depcomp, frontend/get_audio.c, frontend/gtkanal.c,
	  frontend/main.c, frontend/mp3x.c, frontend/parse.c,
	  frontend/timestatus.c:
	  *.c:
	  - use set/get functions
	  *** lame_set_quality(): should it check for <0 and >9 itself?
	  (parse.c)
	  
	  depcomp:
	  - update (autotools)

2001-03-11 11:23  aleidinger

	* libmp3lame/depcomp, libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/psymodel.c, libmp3lame/util.c, libmp3lame/util.h:
	  libmp3lame/encoder.c:
	  - MPEG_mode (JOINT_STEREO) change (related to set/get functions
	  change)
	  - J.D.s improvements to Gabriels ath auto adjust code (VBR)
	  - move ath auto adjust code into seperate function
	  
	  libmp3lame/lame.c:
	  - errorf/msgf/debugf changes
	  - MPEG_mode (JOINT_STEREO/...) change (related to set/get
	  functions change)
	  - remove prototypes of set/get functions
	  
	  libmp3lame/psymodel.c:
	  - MPEG_mode (JOINT_STEREO/...) change (related to set/get
	  functions change)
	  
	  libmp3lame/util.[ch]:
	  - new errorf/msgf/debugf implementation
	  
	  libmp3lame/depcomp:
	  - update (autotools)

2001-03-11 11:22  aleidinger

	* include/lame.h:
	  - set/get functions related cleanup
	  - errorf/msgf/debugf

2001-03-11 11:21  aleidinger

	* libmp3lame/Makefile.in:
	  regen

2001-03-11 11:21  aleidinger

	* libmp3lame/Makefile.am:
	  - add set_get.c

2001-03-11 11:20  aleidinger

	* libmp3lame/set_get.c:
	  - set and get functions for the gfp

2001-03-10 19:15  markt

	* libmp3lame/encoder.c:
	  a little more tuning of tot_energy for ATH adjust

2001-03-10 19:06  markt

	* libmp3lame/encoder.c:
	  modified ATH adjustment code to use tot_ener returned by psy
	  model
	  and not the maximum of the PCM sample values.
	  
	  We need to replace this with JD's formula. We need to decide
	  if this is really worth while. My preference is to take the
	  most conservative ATH possible, but this seems to produce too
	  high bitrates for VBR modes. (this may no longer be true
	  not that masings are computed in all scalefactor bands)

2001-03-10 18:33  markt

	* frontend/get_audio.c:
	  added 127 also, on recommendation of JD, to 8bit -> 16bit
	  conversion routine

2001-03-10 18:29  markt

	* frontend/get_audio.c:
	  fix from "J.D." <jodaman@cegt201.bradley.edu> for
	  8 bit input support.

2001-03-09 17:56  markt

	* test/lametest.py:
	  'os.pathsep' changed to 'os.sep'
	  
	  For example, on unix:
	  
	  os.pathsep = ":" - the character that seperaters search paths in
	  the PATH environment variable
	  
	  os.sep = "/" - the charcter that seperates directories

2001-03-07 12:32  aleidinger

	* test/lametest.py:
	  - use OS independent path seperator

2001-03-07 12:31  aleidinger

	* DEFINES:
	  - NOTABELS is now the default

2001-03-05 20:29  markt

	* libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c:
	  code (disabled) to use minimum value when going from partition
	  bands to scalefactor bands (instead of average value)

2001-03-05 18:21  markt

	* libmp3lame/psymodel.c, libmp3lame/quantize_pvt.c, testcase.mp3:
	  Noise calculation: had to switch calculation of tot_noise and
	  over_noise back to being done in db units. Multiply (to save
	  the 22 log10() function calls) was overflowing.
	  
	  updated testcases to reflect changes

2001-03-05 01:44  markt

	* testcase.mp3:
	  updated testcase.mp3 for NOTABLES code

2001-03-05 01:26  markt

	* libmp3lame/psymodel.c:
	  NOTABLES enabled by default.
	  
	  Psycho acoustics computed in *all* scalefactor bands.
	  
	  cvs tag before this change: before_notables
	  cvs tag after this change: notables

2001-03-05 01:25  markt

	* libmp3lame/psymodel.c:
	  The last commit before NOTABLES becomes the default.
	  tagged with "before.notables"

2001-03-04 20:03  markt

	* libmp3lame/encoder.c, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/util.h:
	  More prep work for enabling NOTABLES.
	  Will now compute masking for all scalefactor bands

2001-03-04 15:30  markt

	* libmp3lame/psymodel.c, libmp3lame/util.c, testcase.mp3:
	  Finished proper normalization of spreading function.
	  
	  updated testcases

2001-03-04 06:56  markt

	* libmp3lame/psymodel.c:
	  more tweaking of s3 normalization

2001-03-04 06:48  markt

	* libmp3lame/psymodel.c:
	  spreading function moved into seperate routine, s3_func().
	  s3 normalized so that:
	  
	  +inf
	  /
	  | s3 [ bark ] d(bark) = 1
	  /
	  -inf
	  
	  which resulted in dividing the result by 0.66.
	  
	  This is the first step in fixing the long outstanding bug
	  mentioned by Takehiro: s3[i][j] is normalized improperly for
	  values of i near the partition band boundaries. For these values,
	  the computed 'norm' is much too small since a lot of the data is
	  missing, resulting in artificially increasing the spreading
	  function
	  (and thus the masking)
	  
	  --nspsytune: s3_l function is used with a normalization of .5
	  (instead of computing the norm). So I changed this to a .5*.66
	  to compensate for the above change.
	  
	  Next step:
	  
	  1.
	  s3[i][j] will be multiplied by the width (in barks) of the
	  j'th partition band (about .35). Once this is done, we will not
	  need to
	  compute and divide by 'norm' later.
	  
	  
	  2.
	  Make NOTABLES the default
	  
	  3.
	  compute psymodel all the way up to 22khz.

2001-03-04 05:51  markt

	* API, libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  added prototypes to lame.c (but functions not yet written)

2001-03-04 05:13  markt

	* include/lame.h:
	  finished last of the prototypes

2001-03-04 05:02  markt

	* frontend/get_audio.c:
	  changed to:
	  
	  gfp->num_samples = data_length / (channels * ((bits_per_sample+7)
	  / 8));
	  
	  thanks J.D.!

2001-03-04 00:20  markt

	* include/lame.h, libmp3lame/lame.c:
	  added:
	  lame_encode_buffer_sample_t, which is called by the following
	  functions
	  in the API:
	  
	  lame_encode_buffer (original, short int interface)
	  lame_encode_buffer_float
	  lame_encode_buffer_long

2001-03-02 22:39  markt

	* frontend/get_audio.c, test/lametest.py:
	  get_audio.c: fixed num_samples computation for case when
	  bits_per_sample < 8. (althouth this case is already trapped for
	  earlier)
	  
	  lametest.py: slight change to reference mp3 filename generation
	  so lametest.py can be run from another directory.

2001-03-02 18:28  markt

	* Makefile.am, doc/html/Makefile.in, doc/man/Makefile.in,
	  frontend/get_audio.c, libmp3lame/i386/Makefile.in, testcase.mp3:
	  error message about non-PCM wave files from
	  "J.D." <jodaman@cegt201.bradley.edu>
	  
	  Added "make testupdate" to Makefile.am, which updates
	  the test file in the CVS repository

2001-02-27 11:54  robert

	* include/lame.h, libmp3lame/lame.c:
	  lame.h typo
	  lame.c fix for possibility of calling lame_init_parms twice

2001-02-27 11:39  aleidinger

	* DEFINES:
	  remove KLEMM_42

2001-02-27 11:38  aleidinger

	* Makefile.in, configure, frontend/Makefile.in, misc/Makefile.in:
	  regen

2001-02-27 11:37  aleidinger

	* frontend/Makefile.am, misc/Makefile.am:
	  remove EXTRA_PROGRAMS on make clean

2001-02-27 11:36  aleidinger

	* Makefile.am:
	  remove testcase.new.mp3 on error

2001-02-27 11:36  aleidinger

	* configure.in:
	  display version of gcc

2001-02-27 11:35  aleidinger

	* include/lame.h:
	  fix typo in comment

2001-02-27 11:33  aleidinger

	* libmp3lame/util.c:
	  include machine/floatingpoint.h on FreeBSD

2001-02-27 09:59  robert

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c,
	  libmp3lame/util.c, libmp3lame/util.h:
	  VbrTag.c fixed gfp/gfc typo
	  lame.c made data flow analysis happier: there was a possible path
	  to use channels before definition. Even though it would be an
	  error following this path, but the compiler doesn't know it
	  util.hc message functions are not intended to change gfc?!?
	  takehiro.c fixed a warning
	  bitstream.c fixed a warning

2001-02-27 06:14  markt

	* frontend/gtkanal.c, include/lame.h, libmp3lame/VbrTag.c,
	  libmp3lame/bitstream.c, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  Implmented Frank's excellent idea about error messages:
	  
	  gfc->errorf, gfc->msgf and gfc->debugf are all FILE *.
	  default is stderr. If you want to disable all internal library
	  messages, set these to NULL with lame_set_errorf(),
	  lame_set_msgf() etc...
	  
	  For a GUI, you can set these to regular files, and then read
	  the files and display the messages in an pop up window.
	  
	  Maybe the default should be NULL?

2001-02-27 05:17  markt

	* include/lame.h:
	  plugging away at a few more prototypes

2001-02-26 22:45  robert

	* frontend/parse.c, libmp3lame/util.c:
	  little change to Roel's --r3mix setting, on request by Roel

2001-02-26 19:08  markt

	* libmp3lame/VbrTag.c:
	  non RH_SEEKTABLE code removed from VbrTag.c

2001-02-26 18:57  markt

	* frontend/lametime.c:
	  removed to //include lines

2001-02-26 18:52  markt

	* frontend/get_audio.c, frontend/lametime.c:
	  incomprehensible gobbledygook replaced by call to standard
	  'stat' function for file size.

2001-02-26 08:06  markt

	* include/lame.h:
	  more prototypes added to lame.h

2001-02-26 07:38  markt

	* USAGE, frontend/parse.c, libmp3lame/lame.c:
	  useTemporal turned on by default
	  --notemp option added.

2001-02-26 07:26  markt

	* include/lame.h:
	  added more of the lame_get/lame_set prototypes to lame.h
	  
	  also added this:
	  
	  
	  // I give up. Klemm has worn me down on this one...
	  typedef lame_global_flags lame_t;

2001-02-26 06:34  markt

	* misc/Lame.vbs:
	  lame.vbs changes submitted by "Paul Reedy" <ka5pmv@webzone.net>

2001-02-25 22:47  markt

	* libmp3lame/vorbis_interface.c:
	  removed a klemm_42 from vorbis_interface.c

2001-02-25 22:47  markt

	* INSTALL, libmp3lame/lame.c, libmp3lame/vorbis_interface.c:
	  klemm_42 enabled by default.
	  
	  This code uses a different formula to choose the lowpass
	  filter. It takes into account stereo vs. jstereo
	  (jstereo gets 25% more bandwidth). This has the effect
	  of stereo 128kbps using a 14khz lowpass filter, which some
	  people may not like.

2001-02-25 22:02  markt

	* libmp3lame/lame.c, test/CBRABR.op:
	  made allow_diff_short the default if mode=stereo and
	  allow_diff_short has not been set to 0 or 1 by the calling
	  program.
	  
	  addes these notes to lame_init_params() :
	  *
	  * We first have some complex code to determine bitrate,
	  * output samplerate and mode. It is complicated by the fact
	  * that we allow the user to set some or all of these parameters,
	  * and need to determine best possible values for the rest of
	  them:
	  *
	  * set some CPU related flags
	  * check if we are mono->mono, stereo->mono or stereo->stereo
	  * compute bitrate and output samplerate:
	  * user may have set compression ratio
	  * user may have set a bitrate
	  * user may have set a output samplerate
	  * set some options which depend on output samplerate
	  * compute the actual compression ratio
	  * set mode based on compression ratio
	  *
	  * The remaining code is much simpler - it just sets options
	  * based on the mode & compression ratio:
	  *
	  * set allow_diff_short based on mode
	  * select lowpass filter based on compression ratio & mode
	  * set the bitrate index, and min/max bitrates for VBR modes
	  * disable VBR tag if it is not appropriate
	  * initialize the bitstream
	  * initialize scalefac_band data
	  * set sideinfo_len (based on channels, CRC, out_samplerate)
	  * write an id3v2 tag into the bitstream
	  * write VBR tag into the bitstream
	  * set mpeg1/2 flag
	  * estimate the number of frames (based on a lot of data)
	  *
	  * now we set more flags:
	  * nspsytune:
	  * see code
	  * VBR modes
	  * see code
	  * CBR/ABR
	  * see code

2001-02-25 21:28  markt

	* frontend/parse.c:
	  removed VBR warning

2001-02-25 20:24  markt

	* libmp3lame/psymodel.c:
	  edited a few comments in psymodel.c

2001-02-25 20:17  markt

	* libmp3lame/encoder.c, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/util.h, test/nores.op:
	  Started some re-originization of psymodel.c
	  created:
	  
	  psymodel_init() function, called by both psymodel and
	  psymodel_ns.
	  
	  I was carefull to make sure no binary changes were introduced
	  with --nspsytune.
	  
	  Modified both psymodel and psymodel_ns to return the
	  energy in each channel. I'm going to request that the
	  ATH level adjust be based on the energy comptuted by the
	  psymodel, rather than computing a magnitude in encoder.c

2001-02-25 17:37  markt

	* Dll/BladeMP3EncDLL.c, frontend/parse.c, include/lame.h,
	  libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  removed mode_fixed variable. If calling program selects a mode
	  (mono, stereo, jstereo) then that mode is used, otherwise
	  lame picks best mode to use.

2001-02-25 01:58  markt

	* include/lame.h:
	  started adding more prototypes into lame.h:
	  
	  int lame_set_variable_name(lame_global_flags *, int)
	  int lame_get_variable_name(lame_global_flags *)
	  
	  _get will return the value (instead of an error code)
	  as suggested.
	  
	  _set will return an error code. But right now, the way lame
	  works there is no error checking until lame_init_params() is
	  called. This is because, for example, bitrate=135
	  is not an error if free_format=1, but it is an error
	  for CBR. So many errors cannot be detected until all options
	  are set.

2001-02-24 19:35  markt

	* include/lame.h, libmp3lame/util.h:
	  commented out a uint_64_t field in util.h to get lame to
	  compile. Do we really need these types of variables?
	  
	  removed some stuff from lame.h. Note that many decisions have
	  already been made about the library interface, so if you are
	  planning on changing this, you need to check with me
	  (mark taylor) first. In particular,
	  there is going to be no file I/O, and all data is in
	  machine order. (no endian stuff in the library).

2001-02-24 13:55  robert

	* Makefile.unix, configMS.h:
	  get Makefile.unix working again (tested for Linux, not for OS/2
	  etc.)

2001-02-21 21:17  robert

	* libmp3lame/lame.c:
	  --vbr-mtrh related

2001-02-21 17:07  robert

	* libmp3lame/vbrquantize.c:
	  take sfb21 into consideration for median search

2001-02-21 17:06  robert

	* libmp3lame/util.c:
	  index transformation in select_kth_int()

2001-02-21 14:59  aleidinger

	* test/lametest.py:
	  do not print the passed/failed statistic in the makeref case

2001-02-21 13:49  robert

	* frontend/parse.c, libmp3lame/lame.c:
	  default quality handling fix

2001-02-21 12:39  aleidinger

	* Dll/BladeMP3EncDLL.c:
	  lame.h has to be included before version.h

2001-02-20 18:34  aleidinger

	* Makefile.in, config.h.in, configure, configure.in,
	  frontend/brhist.c:
	  - Patch (#103821) from guy_hutchison:
	  Fixes compilation problem on systems with ncurses dir
	  - add ChangeLog to Makefile.in

2001-02-20 18:23  aleidinger

	* libmp3lame/encoder.h, libmp3lame/version.h:
	  *** empty log message ***

2001-02-20 18:20  aleidinger

	* doc/html/Makefile.am, doc/html/Makefile.in:
	  - add basic.html

2001-02-20 18:20  aleidinger

	* configMS.h:
	  - add LAME_LIBRARY_BUILD

2001-02-20 18:20  aleidinger

	* libmp3lame/util.h:
	  - more documentation, seems to be partially truncated (Frank)
	  - CPU_features is now a struct

2001-02-20 18:19  aleidinger

	* libmp3lame/takehiro.c:
	  - CPU_features is now a struct

2001-02-20 18:19  aleidinger

	* libmp3lame/lame.c:
	  - CPU_features is now a struct, no need to waste space for those
	  bolean values (Frank)
	  - introduce some KLEMM_44 protected code

2001-02-20 18:18  aleidinger

	* libmp3lame/util.c:
	  - fixes + docu for ATHformula_Frank() (Frank)
	  - don't compile old resampling in KLEMM_44 case

2001-02-20 18:17  aleidinger

	* libmp3lame/version.c:
	  - doxygen parseable documentation
	  - better lclint flags
	  - enable get_lame_version_numerical()

2001-02-20 18:17  aleidinger

	* include/lame.h:
	  - delete prototype of already removed get_lame_about()
	  - add get_lame_version_numerical() and lame_version_t
	  (now dshow didn't needs to include version.h to get the numerical
	  version of lame (dshow not changed, I let pgubanov do this))
	  - introduce some KLEMM_44 protected code (Frank)
	  - rework formatting and documentation to be more (IMO much more)
	  readable
	  (no lines longer than 80 chars, consistent formatting, looks alot
	  cleaner)

2001-02-20 10:41  aleidinger

	* DEFINES:
	  Remove KLEMM_04, there was a previous commit which handled it,
	  see
	  ChangeLog.

2001-02-20 01:56  markt

	* DEFINES, frontend/brhist.c, frontend/get_audio.c,
	  frontend/timestatus.c:
	  KLEMM 5,6,7 made the default & removed

2001-02-20 00:40  markt

	* libmp3lame/lame.c:
	  simplified lame_encoder_buffer_interleaved: it now always calls
	  lame_encoder_buffer()

2001-02-20 00:30  markt

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  fix for --scale problem?
	  
	  made a new routine: fill_buffer()
	  
	  This is the generic routine used to copy data from the input
	  buffer
	  into lame's internal buffer, 'mfbuf[]'.
	  
	  It also calls the resample code if necessary, and the user
	  selected
	  rescaling code if necessary.

2001-02-20 00:26  markt

	* frontend/parse.c:
	  added comment to vbr warning message

2001-02-20 00:24  markt

	* frontend/parse.c:
	  re-added the warning about VBR modes. It seems to have been
	  deleted
	  at some point.
	  
	  This is because this website:
	  
	  http://www.kuro5hin.org/?op=displaystory;sid=2001/1/25/164540/759
	  
	  is comparing Vorbis with lame -V4.
	  
	  While many people like the high bitrate VBR modes,
	  lame -V4 is *often* of lower quality than CBR.

2001-02-19 18:29  aleidinger

	* test/VBR.op:
	  - remove "--silent", lametest.py now handles this
	  
	  Aproved by: robert

2001-02-19 18:28  aleidinger

	* test/lametest.py:
	  - allow relative paths in a more portable manner
	  - should be more OS independand now
	  - more file checks (readable/executable)
	  - internally adds new option "--quiet" to the options
	  
	  This works on my system, it also should work on systems where the
	  previous version worked. It's not as portable as it could be,
	  search for XXX in the source.

2001-02-18 21:18  robert

	* frontend/main.c, frontend/parse.c, libmp3lame/lame.c,
	  libmp3lame/version.h, test/VBR.op:
	  new options --silent and --quiet to be really quiet
	  new option --verbose (default) to turn -S/--silent off again
	  --vbr-mtrh and --vbr-new use now ATH type 2
	  for VBR modes it is allowed to use different block types in
	  stereo mode (not yet joint stereo)
	  bumped version to 3.88 alpha 10

2001-02-18 12:11  aleidinger

	* dshow/Makefile.am, dshow/Makefile.in:
	  Add recently added files to the makefile.

2001-02-17 15:45  pgubanov

	* dshow/Encoder.cpp:
	  gfp is sometimes zero in CEncoder::Encode() - why???

2001-02-17 14:30  aleidinger

	* libmp3lame/pcm.c, libmp3lame/pcm.h:
	  New code from Frank, some needed parts in different files aren't
	  commited
	  yet, they produce huge differences with lametest.py even if
	  KLEMM_44 isn't
	  defined. -> KLEMM_44 (new API + resampling) isn't usable yet

2001-02-17 14:27  aleidinger

	* test/lametest.py:
	  - use "/usr/bin/env python" instead of "/usr/bin/python" in
	  shebang line
	  (portability)
	  - use "-l" instead of "--lines" for wc
	  (portability, has to run on every IEEE 1003.2 ("POSIX.2")
	  compliant
	  system)
	  - bugfix, check for lame2 instead for lame1 in the lame2 part
	  (Note: the entire test didn't work for me if I use e.g.
	  ./lametest.py XXX.op ../../testcases/XXX
	  ../../build.org/frontend/lame ...)

2001-02-17 14:17  pgubanov

	* dshow/aboutprp.cpp, dshow/aboutprp.h:
	  Elecard copyrigth added (executives need it for some reason)

2001-02-17 13:58  pgubanov

	* dshow/Mpegac.cpp, dshow/Property.rc, dshow/UIDS.H,
	  dshow/dshow.dsp, dshow/dshow.dsw, dshow/elogo.ico,
	  dshow/resource.h:
	  Some stability improvements + Elecard copyrigth added (executives
	  need it for some reason)

2001-02-16 17:23  markt

	* doc/html/basic.html:
	  added basic.html to cvs

2001-02-16 10:35  pgubanov

	* dshow/dshow.dsp:
	  STDC_HEADERS defined in project settings

2001-02-16 10:33  pgubanov

	* libmp3lame/lame.c:
	  mono handling in encode_buffer_interleaved caused access
	  violation - fixed

2001-02-14 22:22  aleidinger

	* ChangeLog:
	  Now we have a ChangeLog, generated from the output of "cvs log"

2001-02-14 22:21  aleidinger

	* ChangeLog.header:
	  Header for the ChangeLog, not included at "make dist"

2001-02-14 22:20  aleidinger

	* TODO:
	  remove Makefile.MSVC, robert seems to keep it up to date

2001-02-14 22:20  aleidinger

	* Makefile.B32:
	  try to unbreak it, add configMS.h -> config.h (untested)

2001-02-14 22:20  aleidinger

	* DEFINES:
	  add USE_FFT3DN

2001-02-14 22:19  aleidinger

	* INSTALL:
	  * add some more info how to use some advanced features of our
	  configure script
	  * add some info for Cygwin users (use configure)
	  * add some info about faster code with nasm

2001-02-14 20:50  robert

	* frontend/parse.c:
	  added -V1 to Roel's --remix preset, I assume it was missing by
	  accident

2001-02-14 16:38  aleidinger

	* Dll/Makefile.in, Makefile.am.global, Makefile.in, aclocal.m4,
	  configure, configure.in, debian/Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  use new automake 1.4d (bugfixes)

2001-02-14 14:43  robert

	* libmp3lame/encoder.c, libmp3lame/util.h:
	  tuning of the auto ath adjustment for --vbr-mtrh mode

2001-02-14 14:24  robert

	* libmp3lame/util.c:
	  patch by Roel:
	  tweaking his ATH modification (--athtype 3)

2001-02-13 18:27  markt

	* frontend/parse.c:
	  fixed typo in parse.c

2001-02-13 18:24  markt

	* libmp3lame/psymodel.c:
	  added a lot of documentation on the psycho acoustics
	  to the top of psymodel.c

2001-02-13 18:22  markt

	* mpglib/common.c, mpglib/common.h, mpglib/interface.c:
	  improved head_check() function: When resyncing data stream,
	  only look for headers of the same layer of the previous
	  frames. i.e. dont allow layer 2 and layer 3 frames in the
	  same stream.

2001-02-13 18:21  markt

	* frontend/get_audio.c, frontend/parse.c:
	  added the --r3mix option
	  fixed bug with --decoder-mp3delay option
	  added error message for non-8 and 16 bit input

2001-02-13 16:58  robert

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.c, libmp3lame/util.h:
	  first of all: this patch shows no difference in the CBR/ABR test
	  
	  - made athtype 3 default for VBR modes, 1 still default for
	  CBR/ABR modes
	  - made athadjust default for VBR modes, CBR/ABR modes don't need
	  it
	  - removed --athadjust switch, was only temporally solution
	  - collected all spreaded ATH settings into ATH_t struct
	  if someone needs a new ATH related variable, please add it here

2001-02-12 18:24  markt

	* doc/html/contributors.html, doc/html/history.html,
	  doc/html/index.html, doc/html/node6.html, doc/html/switchs.html:
	  new html docs from Gabriel

2001-02-12 17:47  markt

	* test/lametest.py:
	  print total pass/fail message at end

2001-02-09 16:47  robert

	* libmp3lame/quantize_pvt.c:
	  patch by Gaby:
	  allow ATH adjust for the ATH only modes too

2001-02-08 22:19  robert

	* test/VBR.op:
	  update to the VBR python options file

2001-02-07 15:51  robert

	* libmp3lame/encoder.c, libmp3lame/vbrquantize.c:
	  all patches --vbr-mtrh related
	  
	  encoder.c:
	  - made a variant of Gaby's auto ATH adjust the default
	  
	  vbrquantize.c:
	  - some small code moves
	  - added some safety margin to the global gain

2001-02-06 21:43  robert

	* libmp3lame/psymodel.c, libmp3lame/vbrquantize.c:
	  we should not try to hard to get high freqs thru'
	  so I will let --vbr-mtrh as is. It seems to give
	  reasonable file sizes again, even with Frank's ATH.

2001-02-05 10:18  robert

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/quantize_pvt.c:
	  re-check-in of Gaby's ATH auto adjustment code.
	  You need to specify --athadjust to get this
	  SUBSTANTIALLY QUALITY IMPROVEMENT
	  
	  quote:
	  "Also, we already have gfc->ath_lower (specified in db).
	  No reason to add another variable which does the same thing,
	  and does not even use the same units."
	  
	  That is not true, there is actually no gfc->ath_lower in dB.
	  We have a gfp->ATHlower in dB. But this is of a totally
	  different meaning.
	  gfp->ATHlower will lower the ATH for the whole file by a fixed
	  amount of x dB.
	  gfc->ath_auto_adjust will lower the ATH on a frame by frame basis
	  depending on the peak volume.

2001-02-05 04:58  markt

	* frontend/parse.c:
	  Added warning about using VBR modes.
	  This is because this website:
	  
	  http://www.kuro5hin.org/?op=displaystory;sid=2001/1/25/164540/759
	  
	  is comparing Vorbis with lame -V4.
	  
	  While many people like the high bitrate VBR modes,
	  lame -V4 is *often* of lower quality than CBR.

2001-02-05 02:36  shibatch

	* libmp3lame/psymodel.c:
	  Reverted part of last tweaks of nspsytune.

2001-02-05 02:27  markt

	* HACKING:
	  updated HACKING with directions on how to run the
	  lametest.py python script

2001-02-05 02:24  markt

	* test, test/CBRABR.op, test/VBR.op, test/lametest.py,
	  test/nores.op:
	  Adding a python script used for testing, along with
	  some sample options files.

2001-02-05 02:23  markt

	* HACKING, libmp3lame/encoder.c, libmp3lame/quantize_pvt.c:
	  reverted encoder.c and quantize_pvt.c back to 13:00 3 Feb 2001.
	  These modifications broke my test cases. Any such changes need
	  aproval first.
	  
	  Also, we already have gfc->ath_lower (specified in db).
	  No reason to add another variable which does the same thing,
	  and does not even use the same units.

2001-02-05 00:34  cisc

	* frontend/amiga_mpega.c:
	  MPG_MD_LR_LR define was moved, set mode_ext to 0 instead (since
	  we don't get that info anyway).

2001-02-04 23:56  robert

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/vbrquantize.c:
	  parse.c: small change in default behaviour for --vbr-mtrh
	  lame.c: it makes no sense for a sfb21 extra at sample frequencies
	  below 44.1 kHz
	  vbrquantize.c: oops, last commit was a merger from a wrong branch
	  of mine

2001-02-04 23:28  shibatch

	* libmp3lame/quantize_pvt.c:
	  Enabled auto-adjust ATH for nspsytune.

2001-02-04 23:14  shibatch

	* libmp3lame/psymodel.c:
	  Tweaked nspsytune.

2001-02-03 23:20  robert

	* libmp3lame/quantize_pvt.c, libmp3lame/version.h:
	  bumped version to alpha 9
	  some clean-up

2001-02-03 20:10  robert

	* libmp3lame/encoder.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.c, libmp3lame/util.h, libmp3lame/vbrquantize.c:
	  patch by Gabriel Bouvigne:
	  auto adjust of ATH for low volume tracks
	  a big improvement in quality!
	  therefore on by default
	  
	  patch by Robert Hegemann:
	  further improvement of --vbr-mtrh mode

2001-02-02 01:04  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  --vbr-mtrh tweaks

2001-01-31 19:44  aleidinger

	* configure, configure.in:
	  bugfix

2001-01-31 15:34  robert

	* libmp3lame/quantize.c:
	  fix a typo I introduced by my last commit

2001-01-31 12:53  aleidinger

	* Dll/Makefile.in, Makefile.am.global, Makefile.in, acconfig.h,
	  acinclude.m4, aclocal.m4, config.h.in, configMS.h, configure,
	  configure.in, debian/Makefile.in, depcomp, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in:
	  - config*/Makefile*/ac*:
	  * tests for *int*_t
	  * test for 80 bit floats (for speed optimizations)
	  * add pcm.* to the makefile
	  * removed resample.[ch] from the makefile
	  * depend upon automake 1.4b
	  * fix in check for ccc
	  (submitted by nold via the sourceforge patch page at
	  the lame project page, patch id #103507)
	  * something I may have forgotten
	  
	  Note:
	  Please don't commit any Makefile.in if you didn't have
	  changed the corresponding Makefile.am. CVS didn't stores
	  the time stamps of the files at commit time, it stores
	  the time stamps of the commit time, so we have to commit
	  the files in the correct order, but I'm to lazy to do so
	  and I assume I'm not the only one.
	  The same applies to config*.
	  
	  - depcomp:
	  missing file, the dependancy tracking should work again

2001-01-30 20:53  robert

	* libmp3lame/quantize.c:
	  fixed a design flaw in VBR_iteration_loop() I'm responsible for.
	  affects quality!

2001-01-30 20:52  robert

	* libmp3lame/lame.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/vbrquantize.c:
	  tuning of --vbr-mtrh mode

2001-01-30 20:51  robert

	* frontend/parse.c, include/lame.h:
	  patch by Gaby:
	  changed --athlower to accept floats

2001-01-30 11:02  robert

	* libmp3lame/psymodel.c, libmp3lame/util.c:
	  psymodel.c: fixing compile problem (was: M_LN10 undefined)
	  util.c: Roel's modification to Gaby's ATH, new: --athtype 3

2001-01-29 14:46  robert

	* libmp3lame/vbrquantize.c:
	  --vbr-mtrh and --vbr-new use now the faster and sloppier
	  scalefactor-band noise calculation with quality settings
	  -q 3...9

2001-01-28 12:51  takehiro

	* libmp3lame/quantize.h, libmp3lame/util.h:
	  oops, compile problem fix

2001-01-28 12:39  takehiro

	* libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/vbrquantize.c:
	  magic number 4095 and 4096 is removed

2001-01-27 18:00  takehiro

	* libmp3lame/psymodel.c:
	  make temporal masking effect strong.

2001-01-27 17:52  takehiro

	* libmp3lame/quantize_pvt.c:
	  debug message is removed :)

2001-01-27 17:39  takehiro

	* frontend/parse.c, include/lame.h, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.h:
	  experimental temporal masking support.
	  "--temporal" option to use this.

2001-01-26 22:47  robert

	* frontend/parse.c, libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/vbrquantize.c:
	  tuning of --vbr-mtrh VBR mode

2001-01-25 20:57  robert

	* misc/abx.c:
	  update by Frank:
	  some further endings and audio formats

2001-01-24 23:59  robert

	* libmp3lame/quantize_pvt.c:
	  found and fixed another BUG caused by the "bit optimization."
	  patch

2001-01-24 23:02  robert

	* libmp3lame/pcm.c, libmp3lame/pcm.h:
	  updates by Frank

2001-01-24 22:33  robert

	* Makefile.DJGPP, Makefile.unix:
	  small config.h related tunings

2001-01-24 21:33  robert

	* libmp3lame/vbrquantize.c:
	  quick-fix for code that was broken by Takehiro's last
	  "bit optimization." in function calc_xmin(), quantize_pvt.c

2001-01-23 12:50  takehiro

	* libmp3lame/quantize_pvt.c:
	  bit optimization.

2001-01-23 07:12  shibatch

	* libmp3lame/quantize.c:
	  Bugfix of nspsytune.

2001-01-22 23:00  afaber

	* Dll/BladeMP3EncDLL.c:
	  VBR was always enabled, even when the client did not ask for it

2001-01-22 19:54  aleidinger

	* libmp3lame/vorbis_interface.c:
	  remove trailing garbage

2001-01-22 09:31  shibatch

	* libmp3lame/psymodel.c, libmp3lame/quantize.c:
	  Tweaked nspsytune.

2001-01-22 04:35  markt

	* Dll/Makefile.in, Makefile.in, configure, debian/Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, frontend/parse.c,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, libmp3lame/psymodel.c,
	  misc/Makefile.in, mpglib/Makefile.in:
	  fixed bug preventing -m a mode from working.
	  
	  ran autoconf which changed all the Makefile.in's

2001-01-22 01:07  robert

	* libmp3lame/pcm.c, libmp3lame/pcm.h,
	  libmp3lame/vorbis_interface.c:
	  patches by Frank:
	  
	  vorbis_interface: was/is broken since a long time
	  pcm.[ch]: fixing some transfer errors (email)

2001-01-21 21:14  afaber

	* Dll/BladeMP3EncDLL.c:
	  Fixed problems with writing the Xing MP3 header
	  (gfc->internal_flags structure was already freed, while it was
	  needed to write the Xing MP3 tag)

2001-01-21 19:13  robert

	* libmp3lame/pcm.c, libmp3lame/pcm.h:
	  patches by Frank:
	  
	  scalar.nas - new SIMD assembler routines
	  resample.[ch] - removed
	  pcm.[ch] - new, containing resampling routines and more
	  version.c - some clean up

2001-01-21 19:07  robert

	* libmp3lame/i386/scalar.nas, libmp3lame/resample.c,
	  libmp3lame/resample.h, libmp3lame/version.c:
	  patches by Frank:
	  
	  scalar.nas - new SIMD assembler routines
	  resample.[ch] - removed
	  pcm.[ch] - new, containing resampling routines and more
	  version.c - some clean up

2001-01-21 19:03  robert

	* frontend/lametime.c:
	  it's always a good idea to include its own header file

2001-01-20 09:51  robert

	* libmp3lame/util.c:
	  patch by Gaby:
	  
	  I changed a file from the 19/01 cvs (I added an ath balanced
	  between the
	  original one and Frank's one on the option --athtype 2)

2001-01-19 09:08  shibatch

	* libmp3lame/encoder.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c:
	  Improved reservoir handling(nspsytune).

2001-01-18 22:18  robert

	* libmp3lame/quantize_pvt.c:
	  allow "--athlower n" to modify the ATH for sfb21

2001-01-18 13:04  aleidinger

	* Makefile.DJGPP:
	  add config.h to the clean target

2001-01-18 13:03  aleidinger

	* Makefile.DJGPP:
	  bugfix (please report if it's broken, I can't test the config.h
	  target)

2001-01-17 21:32  robert

	* Makefile.unix:
	  additional fixes

2001-01-17 20:35  robert

	* libmp3lame/VbrTag.c, libmp3lame/util.c, libmp3lame/util.h:
	  RH_SEEK_TABLE enables a different method of generating TOC
	  entries
	  in Xing VBR header. now enabled!
	  
	  PLEASE TEST TRY IT
	  
	  It should fix the problem poeple reported who encode radio shows
	  or
	  alike which last a few hours.
	  Another advantage is, it will not steadily increase LAME's memory
	  footprint
	  while running.

2001-01-17 16:05  shibatch

	* libmp3lame/psymodel.c:
	  Reverted the final change.
	  It degraded encoded quality of gspi*.wav.

2001-01-17 13:26  shibatch

	* libmp3lame/psymodel.c:
	  Tweaked nspsytune.

2001-01-17 09:55  robert

	* Makefile.unix:
	  try to fix the broken Makefile after latest configurification

2001-01-17 09:54  robert

	* libmp3lame/newmdct.c:
	  + replaced M_SQRT2 by SQRT2 which is defined in util.h
	  + replaced "=-" by "= -" to avoid confusion with the
	  old K&R "=-" operator, an equivalent to "-=". Even
	  though new compilers seem to handle it as an
	  assignement and unary minus sequence, but some
	  buggy compiler could be trapped here (as I).

2001-01-16 13:08  robert

	* configMS.h:
	  turned some defines for system include files on

2001-01-16 09:04  robert

	* Makefile.MSVC:
	  fixed typo, sorry

2001-01-15 15:16  aleidinger

	* Dll/Makefile.am, Dll/Makefile.in, Makefile.am,
	  Makefile.am.global, Makefile.in, config.h.in, configMS.h,
	  configure, configure.in, debian/Makefile.am, debian/Makefile.in,
	  doc/Makefile.am, doc/Makefile.in, dshow/Makefile.am,
	  dshow/Makefile.in, frontend/Makefile.am, frontend/Makefile.in,
	  frontend/brhist.c, frontend/console.c, frontend/get_audio.c,
	  frontend/gpkplotting.c, frontend/main.c, frontend/mp3rtp.c,
	  frontend/parse.c, frontend/rtp.c, include/Makefile.am,
	  include/Makefile.in, libmp3lame/Makefile.am,
	  libmp3lame/Makefile.in, libmp3lame/id3tag.c,
	  libmp3lame/machine.h, misc/Makefile.am, misc/Makefile.in,
	  mpglib/Makefile.am, mpglib/Makefile.in, mpglib/common.c,
	  mpglib/decode_i386.c, mpglib/mpg123.h:
	  automake/autoconf cleanup and enhancement

2001-01-15 14:19  shibatch

	* libmp3lame/encoder.c, libmp3lame/psymodel.c, libmp3lame/util.h:
	  More cleanups and bugfixes of --nspsytune.

2001-01-15 09:38  aleidinger

	* doc/html/Makefile.am:
	  remove node[57].html

2001-01-15 09:14  gramps

	* USAGE:
	  Raised track limit from 99 to 255 per user requests.

2001-01-15 09:10  robert

	* Makefile.B32, Makefile.MSVC, Makefile.unix:
	  more RH_AMP clean up

2001-01-15 09:00  gramps

	* doc/html/id3.html, frontend/parse.c:
	  Raised track limit from 99 to 255 per user requests.

2001-01-15 08:48  gramps

	* libmp3lame/id3tag.c:
	  - Simplified prologue of all public functions, removing
	  intermediate "spec"
	  variable, and assuming (like the rest of LAME) that gfp is never
	  null.
	  - Fixed a stupid sizeof bug (mine) in "id3tag_init".
	  - Raised track limit from 99 to 255 per user requests.
	  - No longer write full text of genre in version 2 tag. Now write
	  more
	  cannonical "(num)" format where "num" is the text version of the
	  genre
	  number, more closely following the updated ID3v2.3 spec and usage
	  in id3lib.
	  This also makes the tag smaller.
	  - Now use id3lib-compatible bogus language descriptor "XXX" in
	  version 2
	  comment frames instead of just null bytes.
	  - Changed order of frames written in version 2 tag to match
	  version 1 tag.
	  - Restored some of Frank's optimizations.
	  - Fixed LAME URL.
	  - Reworded some comments and fixed misspellings.
	  - Moved "local_strcasecmp" back into its original position.
	  - Some small reformatting to match original file.

2001-01-14 23:29  gramps

	* USAGE:
	  - Changed "highest" to "higher" to describe recommended quality
	  setting in
	  constant bit rate example.
	  
	  - Changed "CBR" to "VBR" to correct misspelling in variable bit
	  rate
	  description.

2001-01-14 22:40  gramps

	* doc/html/Makefile.in:
	  Removed dependency on obsolete node5.html and node7.html.

2001-01-14 22:35  markt

	* USAGE, frontend/parse.c:
	  documentation of -q parameter in USAGE and lame --longhelp

2001-01-14 22:13  markt

	* doc/html/node5.html, doc/html/node7.html:
	  removed as per Gaby

2001-01-14 22:12  markt

	* libmp3lame/lame.c, libmp3lame/quantize.c:
	  created new routine lame_init_qval() which does all the
	  gfp->quality related settings. Cleaned up some more RH_AMP code:
	  The only way to enable scalefac_scale right now is with
	  -Z or --nspsytune.

2001-01-14 21:01  markt

	* testcase.mp3:
	  updated test case
	  (Takehiro fixed a bug which I introduced, which changes test
	  cases).

2001-01-14 17:28  aleidinger

	* analyzer:
	  obsolete

2001-01-14 17:25  aleidinger

	* DEFINES, Dll/Makefile.am, Dll/Makefile.in, Makefile.in,
	  aclocal.m4, config.guess, config.sub, configure, configure.in,
	  debian/Makefile.in, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, dshow/Makefile.in, frontend/Makefile.in,
	  frontend/depcomp, include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/depcomp, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, misc/depcomp, missing, mkinstalldirs,
	  mpglib/Makefile.in, mpglib/depcomp:
	  DEFINES:
	  remove obsolete defines
	  
	  configure*:
	  minor cleanup
	  
	  Makefile*:
	  - minor cleanup
	  - use automake 2.14b
	  
	  Added files:
	  needed for new automake
	  
	  rest:
	  updated automake files

2001-01-14 15:40  aleidinger

	* misc/abx.c:
	  there are encoders with different delays on different channels
	  (Frank)

2001-01-14 12:21  shibatch

	* libmp3lame/lame.c:
	  Enabled scalefac_scale if --nspsytune is specified.

2001-01-14 12:02  takehiro

	* testcase.mp3:
	  fixed new amp_scalefac_bands
	  (it may use "trigger" variable without any initialization)

2001-01-14 09:42  takehiro

	* libmp3lame/quantize.c:
	  minor bug fix.
	  when subblock gain=1 and scalefactor is -1(means no need to
	  calculate),
	  LAME will try to check the quantization with meanless scalefactor
	  combination.

2001-01-14 09:39  takehiro

	* libmp3lame/quantize.c:
	  subblock_gain is 3-bit and its limit is 7, not 8.

2001-01-14 09:35  takehiro

	* libmp3lame/quantize.c:
	  oops, conflicts are removed

2001-01-14 09:31  takehiro

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  to store the distortion into III_psy_xmin structure, not simple
	  array.
	  this may make array pointer calculation easier and bring some
	  speed up.

2001-01-14 09:26  shibatch

	* libmp3lame/encoder.c, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h:
	  Cleanup and small speed optimization of --nspsytune.

2001-01-14 08:17  markt

	* API:
	  one minor update to the API file

2001-01-14 08:13  markt

	* libmp3lame/lame.c, libmp3lame/quantize.c, testcase.mp3:
	  made a new, simpler unified amp_scalefac_bands():
	  
	  * distort[] = noise/masking
	  * distort[] > 1 ==> noise is not masked
	  * distort[] < 1 ==> noise is masked
	  * max_dist = maximum value of distort[]
	  *
	  * Three algorithms:
	  * noise_shaping_amp
	  * 0 Amplify all bands with distort[]>1.
	  *
	  * 1 Amplify all bands with distort[] >= max_dist^(.5);
	  * ( 50% in the db scale)
	  *
	  * 2 Amplify first band with distort[] >= max_dist;
	  *
	  *
	  * For algorithms 0 and 1, if max_dist < 1, then amplify all bands
	  * with distort[] >= .95*max_dist. This is to make sure we always
	  * amplify at least one band.
	  *
	  
	  noise_shaping_amp=0 -q3 .. -q9
	  noise_shaping_amp=1 -q2
	  noise_shaping_amp=2 -q1, -q0
	  
	  --nspsytune was using what is now noise_shaping_amp=2, but the
	  only
	  way to get this now is to use --nspsytune -q1.
	  Naoki: is this a problem?
	  
	  I also defaulted the clever RH_AMP stoping criterion in
	  outer_loop.

2001-01-14 05:36  markt

	* libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  some tuning of VBR masking_lower because of new ATH function.
	  Disabled VBR_q based lowering of ATH. Is this still needed?
	  You can always achieve the same effect with --athlower option.

2001-01-13 18:01  aleidinger

	* Dll/Makefile.in, Makefile.am, Makefile.in, configure,
	  configure.in, debian/Makefile.am, debian/Makefile.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.am, dshow/Makefile.in, frontend/Makefile.am,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.am, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.am, mpglib/Makefile.in:
	  "make dist" related:
	  - add debian subdir
	  - add missing files
	  
	  different builddir related:
	  - change argument to -l option for nasm

2001-01-13 17:40  aleidinger

	* frontend/Makefile.in.old, libmp3lame/Makefile.in.old:
	  obsolete

2001-01-13 13:45  takehiro

	* libmp3lame/newmdct.c:
	  bit coding hack and cleanup

2001-01-13 12:55  takehiro

	* Dll/Makefile.in, Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  automake generated

2001-01-13 12:54  takehiro

	* libmp3lame/fft.c, libmp3lame/i386/Makefile.am,
	  libmp3lame/i386/fft3dn.nas:
	  now LAME supports 3dnow! fft

2001-01-13 10:37  pgubanov

	* dshow/Encoder.cpp:
	  Removed (char*) cast in lame_encode_buffer() call

2001-01-12 19:20  markt

	* debian/changelog, debian/control, debian/copyright,
	  debian/lame-dev.docs, debian/lame-dev.files,
	  debian/lame-extras.files, debian/lame.docs, debian/lame.files,
	  debian/rules:
	  added debian files from Ingo Saitz

2001-01-12 19:17  markt

	* debian/cron.d.ex, debian/dirs, debian/docs,
	  debian/emacsen-install.ex, debian/emacsen-remove.ex,
	  debian/emacsen-startup.ex, debian/ex.doc-base.package,
	  debian/files, debian/init.d.ex, debian/manpage.1.ex,
	  debian/manpage.sgml.ex, debian/menu.ex,
	  debian/postinst.debhelper, debian/postinst.ex, debian/postrm.ex,
	  debian/preinst.ex, debian/prerm.debhelper, debian/prerm.ex,
	  debian/substvars, debian/watch.ex:
	  New Debian package from Ingo Saitz

2001-01-12 19:11  markt

	* README.Debian:
	  removed README.Debian

2001-01-12 11:13  aleidinger

	* misc/abx.c:
	  further improvements, e.g. reduced latency (Frank)

2001-01-11 22:37  markt

	* HACKING, frontend/mp3rtp.c, frontend/mp3x.c:
	  updated mp3rtp.c and mp3x.c to use lame_global_flags *lame_init()
	  (lame allocates storage for lame_global_flags)
	  rather than lame_init_old(&gf) which required calling
	  program to instantiate gf.
	  
	  Now we need to start writing one routine for every parameter:
	  
	  lame_set_bitrate
	  lame_set_num_channels
	  etc...

2001-01-10 16:03  aleidinger

	* frontend/lame.dsp, frontend/mp3x.dsp, libmp3lame/libmp3lame.dsp,
	  mpglib/mpglib.dsp:
	  work around for a bug in VC (Todd Richmond)

2001-01-09 15:15  aleidinger

	* misc/abx.c:
	  bugfixes

2001-01-08 18:15  aleidinger

	* DEFINES:
	  removed KLEMM_12, on by default

2001-01-08 18:07  aleidinger

	* libmp3lame/vbrquantize.c:
	  Roberts VBR code: speed improvement for vbr-mtrh (ping pong
	  patch, OKed by Mark)

2001-01-07 23:47  markt

	* frontend/portableio.c, frontend/timestatus.c:
	  chage to KLEMM-36 from Frank

2001-01-07 22:53  markt

	* frontend/timestatus.c, libmp3lame/psymodel.c:
	  undefined variable (last = 0, line 214) in timestatis.c
	  commented out.
	  
	  fixed some dimensions in the funtion prototype of l3para_read

2001-01-07 22:35  markt

	* frontend/timestatus.c:
	  Applied Frank's patch (from mp3encoder) to timestatus.c
	  with -R. I hope I didn't undo this by mistake!

2001-01-07 22:33  markt

	* frontend/timestatus.c:
	  added a brhist_jump_back() to timestatus_klemm().
	  sent to me by Magnus Holmgren.

2001-01-07 22:28  markt

	* USAGE, frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/psymodel.c:
	  added the "-ma" option.
	  Ultimate goal:
	  8 kbps Mono
	  16- 96 kbps Intensity Stereo (if available, otherwise Joint
	  Stereo)
	  112-128 kbps Joint Stereo -mj
	  160-192 kbps something between -mj and -ms
	  224-320 kbps Independent Stereo -ms
	  
	  currently enables KLEMM_12. KLEMM_12 is gone.

2001-01-07 16:17  aleidinger

	* libmp3lame/lame.c:
	  Don\'t change floatingpoint mask for FreeBSD on Alpha hardware,
	  obtained from FreeBSD-Ports

2001-01-07 14:48  aleidinger

	* misc/abx.c:
	  cleanup, no functional change

2001-01-07 14:47  aleidinger

	* Dll/Example.dsw, Dll/MP3EncDll.dsw, dshow/dshow.dsw,
	  frontend/lame.dsw, frontend/mp3x.dsw, misc/abx.c:
	  fix *.dsw
	  new version of abc.c (Frank)

2001-01-07 09:14  markt

	* frontend/mp3rtp.c, frontend/mp3x.c:
	  global variables used by frontend code are now instantiated
	  in parse.c.
	  
	  Before, the were defined external in main.h, and instantiated
	  in the frontend, meaning they had to appear in mp3x.c, main.c
	  and mp3rtp.c
	  
	  we need to get rid of these!

2001-01-07 09:11  markt

	* frontend/main.c, frontend/main.h, frontend/parse.c,
	  libmp3lame/lame.c:
	  more tweaks to the --decode-mp3delay option

2001-01-07 08:28  markt

	* libmp3lame/lame.c, libmp3lame/quantize.c, libmp3lame/util.h:
	  added a variable, noise_shaping_amp. It does nothing right now.

2001-01-07 08:24  markt

	* frontend/main.c, frontend/main.h, frontend/parse.c:
	  added a undocumented option --decode-mp3delay <n>
	  to allow someone to adjust the number of samples
	  lame --decode will truncate. David Robinson requirested this
	  since he didn't like that lame --decode would truncate 1106
	  samples
	  (assuming a encoder delay of 576) since BladeEnc has an encoder
	  delay
	  of only 528.)

2001-01-07 05:25  markt

	* frontend/main.c, libmp3lame/util.c:
	  main.c: uses lame_init() instead of lame_init_old().
	  util.c: added frank's changes: 32 point filter instead of 19.
	  prep work for frank's new resampling/filtering code.
	  updated test cases.

2001-01-06 23:12  markt

	* libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.c:
	  ATHformula modifed to work in Hz instead of Khz.
	  
	  Minor change to a bunch of routines + regeneration of
	  test cases.

2001-01-06 22:57  markt

	* testcase.mp3:
	  updated testcase

2001-01-06 22:55  markt

	* libmp3lame/fft.c:
	  Switched to the Blackman window for the FFT energy estimator.
	  This is just to merge some nspsytune code into default

2001-01-06 07:40  gramps

	* libmp3lame/id3tag.c:
	  Backed out some large changes done in November which introduced
	  at least two bugs. Also fixed other older bug in id3tag_write_v2.

2001-01-06 01:11  markt

	* include/lame.h, libmp3lame/lame.c, libmp3lame/psymodel.c:
	  more prep work to remove all psycho acoustic tables

2001-01-06 01:00  markt

	* libmp3lame/VbrTag.h, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/psymodel.c:
	  added formulas to replace the last of the table data.
	  some changes to reflect that Xing VBR header parsing code
	  has been moved into mpglib

2001-01-05 22:15  markt

	* mpglib/interface.c, mpglib/mpglib.h:
	  added XingVBR header parsing code

2001-01-05 22:13  markt

	* frontend/get_audio.c, frontend/gtkanal.c:
	  fixed bug in frame analyzer when analyzing mp3 files.
	  frame counter incremented too soon. This was introduced
	  when someone (maybe me, but I dont remember!) replaced the
	  use of gfp->frameNum with a global variable limited in scope to
	  just gtkanal.c
	  
	  get_audio.c: removed VBR Xing header checking code.
	  this has been moved into mpglib.

2001-01-05 15:26  aleidinger

	* Makefile.in.old, configure.in.old:
	  obsolete

2001-01-05 15:20  aleidinger

	* Makefile.B32, Makefile.unix, frontend/brhist.c,
	  frontend/get_audio.c, frontend/gpkplotting.c, frontend/gtkanal.c,
	  frontend/lame.dsp, frontend/lametime.c, frontend/main.c,
	  frontend/mp3rtp.c, frontend/mp3x.c, frontend/mp3x.dsp,
	  frontend/parse.c, frontend/portableio.c, frontend/rtp.c,
	  frontend/timestatus.c, libmp3lame/VbrTag.c,
	  libmp3lame/bitstream.c, libmp3lame/debugscalefac.c,
	  libmp3lame/encoder.c, libmp3lame/fft.c, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/libmp3lame.dsp,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/resample.c,
	  libmp3lame/reservoir.c, libmp3lame/tables.c,
	  libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c, libmp3lame/version.c,
	  libmp3lame/vorbis_interface.c, misc/abx.c, mpglib/common.c,
	  mpglib/dct64_i386.c, mpglib/decode_i386.c, mpglib/interface.c,
	  mpglib/layer1.c, mpglib/layer2.c, mpglib/layer3.c,
	  mpglib/mpglib.dsp, mpglib/tabinit.c:
	  misc/abx.c: (Frank Klemm)
	  - level adjustment
	  - clean up
	  - A-B repeat ("chunks")
	  
	  Makefile.*:
	  - nuke NOTERMCAP, it's HAVE_TERMCAP now
	  (defined where neccessary, please report slipped in breakage)
	  
	  (frontend|mp3x|libmp3lame|mpglib).dsp:
	  - copy configMS.h config.h (Todd Richmond)
	  - unbreak (Todd Richmond)
	  - nuke NOTERMCAP
	  - change HAVE([A-Z]*) to HAVE_$1
	  
	  *.c:
	  - get rid of configMS.h, we finally have support in *.dsp

2001-01-03 13:06  aleidinger

	* Makefile.DJGPP, Makefile.am, Makefile.in, Makefile.unix,
	  aclocal.m4, config.h.in, configure, configure.in,
	  frontend/Makefile.am, frontend/Makefile.in, frontend/brhist.c,
	  frontend/console.c, frontend/get_audio.c, frontend/gpkplotting.c,
	  frontend/gtkanal.c, frontend/lametime.c, frontend/main.c,
	  frontend/mp3rtp.c, frontend/mp3x.c, frontend/parse.c,
	  frontend/portableio.c, frontend/rtp.c, frontend/timestatus.c,
	  libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/debugscalefac.c, libmp3lame/encoder.c,
	  libmp3lame/fft.c, libmp3lame/id3tag.c, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/resample.c,
	  libmp3lame/reservoir.c, libmp3lame/tables.c,
	  libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c, libmp3lame/version.c,
	  libmp3lame/vorbis_interface.c, misc/Makefile.am,
	  misc/Makefile.in, mpglib/common.c, mpglib/dct64_i386.c,
	  mpglib/decode_i386.c, mpglib/interface.c, mpglib/layer1.c,
	  mpglib/layer2.c, mpglib/layer3.c, mpglib/tabinit.c:
	  Makefile.{unix,DJGPP}: Bugfixes (Includes)
	  Makefile.am: follow the move of some scripts to misc
	  *: add dmalloc support (configure --with-dmalloc)
	  (it seems we have a memleak somewhere, I will have a closer look
	  soon)

2001-01-03 11:51  shibatch

	* libmp3lame/encoder.c, libmp3lame/psymodel.c:
	  Tweaking and bugfix of --nspsytune.

2001-01-02 06:02  markt

	* frontend/get_audio.c, libmp3lame/encoder.c:
	  typo in read_samples_pcm fixed

2001-01-01 17:35  markt

	* API, HACKING:
	  minor edits to API and HACKING

2001-01-01 17:26  markt

	* Lame.vbs, README.WINGTK, TODO, USAGE, lame4dos.bat,
	  misc/Lame.vbs, misc/lame4dos.bat:
	  moved some more scripts into 'misc' directory

2001-01-01 17:12  markt

	* INSTALL, README.WINGTK, USAGE, auenc, frontend/README.Win32,
	  misc/auenc, misc/mlame, mlame:
	  moved scripts into misc directory
	  moved README.Win32 (which described compiling mp3x under windows)
	  back to its original name, README.WINGTK
	  minor updates to INSTALL and USAGE

2000-12-27 13:13  aleidinger

	* misc/abx.c:
	  - listen to the difference of both signals
	  - highlighting of "errors"
	  - DC canceling
	  - support for monaural signals
	  - cleanup
	  - bugfixes
	  - correct shell quoting

2000-12-25 10:49  shibatch

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.h:
	  Tweaked --nspsytune.

2000-12-22 15:36  aleidinger

	* misc/abx.c:
	  now with cross correlation (Frank)

2000-12-20 22:55  aleidinger

	* misc/abx.c:
	  - files to test have to have the same samplerate (Frank)
	  - check for correct number of arguments (Alexander)

2000-12-20 15:14  aleidinger

	* DEFINES, libmp3lame/psymodel.c, libmp3lame/tables.c,
	  libmp3lame/tables.h:
	  Patches for KLEMM_43 protected code

2000-12-19 13:09  aleidinger

	* DEFINES, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/reservoir.c, libmp3lame/tables.c,
	  libmp3lame/vbrquantize.c:
	  Ping pong #3, OKed by Mark.

2000-12-19 08:16  markt

	* LICENSE:
	  simplifed the LICENSE file describing how to use
	  LAME in commercial software

2000-12-19 07:51  markt

	* INSTALL, README, README.B32:
	  updated README files with new configure information.
	  debuging statement added to interface.c, then removed.

2000-12-19 07:46  markt

	* frontend/brhist.c, frontend/get_audio.c:
	  more size_t's removed

2000-12-19 07:43  markt

	* frontend/main.c, frontend/mp3rtp.c, frontend/parse.c,
	  frontend/rtp.c, frontend/rtp.h:
	  removed some size_t's.
	  Added some windows code to rtp.c/mp3rtp.c: It is supposed to
	  be much improved:
	  
	  
	  LAME is really great project.
	  
	  However, RTP code for multicasting isn't very good.
	  
	  I've attached some modification, and tested it on Win NT4 and
	  Win2K with
	  FreeAmp on ordinary 10 and 100 Mbit Ethernet with encoding in
	  neverending
	  loop. It was running one week w/o any probs.
	  
	  I would like to please you to not include my name anywhere (there
	  are
	  several reason why, at the moment).
	  
	  However, code is very MSVC specific, so it is commented out
	  with #if 0. Most of the code was added to the end of rtp.c.

2000-12-19 07:00  markt

	* frontend/main.c:
	  deleted stupid C++ typecasts. LAME is written in C.
	  integer arguments to fwrite() do not need typecasts.

2000-12-16 05:43  shibatch

	* libmp3lame/lame.c, libmp3lame/psymodel.c:
	  1. Old ATH formula is now default for --nspsytune.
	  2. Tweaked a parameter of --nspsytune.
	  3. Minor code cleanup.

2000-12-14 13:32  aleidinger

	* Dll/Makefile.in, Makefile.in, configure, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  regen

2000-12-14 13:31  aleidinger

	* configure.in, libmp3lame/i386/Makefile.am:
	  force nasm to generate coff output when building under cygwin
	  (Egor Duga)

2000-12-14 10:05  aleidinger

	* Dll/Makefile.in, Makefile.in, config.h.in, configure,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  dshow/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  regen

2000-12-14 10:04  aleidinger

	* Makefile.DJGPP, Makefile.am, Makefile.unix, configMS.h,
	  configure.in, frontend/Makefile.am, libmp3lame/Makefile.am,
	  libmp3lame/i386/Makefile.am, mpglib/Makefile.am:
	  configure.in:
	  - determine the size of various types at configure time
	  - add --enable-debug=anoying
	  
	  configMS.h:
	  - add SIZEOF_... defines, see configure.in
	  
	  Makefile.unix:
	  - bugfix from Robert (shame on me)
	  
	  Makefile.DJGPP:
	  - add cpu_feat.nas and scalar.nas to the build with NASM
	  
	  Submitted by Egor Duda <deo@logos-m.ru> (thank you, works great):
	  * Makefile.am: add $(EXEEXT) to executables for cygwin build
	  * Makefile.am: take source .wav file from srcdir
	  * configure.in: require autoconf 2.13
	  * configure.in: check for executable files extension
	  * frontend/Makefile.am: use builddir to get autogenerated headers
	  and libraries from, in case we're doing build in separate
	  directory;
	  take normal headers from srcdir instead of current.
	  * libmp3lame/Makefile.am: ditto
	  * libmp3lame/i386/Makefile.am: ditto
	  * mpglib/Makefile.am: ditto

2000-12-13 02:28  shibatch

	* libmp3lame/lame.c:
	  Minor bugfix.

2000-12-13 02:21  shibatch

	* libmp3lame/psymodel.c:
	  Minor bugfix of --nspsytune.
	  (forgot to merge.)

2000-12-12 04:50  shibatch

	* frontend/parse.c, include/lame.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.c, libmp3lame/util.h:
	  Added option --athtype which selects ATH formula.
	  
	  Some major changes to --nspsytune.
	  1. --nspsytune doesn't use pe.
	  2. MS psymodel is modified. Added new option --nssafejoint.
	  3. some more tunings of short block preecho control.

2000-12-12 01:11  markt

	* libmp3lame/lame.c, libmp3lame/util.c:
	  Latest checkin from Alex include an rewritten 'Bitrate()'
	  routine.
	  
	  This breaks the --abr code, since new routine considers a bitrate
	  of 0
	  invalid. Now fixed.

2000-12-11 22:15  markt

	* misc/Makefile.am:
	  added mlame_corr.c to the
	  EXTRA_DIST line of Makefile.am

2000-12-11 16:03  aleidinger

	* DEFINES, Makefile.in, frontend/main.c, include/lame.h,
	  libmp3lame/encoder.c, libmp3lame/encoder.h, libmp3lame/fft.c,
	  libmp3lame/fft.h, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/newmdct.h, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.c, libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/vorbis_interface.c, mpglib/interface.c,
	  mpglib/interface.h:
	  tries to include every change before the CVS rollback which
	  didn't changes output

2000-12-11 14:48  aleidinger

	* frontend/gtkanal.c, libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/newmdct.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/reservoir.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c, libmp3lame/vorbis_interface.c:
	  redo gfc->stereo -> gfc->channels_out conversation

2000-12-11 01:32  markt

	* USAGE:
	  updated info regarding the MAD decoder

2000-12-11 00:34  markt

	* Makefile.am, libmp3lame/lame.c:
	  applied patch for cygwin FPE stuff

2000-12-11 00:01  markt

	* USAGE, configure, configure.in, frontend/get_audio.c,
	  frontend/get_audio.h, include/lame.h,
	  libmp3lame/mpglib_interface.c, misc/mlame_corr.c, mlame_corr.c:
	  configure: removed all the debug warning flags. it was impossible
	  to find real problems when compiling in debug mode.
	  
	  moved mlame_corr.c into 'misc' directory
	  
	  removed some 'size_t' and 'unsigned' from get_audio.c
	  
	  made mp3buffer 'unsigned char*' for the mpglib interface
	  routines.

2000-12-08 13:47  aleidinger

	* auenc, misc/mugeco.sh:
	  more portability fixes from Fredrik Mellstrom

2000-12-07 18:58  aleidinger

	* mlame:
	  use sh instead of bash

2000-12-07 18:55  aleidinger

	* auenc:
	  portability patch from Fredrik Mellstrom

2000-12-05 15:48  aleidinger

	* DEFINES:
	  update

2000-12-05 15:46  aleidinger

	* Makefile.MSVC:
	  add scalar.nas

2000-12-05 15:40  aleidinger

	* Dll/Makefile.in, configure, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, mpglib/Makefile.in:
	  regen

2000-12-05 15:37  aleidinger

	* Makefile.B32, Makefile.MSVC, Makefile.in, config.h.in,
	  configMS.h, configure.in, frontend/Makefile.am,
	  frontend/get_audio.c, frontend/get_audio.h,
	  frontend/portableio.c, libmp3lame/Makefile.am,
	  libmp3lame/i386/Makefile.in, libmp3lame/i386/cpu_feat.nas,
	  misc/ath.c, mpglib/Makefile.am:
	  play ping-pong with some files (we're now at 2000-12-03 again),
	  ok'ed by Mark

2000-12-05 01:17  markt

	* DEFINES:
	  restored DEFINES.
	  
	  Sorry Alexendar - I tried to only revert the *.c and *.h files
	  to their Nov 21 values (because of test case problems) but
	  I was unable to get this to work. I was lazy and reverted
	  entire repository back to Nov 21.

2000-12-05 01:15  markt

	* DEFINES:
	  trying to restore DEFINES

2000-12-04 14:36  aleidinger

	* Makefile.unix:
	  add scalar.nas to build

2000-12-04 06:31  pgubanov

	* dshow/PropPage.cpp, dshow/PropPage.h:
	  Target bitrate fix

2000-12-04 06:30  pgubanov

	* dshow/Mpegac.cpp, dshow/Mpegac.def, dshow/Mpegac.h,
	  dshow/PropPage_adv.cpp, dshow/PropPage_adv.h, dshow/Property.rc,
	  dshow/README, dshow/REG.CPP, dshow/REG.H, dshow/UIDS.H,
	  dshow/dshow.dsp, dshow/iaudioprops.h, dshow/resource.h:
	  CR/LF fix

2000-12-03 23:27  markt

	* libmp3lame/util.c, testcase.mp3:
	  KLEMM_01 defaulted.
	  
	  This enables the new ATH formula.

2000-12-03 23:03  markt

	* libmp3lame/tables.c, libmp3lame/tables.h:
	  huffmantab: int changed back to short.

2000-12-03 23:00  markt

	* DEFINES, Dll/Makefile.in, Makefile.B32, Makefile.MSVC,
	  Makefile.in, config.h.in, configMS.h, configure, configure.in,
	  doc/Makefile.in, doc/html/Makefile.in,
	  doc/html/contributors.html, doc/html/examples.html,
	  doc/html/history.html, doc/html/id3.html, doc/html/index.html,
	  doc/html/lame.css, doc/html/modes.html, doc/html/node6.html,
	  doc/html/switchs.html, doc/man/Makefile.in, frontend/Makefile.am,
	  frontend/Makefile.in, frontend/get_audio.c, frontend/get_audio.h,
	  frontend/gtkanal.c, frontend/main.c, frontend/parse.c,
	  frontend/portableio.c, include/Makefile.in, include/lame.h,
	  libmp3lame/Makefile.am, libmp3lame/Makefile.in,
	  libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/encoder.c, libmp3lame/encoder.h, libmp3lame/fft.c,
	  libmp3lame/fft.h, libmp3lame/i386/Makefile.am,
	  libmp3lame/i386/Makefile.in, libmp3lame/i386/cpu_feat.nas,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/newmdct.c, libmp3lame/newmdct.h, libmp3lame/pcm.c,
	  libmp3lame/pcm.h, libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/tables.c, libmp3lame/tables.h,
	  libmp3lame/util.c, libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/version.h, libmp3lame/vorbis_interface.c,
	  mpglib/Makefile.am, mpglib/Makefile.in, mpglib/interface.c,
	  mpglib/interface.h:
	  Reverted back to Nov 21 version.
	  
	  Added HTML doc from GB

2000-12-03 17:38  aleidinger

	* configure:
	  regen

2000-12-03 17:37  aleidinger

	* configure.in:
	  fix in debugging options

2000-12-03 14:16  aleidinger

	* DEFINES:
	  add KLEMM_36

2000-12-03 14:14  aleidinger

	* frontend/portableio.c:
	  Portability fix from Frank, you need to define KLEMM_36 to enable
	  it
	  (no functional change if you leave it disabled):
	  ---snip---
	  It is a hoax to call this code portable-IO:
	  
	  - It doesn't work on machines with CHAR_BIT != 8
	  - it also don't test this error condition
	  - otherwise it tries to handle CHAR_BIT != 8 by things like
	  masking 'putc(i&0xff,fp)'
	  - It doesn't handle EOF in any way
	  - it only works with ints with 32 or more bits
	  - It is a collection of initial buggy code with patching the
	  known errors
	  instead of CORRECTING them!
	  For that see comments on the old Read16BitsHighLow()
	  ---snip---

2000-12-03 13:14  aleidinger

	* misc/abx.c:
	  Improvements from Frank:
	  - more input formats (detected by filename)
	  Improvements by me:
	  - sanity checks
	  - works on FreeBSD

2000-12-01 17:42  aleidinger

	* libmp3lame/psymodel.c, libmp3lame/tables.c, libmp3lame/tables.h:
	  Patches from Frank:
	  ---snip---
	  defined 4 new types for holding psydata information
	  psydata is divided into 25 normal tables (the 24 tables + 1
	  master table)
	  access via "normal" index instead of scanning through the data
	  tables are now sparsable, i.e. it is possible to remove unneeded
	  parts
	  but this is not yet done to avoid to do to much things at once
	  tables are now smaller
	  ---snip---
	  Hi also did some tests to verify this didn't changes the output.
	  
	  I also verified it with "make test" and some regular encoding.

2000-11-30 10:23  aleidinger

	* DEFINES:
	  add description of some compile time defines

2000-11-30 10:20  aleidinger

	* Makefile.B32:
	  sync with recent changes in the other Makefiles:
	  - HAVExxx -> HAVE_xxx
	  - HAVE_CONFIG_MS_H

2000-11-30 10:17  aleidinger

	* frontend/get_audio.c, frontend/get_audio.h:
	  - remove check for CHAR_BITS, was a typo from Frank (confirmed by
	  Frank)
	  - add optimization for unknown SIZEOF_UNSIGNED_LONG (submitted by
	  Frank)
	  in optimized version of SwapBytesInWords()
	  - make the optimized version of SwapBytesInWords() the default
	  (OKed by Mark)
	  - remove DetermineByteOrder(), we have compile time support for
	  it

2000-11-30 10:16  aleidinger

	* Dll/Makefile.in, Makefile.in, configure, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  mpglib/Makefile.in:
	  regen

2000-11-30 10:11  aleidinger

	* libmp3lame/pcm.c:
	  - some modifications from Frank

2000-11-30 10:08  aleidinger

	* configure.in:
	  - compile time check for sizeof(int)
	  - more compiler warnings for debugging

2000-11-30 10:07  aleidinger

	* config.h.in, configMS.h:
	  - compile time value for sizeof(int)

2000-11-30 10:07  aleidinger

	* frontend/Makefile.am, libmp3lame/Makefile.am, mpglib/Makefile.am:
	  - additional options for lclint

2000-11-30 10:06  aleidinger

	* libmp3lame/i386/Makefile.am:
	  - *.lst instead of specifying every file in "CLEANFILES"

2000-11-27 16:39  aleidinger

	* libmp3lame/pcm.c, libmp3lame/pcm.h:
	  submitted by Frank, first revision of a new lame API, not
	  connected to the build yet

2000-11-26 20:58  robert

	* libmp3lame/i386/cpu_feat.nas, libmp3lame/lame.c,
	  libmp3lame/util.c, libmp3lame/util.h:
	  Part2 of Frank's Patch
	  
	  His List of Changes (part 1 + 2):
	  
	  - changed the type of the MP3 byte stream from 'char' to
	  'unsigned char'
	  - added some 'const' attributes
	  - added to lame_internal_flags: gfp, frame_size, coding,
	  frame_count
	  - one code move inside lame.c (mode_ext), one remove
	  (channels_out)
	  - two new types in util.c
	  - some bug fixes with Ogg
	  - SIMD2 detection

2000-11-26 10:50  robert

	* frontend/get_audio.c, frontend/main.c, include/lame.h,
	  libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/encoder.c, libmp3lame/encoder.h, libmp3lame/fft.c,
	  libmp3lame/fft.h, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/newmdct.h, libmp3lame/psymodel.c,
	  libmp3lame/psymodel.h, libmp3lame/util.h,
	  libmp3lame/vorbis_interface.c, mpglib/interface.c,
	  mpglib/interface.h:
	  Patch by Frank:
	  changes from char to unsigned char

2000-11-25 22:59  robert

	* libmp3lame/quantize.c:
	  bug fix for -X6, could result in clicking noises

2000-11-25 20:40  robert

	* libmp3lame/quantize.c:
	  typo "powf()" lend to crashes under windows, fixed

2000-11-25 18:56  robert

	* libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/vbrquantize.c:
	  small fix for vbr-mtrh mode

2000-11-25 18:18  robert

	* libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/util.h, libmp3lame/vbrquantize.c:
	  removed superfluous memset in vbr_noise_shaping2()
	  some veryminor speed up for cbr
	  amp_scalefac cleanup

2000-11-24 17:28  robert

	* frontend/gtkanal.c:
	  Patches by Frank:
	  
	  - documentation
	  - introduced gfc->channels_in
	  - renamed gfc->stereo into gfc->channels_out
	  (note by me: stereo would imply a boolean type stereo ? yes/no
	  what's really annoying and confusing is that we didn't used it
	  that way,
	  but instead stereo=1 -> mono/1-channel, stereo=2 ->
	  stereo/2-channels)
	  - minor bug fixes for BitrateIndex search

2000-11-24 17:06  robert

	* include/lame.h, libmp3lame/bitstream.c, libmp3lame/encoder.c,
	  libmp3lame/lame.c, libmp3lame/newmdct.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/reservoir.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  Patches by Frank:
	  
	  - documentation
	  - introduced gfc->channels_in
	  - renamed gfc->stereo into gfc->channels_out
	  (note by me: stereo would imply a boolean type stereo ? yes/no
	  what's really annoying and confusing is that we didn't used it
	  that way,
	  but instead stereo=1 -> mono/1-channel, stereo=2 ->
	  stereo/2-channels)
	  - minor bug fixes for BitrateIndex search

2000-11-23 23:19  robert

	* libmp3lame/vbrquantize.c:
	  some loops manually unrolled into "Duff's Device"
	  speed improvement for --vbr-mtrh GCC compiled ca. 5 % overall
	  speed improvement for --vbr-mtrh Intel 4.5 compiled ca. 1-2 %
	  looks like the intel compiler does a better job on loops
	  compared to GCC.

2000-11-23 15:14  robert

	* Makefile.MSVC:
	  compromise between size/speed/multi CPU, thanks to Dmitry

2000-11-23 13:06  aleidinger

	* misc/ath.c:
	  readd support for other systems than linux (at least FreeBSD)

2000-11-22 21:03  robert

	* Makefile.MSVC:
	  added little howto into Makefile.MSVC
	  think MS C compiler works now again

2000-11-22 01:09  robert

	* frontend/parse.c:
	  cosmetics

2000-11-22 00:30  robert

	* libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/tables.c,
	  libmp3lame/tables.h, libmp3lame/util.c, libmp3lame/version.h,
	  misc/ath.c:
	  some patches by Frank
	  
	  - replace ath.c (has with Lame nothing to do)
	  - lame.c: A line in my new code exchange (see ATHformula
	  problems)
	  - util.c: ATHformula changed over from kHz to cycles per second
	  (last util formula with kHz been)
	  - ATHformula below 20 cycles per second (instead of 10 cycles per
	  second) limits (< 20 cycles per second are Infraschall,
	  additionally the function becomes there completely useless)
	  - quantize_pvt.c: Effects of ATHformula
	  - tables.c:
	  - something documents -
	  -number of entries over one increases
	  (20=>21, 11=>12, etc..)
	  - SNR_s converted and to the obvious values
	  0.15/0.18/0.20/0.25/0.30/0.35/0.40 rounded.
	  - w1 and w2 with 576 multiply (error is <
	  0.91*10^-3, it seems in the tables both to have been rounded off
	  as well as with
	  minimum error have been rounded) - end to OF DATA marker to the
	  integrity test
	  inserted - a reading in of the table revises: psymodel.c - fewer
	  error-prone code (fewer
	  * p++) - meaningful error messages for problem localization -
	  Peeeeeeps in the case of
	  error - removing from loop (now j, is used only farther back; and
	  adapted by k2) -
	  ATHformula - util.c - some amusing " while" Konstrukte, which
	  originate still from the
	  ISO code, transformed into the appropriate " for" Konstrukte
	  (with it first some amusing
	  things are noticeable, so e.g. why starting from index 0 instead
	  of 1 are looked up?)

2000-11-21 22:15  robert

	* Makefile.MSVC, frontend/gtkanal.c:
	  some reorganization in Makefile.MSVC
	  it's now more commandline configurable

2000-11-20 20:45  robert

	* frontend/mp3x.c, frontend/parse.c:
	  fix for mp3x.c: showed usage twice
	  fix for parse.c: lame_version_print()
	  in some situations the text is longer than 80 characters
	  now the url will be wrapped into the next line, right aligned
	  if 80 character aren't enough

2000-11-20 15:08  robert

	* Makefile.MSVC, Makefile.unix, frontend/parse.c,
	  libmp3lame/lame.c, libmp3lame/quantize.c:
	  overhauled Makefile.MSVC for easier configuration
	  design bug in parse.c fixed (should not include files from
	  engine)

2000-11-20 01:48  robert

	* Dll/BladeMP3EncDLL.c, Makefile.MSVC:
	  it's now possible to compile lame_enc.dll with Makefile.MSVC
	  tried the DLL with CD-ex, seems to work

2000-11-20 00:16  robert

	* Makefile.MSVC:
	  updated Compile Flags due to crashes
	  added rule to generate config.h out of configMS.h
	  first try to let the dll compile, still does not work

2000-11-19 14:12  robert

	* Makefile.MSVC, Makefile.unix, frontend/gtkanal.c,
	  libmp3lame/lame.c:
	  Some bugs fixed introduced due to configure stuff and others

2000-11-19 11:26  aleidinger

	* Makefile.MSVC:
	  bugfix for decoder define

2000-11-18 12:13  aleidinger

	* frontend/Makefile.in:
	  regen (unbreak build)

2000-11-18 12:12  aleidinger

	* frontend/Makefile.am:
	  bugfix after removing ieeefloat.*

2000-11-18 10:52  aleidinger

	* frontend/brhist.c, frontend/get_audio.c, frontend/gpkplotting.c,
	  frontend/gtkanal.c, frontend/lame.dsp, frontend/lametime.c,
	  frontend/main.c, frontend/mp3rtp.c, frontend/mp3x.c,
	  frontend/parse.c, frontend/portableio.c, frontend/rtp.c,
	  frontend/timestatus.c, libmp3lame/VbrTag.c,
	  libmp3lame/bitstream.c, libmp3lame/debugscalefac.c,
	  libmp3lame/encoder.c, libmp3lame/fft.c, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/libmp3lame.dsp,
	  libmp3lame/mpglib_interface.c, libmp3lame/newmdct.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/reservoir.c,
	  libmp3lame/tables.c, libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c, libmp3lame/version.c,
	  libmp3lame/vorbis_interface.c, misc/abx.c, misc/ath.c,
	  mpglib/common.c, mpglib/dct64_i386.c, mpglib/decode_i386.c,
	  mpglib/interface.c, mpglib/layer1.c, mpglib/layer2.c,
	  mpglib/layer3.c, mpglib/mpglib.dsp, mpglib/tabinit.c:
	  include configMS.h on Windows

2000-11-18 10:49  aleidinger

	* Makefile.DJGPP, Makefile.MSVC:
	  define HAVE_CONFIG_MS_H for Windows

2000-11-18 10:49  aleidinger

	* Makefile.in:
	  regen (configMS.h)

2000-11-18 10:48  aleidinger

	* Makefile.am, configMS.h:
	  introduce configMS.h for Windows

2000-11-18 05:40  markt

	* README.Debian, debian, debian/changelog, debian/control,
	  debian/copyright, debian/cron.d.ex, debian/dirs, debian/docs,
	  debian/emacsen-install.ex, debian/emacsen-remove.ex,
	  debian/emacsen-startup.ex, debian/ex.doc-base.package,
	  debian/files, debian/init.d.ex, debian/manpage.1.ex,
	  debian/manpage.sgml.ex, debian/menu.ex,
	  debian/postinst.debhelper, debian/postinst.ex, debian/postrm.ex,
	  debian/preinst.ex, debian/prerm.debhelper, debian/prerm.ex,
	  debian/rules, debian/substvars, debian/watch.ex,
	  frontend/brhist.c, frontend/get_audio.c, frontend/main.c:
	  added debian stuff. Make one of the KLEMM's the default

2000-11-18 04:30  markt

	* Dll/Makefile.in, Makefile.B32, Makefile.DJGPP, Makefile.MSVC,
	  Makefile.in, Makefile.unix, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, dshow/Makefile.in,
	  frontend/get_audio.c, frontend/portableio.c, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  mpglib/Makefile.in:
	  Makefiles: removed ieeefloat.* from makefiles.
	  get_audio.c: replaced DetermineByteOrder with original routine.

2000-11-18 04:24  markt

	* frontend/Makefile.am, frontend/Makefile.in, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/ieeefloat.c, frontend/ieeefloat.h,
	  frontend/portableio.c, frontend/portableio.h:
	  Removed ieeefloat.*

2000-11-15 18:52  aleidinger

	* frontend/get_audio.c:
	  only allow fast SwapBytesInWords if SIZEOF_UNSIGNED_LONG is
	  defined

2000-11-14 12:25  aleidinger

	* libmp3lame/resample.h:
	  remove define of INLINE

2000-11-14 12:24  aleidinger

	* frontend/ieeefloat.c:
	  let configure handle little/big endianess

2000-11-14 12:20  aleidinger

	* libmp3lame/bitstream.c, libmp3lame/fft.c, libmp3lame/newmdct.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/takehiro.c, libmp3lame/util.c:
	  INLINE -> inline

2000-11-14 12:19  aleidinger

	* libmp3lame/machine.h:
	  As discussed on the mailinglist:
	  - correct when to define INLINE/inline
	  - INLINE -> inline, configure already does something similar

2000-11-14 12:18  aleidinger

	* libmp3lame/encoder.c:
	  compile time resolvable assertions switched to compile time
	  errors

2000-11-14 12:18  aleidinger

	* libmp3lame/lame.c:
	  - compile time resolvable assertions switched to compile time
	  errors
	  - adjust version string

2000-11-14 12:17  aleidinger

	* frontend/parse.c:
	  - adjust version string
	  - sort system includes

2000-11-14 12:16  aleidinger

	* include/lame.h:
	  - quiet 4 lclint "errors"
	  - disable get_lame_about()

2000-11-14 12:15  aleidinger

	* libmp3lame/version.c:
	  - switched from run time generated version strings to compile
	  time
	  generated ones
	  - implement void get_lame_version_numerical(lame_version_t
	  *const)
	  (#ifdef'ed out)
	  - disable get_lame_about(), snprintf isn't available in Win32

2000-11-13 14:11  aleidinger

	* dshow/Makefile.in:
	  regen (add dshow.dsw)

2000-11-13 14:10  aleidinger

	* dshow/Makefile.am:
	  add dshow.dsw

2000-11-13 14:10  aleidinger

	* misc/mugeco.sh:
	  shell math instead of awk math (no leading zeros needed)

2000-11-13 13:42  pgubanov

	* include/lame.h:
	  Explicit CDECL for API

2000-11-13 13:31  pgubanov

	* dshow/Encoder.cpp, dshow/Encoder.h, dshow/dshow.dsp:
	  Synchronized with current LAME code

2000-11-13 10:59  aleidinger

	* frontend/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in:
	  regen

2000-11-13 10:59  aleidinger

	* frontend/Makefile.am, libmp3lame/Makefile.am:
	  - support for CONFIG_DEFS
	  - possibility to exclude code to lint

2000-11-13 10:57  aleidinger

	* libmp3lame/i386/Makefile.am:
	  support for CONFIG_DEFS

2000-11-13 10:53  aleidinger

	* frontend/ieeefloat.c:
	  remove CVS log tag/information, we have "cvs log" for this

2000-11-13 10:50  aleidinger

	* frontend/get_audio.c:
	  - sort includes
	  - compile time resolvable assertions switched to compile time
	  errors
	  - we didn't need to determine the size of "unsigned long" at
	  runtime
	  (KLEMM_10 case), we're able to do it at compile time (with the
	  help
	  of configure)

2000-11-13 10:48  aleidinger

	* libmp3lame/vbrquantize.c:
	  - move define of TAKEHIRO_IEEE754_HACK into configure.in
	  - commented out code -> ifdef'ed out code
	  (I'm on a anti-warning rampage)

2000-11-13 10:47  aleidinger

	* libmp3lame/quantize_pvt.c:
	  move define of TAKEHIRO_IEEE754_HACK into configure.in

2000-11-13 10:45  aleidinger

	* misc/abx.c, misc/ath.c:
	  - include config.h
	  - include correct soundcard.h (with support from configure)

2000-11-13 10:44  aleidinger

	* Makefile.in, config.h.in, configure:
	  regen

2000-11-13 10:41  aleidinger

	* configure.in:
	  - handle TAKEHIRO_IEEE754_HACK
	  - support for 'CONFIG_DEFS="-Dxxx" ./configure'
	  - handle endianess
	  - check for soundcard.h

2000-11-13 10:39  aleidinger

	* mpglib/Makefile.in:
	  regen (add possibility to exclude code for linting)

2000-11-13 10:38  aleidinger

	* mpglib/Makefile.am:
	  add possibility to exclude code for linting

2000-11-13 10:35  aleidinger

	* misc/mugeco.sh:
	  add start and stop date/time

2000-11-13 09:30  pgubanov

	* dshow/dshow.dsw:
	  Initial release

2000-11-12 22:18  aleidinger

	* misc/mugeco.sh:
	  add id

2000-11-12 22:17  aleidinger

	* misc/mugeco.sh:
	  bugfix

2000-11-12 20:45  aleidinger

	* misc/Makefile.in:
	  regen (add mugeco.sh)

2000-11-12 20:45  aleidinger

	* misc/Makefile.am:
	  add mugeco.sh

2000-11-12 20:43  aleidinger

	* misc/mugeco.sh:
	  first version of the multi generation encoder script

2000-11-12 10:57  aleidinger

	* Makefile.in, configure:
	  regen (readd dshow)

2000-11-12 10:56  aleidinger

	* Makefile.am, configure.in:
	  readd dshow

2000-11-11 23:20  pfk

	* Makefile.DJGPP, Makefile.am, Makefile.in, Makefile.in.old,
	  Makefile.unix, USAGE, configure, configure.in,
	  frontend/get_audio.c, frontend/get_audio.h, frontend/main.c,
	  frontend/parse.c, libmp3lame/bitstream.c, libmp3lame/encoder.h,
	  libmp3lame/i386/scalar.nas, libmp3lame/lame.c,
	  libmp3lame/resample.c, libmp3lame/resample.h,
	  libmp3lame/reservoir.c, libmp3lame/tables.c, libmp3lame/tables.h,
	  misc/abx.c:
	  A lot of very small changes. Most important is prparing of a more
	  sofisticated method to determine the bandwidth and the MP3
	  sampling
	  frequency.

2000-11-11 17:29  aleidinger

	* Makefile.in, configure:
	  regen (add dshow)

2000-11-11 17:28  aleidinger

	* Makefile.am, configure.in:
	  add dshow

2000-11-11 17:26  aleidinger

	* dshow/Makefile.am, dshow/Makefile.in:
	  initial version

2000-11-11 05:56  markt

	* libmp3lame/encoder.c, libmp3lame/machine.h, libmp3lame/util.c,
	  testcase.mp3:
	  updated testcases after changing some default settings

2000-11-11 05:47  markt

	* libmp3lame/encoder.c:
	  make RH_VALIDATE_MS the default

2000-11-11 04:40  markt

	* frontend/parse.c, libmp3lame/lame.c, libmp3lame/util.c:
	  undid some of Frank's changes to the resample code.

2000-11-11 01:55  markt

	* dshow/Encoder.cpp:
	  *** empty log message ***

2000-11-11 01:53  markt

	* dshow, dshow/Encoder.cpp, dshow/Encoder.h, dshow/Mpegac.cpp,
	  dshow/Mpegac.def, dshow/Mpegac.h, dshow/PropPage.cpp,
	  dshow/PropPage.h, dshow/PropPage_adv.cpp, dshow/PropPage_adv.h,
	  dshow/Property.rc, dshow/README, dshow/REG.CPP, dshow/REG.H,
	  dshow/UIDS.H, dshow/dshow.dsp, dshow/iaudioprops.h,
	  dshow/resource.h:
	  directshow filter for LAME from Peter Gubanov

2000-11-10 23:18  pfk

	* frontend/get_audio.c, libmp3lame/i386/scalar.nas,
	  libmp3lame/lame.c, libmp3lame/machine.h, libmp3lame/resample.c,
	  libmp3lame/resample.h, libmp3lame/util.c, libmp3lame/util.h,
	  misc/scalartest.c:
	  Resample continued.
	  
	  The old code now uses a 31 Tap FIR for HQ coding and a 7 Tap FIR
	  for MQ and
	  LQ coding.

2000-11-10 14:42  aleidinger

	* frontend, libmp3lame, mpglib:
	  add lclint.txt

2000-11-10 14:41  aleidinger

	* frontend/Makefile.in, libmp3lame/Makefile.in, mpglib/Makefile.in:
	  regen (lclint target)

2000-11-10 14:40  aleidinger

	* config.h.in:
	  regen (HAVE_EFENCE)

2000-11-10 14:39  aleidinger

	* configure:
	  regen (disalbe-shared/HAVE_EFENCE)

2000-11-10 14:36  aleidinger

	* configure.in:
	  - new default: no shared lib
	  - define HAVE_EFENCE if apropriate

2000-11-10 14:33  aleidinger

	* mpglib/Makefile.am:
	  add lclint target (564 code "errors" found so far)

2000-11-10 14:33  aleidinger

	* frontend/Makefile.am:
	  add lclint target (471 code "errors" found so far)

2000-11-10 14:32  aleidinger

	* libmp3lame/Makefile.am:
	  add lclint target (1505 code "errors" found so far)

2000-11-10 14:31  aleidinger

	* libmp3lame/quantize.c:
	  fix RH_AMP (gfc = gfp->internal_flags)

2000-11-09 22:01  pfk

	* include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/i386/scalar.nas, libmp3lame/resample.c,
	  libmp3lame/resample.h, misc/ath.c, misc/scalartest.c:
	  Start of resampling.

2000-11-09 00:29  pfk

	* libmp3lame/bitstream.c, libmp3lame/id3tag.c,
	  libmp3lame/machine.h, libmp3lame/version.h, misc/ath.c,
	  testcase.mp3:
	  Small changes.
	  
	  - removed an old function strcasecmp (replaced by fuzzy_strcmp)
	  - optimized region 2 bitcoder
	  - remark about version.h (stringification)
	  - INLINE = inline + static

2000-11-08 19:59  aleidinger

	* libmp3lame/resample.c, mpglib/layer1.c:
	  include assert.h, seems to be forgotten in last commit (how could
	  this
	  compile?)

2000-11-07 23:06  pfk

	* frontend/get_audio.c, frontend/parse.c, frontend/timestatus.c,
	  include/lame.h, libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/id3tag.c, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/version.c,
	  libmp3lame/version.h, misc/ath.c, mpglib/layer1.c:
	  Changed back to old report interface until we've discussed all
	  possible
	  versions and make a decision.
	  
	  Some bugfixes. Some remarks.

2000-11-07 15:36  aleidinger

	* configure:
	  regen after:
	  - HAVE{MPGLIB,VORBIS,GTK} -> HAVE_$1
	  - --disable-analyzer-hooks
	  - --enable-efence

2000-11-07 15:36  aleidinger

	* config.h.in, frontend/get_audio.c, frontend/main.c,
	  frontend/parse.c, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/vorbis_interface.c:
	  HAVE{MPGLIB,VORBIS,GTK} -> HAVE_$1

2000-11-07 15:35  aleidinger

	* configure.in:
	  - HAVE{MPGLIB,VORBIS,GTK} -> HAVE_$1
	  - --disable-analyzer-hooks
	  - --enable-efence (use the ElectricFence malloc debugging
	  library)

2000-11-07 12:42  aleidinger

	* frontend/Makefile.in, libmp3lame/Makefile.in:
	  regen after amiga_mpega.c move

2000-11-07 12:41  aleidinger

	* frontend/Makefile.am, libmp3lame/Makefile.am:
	  follow the move of amiga_mpega.c from libmp3lame to frontend

2000-11-07 01:59  pfk

	* frontend/brhist.c, frontend/get_audio.c, frontend/get_audio.h,
	  frontend/main.c, libmp3lame/bitstream.c, libmp3lame/version.h:
	  minor changes.
	  
	  some options can be passed via "export LAMEOPT=..."

2000-11-07 00:58  cisc

	* Makefile.unix, frontend/amiga_mpega.c, libmp3lame/amiga_mpega.c:
	  Moved amiga_mpega.c to frontend where it belongs, and added it to
	  Makefile.unix

2000-11-07 00:14  markt

	* libmp3lame/bitstream.c, libmp3lame/encoder.c, libmp3lame/lame.c,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/util.c, libmp3lame/util.h, libmp3lame/vbrquantize.c:
	  removed lame_global_flags pointer from lame_internal_flags.

2000-11-06 23:19  markt

	* libmp3lame/lame.c, libmp3lame/quantize_pvt.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  fixed resample bug when resampling by integer amounts.

2000-11-06 19:26  cisc

	* API, frontend/gtkanal.c, frontend/parse.c, include/lame.h,
	  libmp3lame/VbrTag.c, libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/version.c, libmp3lame/version.h:
	  Improved get_lame_xxx() API a little bit more

2000-11-06 17:29  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  fixed segmentation fault in Frank's noise calculation
	  implemented Todd Richmond's idea to speed up the outer_loop()

2000-11-06 12:31  aleidinger

	* libmp3lame/i386:
	  add .lst files

2000-11-06 08:22  markt

	* Makefile.am, configure.in, frontend/.indent.pro,
	  frontend/brhist.c, libmp3lame/.indent.pro:
	  went back to frame counts in VBR display

2000-11-06 04:24  markt

	* frontend/get_audio.c, mpglib/interface.c, mpglib/mpglib.h:
	  cleaned up mpeg header search code in get_audio.c
	  modified mpglib so we can move VBR header code into mpglib
	  and out of the frontend code. a little more work is still needed.

2000-11-06 01:13  pfk

	* libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/id3tag.c, libmp3lame/util.o:
	  Small changes for fuzzy ID3 tag compare.

2000-11-06 00:58  cisc

	* frontend/gtkanal.c, frontend/parse.c, include/lame.h,
	  libmp3lame/VbrTag.c, libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/version.c, libmp3lame/version.h:
	  Modified get_lame_xxx() API so it's more re-entrance friendly
	  (returning a const pointer from a static var that's changed every
	  time the function is called is really dirty!).

2000-11-05 22:27  pfk

	* frontend/parse.c, frontend/rtp.h, frontend/timestatus.c,
	  libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  testcase.mp3:
	  Minor changes.
	  Added -X8.

2000-11-05 18:29  pfk

	* frontend/get_audio.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/quantize.c, libmp3lame/util.h, libmp3lame/util.o:
	  Changed lame_init_???_init to Class_ID.
	  On initialization this variable gets the value LAME_ID.
	  Advantages documented.
	  
	  Some little Bugs fixed.

2000-11-05 17:13  takehiro

	* Dll/Makefile.in, Makefile.in, configure, configure.in,
	  doc/Makefile.in, doc/html/Makefile.in, doc/man/Makefile.in,
	  frontend/Makefile.in, frontend/main.c, frontend/parse.c,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, libmp3lame/newmdct.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  misc/Makefile.in, mpglib/Makefile.in:
	  configure.in: typo fix
	  newmdct.c: precision fix(?)

2000-11-05 14:52  aleidinger

	* libmp3lame/i386/Makefile.in:
	  regen (clean of generated .lst files)

2000-11-05 14:52  aleidinger

	* libmp3lame/i386/Makefile.am:
	  add generated .lst files to files to clean

2000-11-05 14:41  aleidinger

	* misc/Makefile.in:
	  regen after adding scalartest.c

2000-11-05 14:40  aleidinger

	* misc/Makefile.am:
	  add scalartest.c

2000-11-05 14:34  aleidinger

	* libmp3lame/util.c~, libmp3lame/util.d, libmp3lame/util.o:
	  remove generated/backup files, seems they are commited by
	  accident

2000-11-05 13:15  pfk

	* libmp3lame/tools.c, libmp3lame/tools.h:
	  add file for generic tools.

2000-11-05 13:09  pfk

	* libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/i386/scalar.nas, libmp3lame/lame.c,
	  libmp3lame/lameerror.h, libmp3lame/psymodel.c,
	  libmp3lame/resample.c, libmp3lame/resample.h,
	  libmp3lame/tables.c, libmp3lame/tables.h, libmp3lame/util.c,
	  libmp3lame/util.c~, libmp3lame/util.d, libmp3lame/util.h,
	  libmp3lame/util.o, libmp3lame/version.h, misc/scalartest.c,
	  testcase.mp3:
	  * Changed psy_data from double to float (this ssem not to change
	  any bit in
	  MP3, but saves 20 KB).
	  
	  * Other, but small modifications

2000-11-05 00:19  pfk

	* frontend/brhist.c:
	  bugfix. Sorry

2000-11-04 18:15  pfk

	* frontend/brhist.c, frontend/brhist.h, frontend/mp3rtp.c,
	  frontend/rtp.c, frontend/rtp.h, frontend/timestatus.c,
	  include/lame.h, libmp3lame/i386/Makefile.am,
	  libmp3lame/i386/Makefile.in, libmp3lame/i386/scalar.nas,
	  libmp3lame/resample.c, libmp3lame/resample.h,
	  libmp3lame/tables.c, libmp3lame/version.c, misc/ath.c:
	  small changes in brhist.
	  some tests and modifications for streaming MP3.
	  other unremarkable stuff

2000-11-04 02:18  pfk

	* doc/html/switchs.html, frontend/get_audio.c,
	  frontend/timestatus.c:
	  Unimportant changes.

2000-11-03 14:21  aleidinger

	* libmp3lame/id3tag.c:
	  more 'const'nes, also saves some bytes in the generated .o

2000-11-03 14:13  aleidinger

	* libmp3lame/i386/Makefile.in:
	  regen after change in Makefile.am

2000-11-03 14:13  aleidinger

	* libmp3lame/i386/Makefile.am:
	  remove trailing backslash in last line of EXTRA_DIST

2000-11-03 14:08  aleidinger

	* doc/html/contributors.html:
	  add myself to contributors

2000-11-03 14:07  aleidinger

	* config.h.in, configure, configure.in, frontend/brhist.c:
	  NOTERMCAP/TERMCAP_AVAILABLE -> HAVE_TERMCAP

2000-11-03 14:00  aleidinger

	* frontend/parse.c:
	  add missing include

2000-11-03 00:47  pfk

	* frontend/brhist.c, frontend/console.c, frontend/console.h,
	  frontend/mp3rtp.c, frontend/parse.c, frontend/timestatus.c,
	  include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/i386/Makefile.am, libmp3lame/i386/Makefile.in,
	  libmp3lame/i386/scalar.nas, libmp3lame/lame.c,
	  libmp3lame/mpglib_interface.c, libmp3lame/resample.h,
	  libmp3lame/util.c, libmp3lame/version.c, libmp3lame/version.h:
	  Minor changes. Biggest ist in bitstream.c, but triggered by
	  KLEMM_14.

2000-11-02 17:16  pfk

	* TODO, frontend/brhist.c, frontend/console.c, frontend/console.h,
	  libmp3lame/i386/scalar.nas, libmp3lame/resample.c,
	  libmp3lame/resample.h, libmp3lame/version.c:
	  These are minor changes to a version which fails on my computer.
	  
	  * MMX reporting added to get_lame_version(9 as requested by TODO.
	  * Added 5 files for resampling and console IO.

2000-11-02 12:57  aleidinger

	* mpglib/common.c, mpglib/common.h:
	  bugfix for --disable-decoder-layer1 case

2000-11-01 21:56  markt

	* LICENSE, frontend/parse.c, include/lame.h:
	  updated LICENSE to remove comments about libsndfile,
	  since libsndfile no longer included in library

2000-11-01 21:54  markt

	* libmp3lame/.indent.pro, libmp3lame/bitstream.c,
	  libmp3lame/version.h:
	  added indent.pro back to CVS

2000-11-01 18:25  markt

	* STYLEGUIDE, libmp3lame/encoder.c:
	  some more context fixes

2000-11-01 18:22  markt

	* libmp3lame/newmdct.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c:
	  removed more 'context'.
	  
	  lame now segfaults.

2000-11-01 18:06  markt

	* libmp3lame/.indent.pro, libmp3lame/VbrTag.c, libmp3lame/fft.c,
	  libmp3lame/fft.h, libmp3lame/id3tag.c, libmp3lame/lame.c,
	  libmp3lame/newmdct.h, libmp3lame/psymodel.h,
	  libmp3lame/quantize.h, libmp3lame/quantize_pvt.h:
	  removing 'context'

2000-11-01 17:56  markt

	* libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/util.h:
	  restored bitstream.c to original.
	  drain_into_ancillary_data was written the way it is
	  on purpose. dont change it without checking with me first

2000-11-01 17:46  markt

	* libmp3lame/util.h:
	  *** empty log message ***

2000-11-01 17:40  markt

	* libmp3lame/util.h:
	  removed 'context'
	  why was lame_internal_flags renamed context?
	  This is under the catagory of cosmetic.

2000-11-01 17:32  markt

	* STYLEGUIDE, frontend/get_audio.c, frontend/parse.c,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c:
	  removed some cosmetic changes.

2000-11-01 16:09  robert

	* frontend/mp3rtp.c, frontend/rtp.c:
	  mp3rtp was broken, does now compile again

2000-11-01 14:35  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.h:
	  only cosmetic changes, done by indent

2000-11-01 14:31  robert

	* frontend/.indent.pro, libmp3lame/.indent.pro, misc/.indent.pro:
	  these files are used by the indent program to pretty format
	  sources

2000-11-01 00:01  robert

	* frontend/brhist.c, frontend/brhist.h, frontend/main.c,
	  frontend/timestatus.c:
	  some more brhist pretty print

2000-10-31 13:58  aleidinger

	* frontend/Makefile.in:
	  regen after bugfix in Makefile.am

2000-10-31 13:57  aleidinger

	* frontend/Makefile.am:
	  fix build of mp3x

2000-10-31 13:18  robert

	* Makefile.MSVC, frontend/brhist.c, frontend/lametime.h:
	  fixes for bitrate histogram and MS compilation

2000-10-31 12:18  robert

	* frontend/lametime.c, frontend/parse.c, frontend/timestatus.c:
	  quick fixes for Console_IO stuff
	  Console_IO is declared in brhist.c, but when compiling
	  without bitrate histogram capabilities the code would
	  not compile anymore

2000-10-30 23:58  robert

	* Makefile.MSVC, Makefile.unix:
	  analyzer stuff was moved to frontend
	  because of automake/configure

2000-10-30 22:34  robert

	* libmp3lame/vbrquantize.c:
	  Takehiro's IEEE754 hack for VBR modes new and mtrh

2000-10-30 11:28  aleidinger

	* Makefile.in:
	  we didn't need acconfig.h, so don't depend on it

2000-10-30 10:51  robert

	* libmp3lame/util.c:
	  BUG fix for debug code for BUG fix for fill_buffer_resample() ;-(

2000-10-30 10:48  robert

	* libmp3lame/util.c:
	  BUG fix for fill_buffer_resample()

2000-10-29 22:26  robert

	* libmp3lame/util.c:
	  small fix for fill_buffer_resample()
	  some buffers where too small allocated, but this will
	  not really help, the routine has some ugly looking points
	  left that need to be fixed.

2000-10-29 21:11  robert

	* libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/encoder.c,
	  libmp3lame/id3tag.c, libmp3lame/lame.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/reservoir.c,
	  libmp3lame/reservoir.h, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/vbrquantize.c:
	  gfp gfc stuff FINISHED for now

2000-10-29 20:03  robert

	* libmp3lame/encoder.c, libmp3lame/fft.c, libmp3lame/fft.h,
	  libmp3lame/newmdct.c, libmp3lame/newmdct.h,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c:
	  a step further with this gfp gfc stuff

2000-10-29 19:15  robert

	* libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c,
	  libmp3lame/util.h:
	  some more gfc stuff

2000-10-29 18:12  robert

	* libmp3lame/encoder.c, libmp3lame/quantize.c,
	  libmp3lame/quantize.h, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c:
	  more on gfp gfc reorganization

2000-10-29 16:37  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h, libmp3lame/takehiro.c,
	  libmp3lame/vbrquantize.c:
	  gfp gfc reorganization

2000-10-29 13:58  robert

	* libmp3lame/fft.c, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.h:
	  data encapsulating Naoki's psymode

2000-10-29 12:54  aleidinger

	* analyzer/Makefile.in, analyzer/Makefile.in.old,
	  analyzer/README.Win32, analyzer/gpkplotting.c,
	  analyzer/gpkplotting.h, analyzer/gtkanal.c, analyzer/gtkanal.h,
	  analyzer/mp3x.c, analyzer/mp3x.dsp, analyzer/mp3x.dsw:
	  they life now in frontend/...

2000-10-29 12:52  aleidinger

	* .:
	  add some config.h related files

2000-10-29 12:51  aleidinger

	* configure:
	  now with support for config.h

2000-10-29 12:50  aleidinger

	* Dll/Makefile.in, Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, misc/Makefile.in,
	  mpglib/Makefile.in:
	  regenerated after adding support for config.h

2000-10-29 12:49  aleidinger

	* aclocal.m4, configure.in, frontend/Makefile.am,
	  frontend/brhist.c, frontend/get_audio.c, frontend/gpkplotting.c,
	  frontend/gtkanal.c, frontend/ieeefloat.c, frontend/lametime.c,
	  frontend/main.c, frontend/mp3rtp.c, frontend/mp3x.c,
	  frontend/parse.c, frontend/portableio.c, frontend/rtp.c,
	  frontend/timestatus.c, libmp3lame/Makefile.am,
	  libmp3lame/VbrTag.c, libmp3lame/amiga_mpega.c,
	  libmp3lame/bitstream.c, libmp3lame/debugscalefac.c,
	  libmp3lame/encoder.c, libmp3lame/fft.c, libmp3lame/id3tag.c,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/newmdct.c, libmp3lame/psymodel.c,
	  libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/reservoir.c, libmp3lame/tables.c,
	  libmp3lame/takehiro.c, libmp3lame/util.c,
	  libmp3lame/vbrquantize.c, libmp3lame/version.c,
	  libmp3lame/vorbis_interface.c, mpglib/Makefile.am,
	  mpglib/common.c, mpglib/dct64_i386.c, mpglib/decode_i386.c,
	  mpglib/interface.c, mpglib/layer1.c, mpglib/layer2.c,
	  mpglib/layer3.c, mpglib/tabinit.c:
	  support for config.h

2000-10-29 12:47  aleidinger

	* config.h.in, stamp-h.in:
	  config.h support

2000-10-28 21:57  pfk

	* configure, libmp3lame/bitstream.c, libmp3lame/bitstream.h,
	  libmp3lame/lame.c:
	  CRC table based code removed
	  
	  Ease CRC header calculation

2000-10-28 14:23  pfk

	* frontend/get_audio.c, frontend/lametime.c, frontend/lametime.h,
	  frontend/main.c, frontend/mp3x.c, frontend/parse.c,
	  include/lame.h, libmp3lame/bitstream.c, libmp3lame/lame.c,
	  libmp3lame/util.c, libmp3lame/util.h:
	  Moved 2 functions from util.c to lametime.c
	  
	  Fixed a memory leak caused by an early return
	  
	  bitstream: Local crc moved to gfp->crcvalue, next step can be to
	  remove a
	  function parameter
	  
	  minor changes

2000-10-28 12:31  robert

	* include/lame.h, libmp3lame/VbrTag.c, libmp3lame/bitstream.c,
	  libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  hunting some BUGs, but still not found
	  encoding a mono VBR resampled to 11 kHz shows some strange
	  behaviour

2000-10-28 10:47  aleidinger

	* Dll/Makefile.in, doc/Makefile.in, doc/html/Makefile.in,
	  doc/man/Makefile.in, frontend/Makefile.in, include/Makefile.in,
	  libmp3lame/Makefile.in, libmp3lame/i386/Makefile.in,
	  misc/Makefile.in, mpglib/Makefile.in:
	  don't use gtar instead of tar

2000-10-28 10:42  aleidinger

	* lame.spec:
	  lame.spec gets now generated at configure time from lame.spec.in

2000-10-28 10:41  aleidinger

	* Makefile.am, Makefile.in:
	  - add lame.spec.in to distribution
	  - remove lame.spec from distribution
	  - show how long lame needs for the testcase

2000-10-28 10:36  aleidinger

	* configure, configure.in:
	  - generate lame.spec from lame.spec.in
	  - CPU specific optimizations (x86, gcc 2.7.3.x or better)

2000-10-28 10:32  aleidinger

	* lame.spec.in:
	  template for configure to generate lame.spec

2000-10-28 01:29  pfk

	* include/lame.h, libmp3lame/lame.c, libmp3lame/util.c,
	  libmp3lame/util.h, libmp3lame/version.h:
	  short => sample_t in lame_encode_buffer and all the related stuff
	  
	  some essential but simple speedups in resampling code
	  
	  resulting MP3 will be different from previous
	  
	  code modification not finished, this is only half of the work

2000-10-27 23:04  robert

	* libmp3lame/vbrquantize.c:
	  max_range tables

2000-10-27 21:50  pfk

	* frontend/main.c, libmp3lame/lame.c, libmp3lame/psymodel.c,
	  libmp3lame/util.c, libmp3lame/util.h:
	  KLEMM_04(old) removed, scalar routines introduced.

2000-10-27 21:34  robert

	* libmp3lame/vbrquantize.c:
	  LSF needs some different max_range tables
	  I plugged some in and it seems to work
	  they are designed after max_sfac_tab[6][4]
	  you may look into scale_bitcount_lsf()

2000-10-27 16:30  robert

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c:
	  LSF hack for vbr-mtrh VBR mode

2000-10-27 15:48  takehiro

	* libmp3lame/newmdct.c:
	  bit optimization

2000-10-27 11:27  robert

	* libmp3lame/quantize_pvt.c, mpglib/layer3.c:
	  BUG in frame analyzer fixed, there was some wild pointer used
	  in layer3.c in preflag condition

2000-10-27 11:19  takehiro

	* libmp3lame/i386/choose_table.nas, libmp3lame/machine.h,
	  libmp3lame/tables.c, libmp3lame/tables.h:
	  64bit int lookup table moved from C code to asm code.

2000-10-26 19:07  pfk

	* frontend/brhist.c, frontend/get_audio.c, frontend/mp3rtp.c,
	  frontend/timestatus.c, include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/psymodel.c, libmp3lame/util.c, misc/ath.c:
	  Make some modifications active.
	  
	  No Termcap terms are now handled in the same way like termcap
	  terms.

2000-10-25 22:18  robert

	* libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.h, libmp3lame/vbrquantize.c:
	  improved vbr_mtrh VBR mode

2000-10-25 14:03  aleidinger

	* frontend/Makefile.am:
	  prevent the CVS Id from being integrated into Makefile.in

2000-10-24 22:47  robert

	* Makefile.MSVC, Makefile.unix, libmp3lame/quantize.c:
	  improved RH_AMP

2000-10-24 22:17  pfk

	* frontend/get_audio.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/psymodel.c:
	  Better syncword recognization

2000-10-24 18:11  takehiro

	* configure, configure.in, frontend/Makefile.in,
	  libmp3lame/machine.h:
	  configure/configure.in
	  added "--enable-all-float", which make FLOAT8 as float
	  
	  machine.h
	  FLOAT (and FLOAT8) typedef fix for alpha processor

2000-10-24 17:44  takehiro

	* libmp3lame/newmdct.c:
	  loop integration.
	  "make test" may find some difference caused by some rounding
	  problem

2000-10-24 15:01  robert

	* libmp3lame/quantize.c:
	  fix for -q7 mode

2000-10-24 13:54  robert

	* libmp3lame/quantize.c:
	  typo

2000-10-24 13:28  robert

	* libmp3lame/quantize.c, libmp3lame/quantize_pvt.c,
	  libmp3lame/quantize_pvt.h:
	  some simplifications

2000-10-24 11:40  robert

	* Makefile.MSVC, Makefile.unix, libmp3lame/quantize.c:
	  improvements, need some compile time defines, look into
	  Makefile.unix

2000-10-23 18:59  robert

	* libmp3lame/machine.h:
	  FLOAT is float, FLOAT8 is double again ;-)

2000-10-23 15:50  takehiro

	* Makefile.am, Makefile.in, frontend/Makefile.in:
	  more make test update

2000-10-23 15:44  takehiro

	* libmp3lame/machine.h:
	  sorry, mistaken check in....

2000-10-23 15:40  takehiro

	* Dll/Makefile.in, Makefile.am, Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, frontend/Makefile.in,
	  include/Makefile.in, lame.spec, libmp3lame/Makefile.in,
	  libmp3lame/i386/Makefile.in, libmp3lame/machine.h,
	  misc/Makefile.in, mpglib/Makefile.in:
	  "make test" update

2000-10-23 15:40  takehiro

	* libmp3lame/bitstream.c:
	  cosmetic change

2000-10-23 10:16  aleidinger

	* Makefile.in:
	  update after change to Makefile.am

2000-10-23 10:16  aleidinger

	* Makefile.am:
	  let test targetbuild frontend/lame if it isn't already build
	  - didn't work with "make -j 4 test"
	  - IMHO the previous version was a feature
	  (requested by Takehiro-san)
	  
	  splittet test target into more subtargets
	  
	  corrected testg target
	  
	  added CVS_RSH=ssh to update/diff target

2000-10-22 22:50  robert

	* libmp3lame/psymodel.c:
	  fix for short block I introduced 2000-10-19

2000-10-22 18:20  pfk

	* Makefile.unix, frontend/brhist.c, frontend/brhist.h,
	  frontend/get_audio.c, frontend/main.c, frontend/parse.c,
	  frontend/parse.h, include/lame.h, libmp3lame/bitstream.c,
	  libmp3lame/encoder.c, libmp3lame/encoder.h, libmp3lame/lame.c,
	  libmp3lame/reservoir.c, libmp3lame/util.c, libmp3lame/util.h,
	  libmp3lame/version.c, libmp3lame/version.h, misc/ath.c:
	  A lot of small changes, most of them must be switched on via
	  compile time
	  defines.
	  
	  vbr displays shows MS and LR frames different.
	  
	  vbr displays depends on the current display width.

2000-10-22 15:36  takehiro

	* Makefile.am:
	  do "make test" with one step

2000-10-22 15:28  takehiro

	* ., Dll, doc, doc/html, doc/man, frontend, include, libmp3lame,
	  libmp3lame/i386, misc, mpglib:
	  igonore Makefile.am things (.libs, .deps, and so on)

2000-10-22 15:17  takehiro

	* libmp3lame, mpglib:
	  added *.lo

2000-10-22 12:06  aleidinger

	* ltconfig, ltmain.sh, missing, mkinstalldirs:
	  generated by automake/libtool

2000-10-22 12:03  aleidinger

	* frontend/Makefile.in, libmp3lame/Makefile.in, mpglib/Makefile.in:
	  generated by automake

2000-10-22 12:01  aleidinger

	* configure:
	  now with automake/libtool support

2000-10-22 12:00  aleidinger

	* confdefs.h:
	  seems to get generated at configure time

2000-10-22 11:58  aleidinger

	* aclocal.m4:
	  generated by aclocal

2000-10-22 11:58  aleidinger

	* configure.in:
	  - add asm support if nasm is found
	  (if we support this, see ASM_FOR_ARCH)
	  - added automake support
	  - added libtool support
	  (a Dll user should look into "AC_LIBTOOL_WIN32_DLL" of the
	  libtool
	  documentation and send apropriate patches)
	  - --enable-{mp3x,mp3rtp}
	  - only build analyzer if GTK is present
	  - set CPUTYPE (to use asm routines where supported)
	  - --enable-decoder* -> --disable-decoder*

2000-10-22 11:57  aleidinger

	* acinclude.m4:
	  remove GTK entry, "aclocal" takes care of it in aclocal.m4

2000-10-22 11:55  aleidinger

	* Dll/Makefile.in, Makefile.in, doc/Makefile.in,
	  doc/html/Makefile.in, doc/man/Makefile.in, include/Makefile.in,
	  libmp3lame/i386/Makefile.in, misc/Makefile.in:
	  generated by automake

2000-10-22 11:53  aleidinger

	* misc/Makefile.am:
	  Initial automake support

2000-10-22 11:50  aleidinger

	* frontend/README.Win32, frontend/gpkplotting.c,
	  frontend/gpkplotting.h, frontend/gtkanal.c, frontend/gtkanal.h,
	  frontend/mp3x.c, frontend/mp3x.dsp, frontend/mp3x.dsw:
	  moved from analyzer/... because automake didn't allows sources
	  for one
	  program/library to be in more than one directory (at least
	  without
	  hacks)

2000-10-22 11:46  aleidinger

	* Makefile.am:
	  - initial automake support
	  - delete testcase.new.mp3 on "clean"
	  - added "update" target, does a cvs update
	  - added "diff" target, does a cvs diff -u

2000-10-22 11:44  aleidinger

	* frontend/Makefile.am:
	  - initial automake support
	  - conditionally build mp3x/mp3rtp

2000-10-22 11:43  aleidinger

	* libmp3lame/Makefile.am, libmp3lame/i386/Makefile.am:
	  Initial automake/libtool support:
	  - if nasm is detected and we have asm for this @CPUTYPE@:
	  * include asm support
	  - didn't build without decoder: "undefined reference to
	  `mpg123_pinfo'"

2000-10-22 11:41  aleidinger

	* mpglib/Makefile.am:
	  Initial automake/libtool support:
	  - mpglib gets linked to libmp3lame.{a,so}
	  - isn't installed on it's own

2000-10-22 11:39  aleidinger

	* Dll/Makefile.am, doc/Makefile.am, doc/html/Makefile.am,
	  doc/man/Makefile.am, include/Makefile.am:
	  Initial automake support.

2000-10-22 10:19  takehiro

	* libmp3lame/l3side.h, libmp3lame/newmdct.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/takehiro.c, libmp3lame/vbrquantize.c:
	  1. changed sfb_smax to sfb_smin. it is really minimum number, and
	  so confusing.
	  2. prepare for mixed block
	  3. prepare for asm quantize code

2000-10-22 09:43  takehiro

	* libmp3lame/fft.c, libmp3lame/util.h:
	  no need to be in internal flag for constants

2000-10-22 09:41  takehiro

	* libmp3lame/fft.c:
	  no need to be in internal flag for constants

2000-10-22 09:27  takehiro

	* libmp3lame/Makefile.in, libmp3lame/encoder.c,
	  libmp3lame/newmdct.c, libmp3lame/newmdct.h, libmp3lame/util.h:
	  1 (util.h) deleted unused flag from internal flags
	  2 (newmdct) in place code of mdct_short
	  3 some cosmetic change

2000-10-22 08:25  takehiro

	* libmp3lame/newmdct.c:
	  code clean up

2000-10-22 06:56  takehiro

	* libmp3lame/machine.h:
	  deleted old remark about asm routine

2000-10-22 06:56  takehiro

	* libmp3lame/newmdct.c:
	  code cleanup

2000-10-21 19:28  takehiro

	* libmp3lame/i386/choose_table.nas, libmp3lame/quantize_pvt.c:
	  reduce memory assumption little and bit faster choose_table

2000-10-21 19:11  takehiro

	* libmp3lame/newmdct.c:
	  add coment and remove unused code

2000-10-21 19:01  takehiro

	* libmp3lame/newmdct.c:
	  bit optimization and now thread safe, I think.

2000-10-21 17:24  takehiro

	* libmp3lame/newmdct.c:
	  brute force loop unrolling and some elimination of multiple.
	  bit faster.

2000-10-21 13:05  robert

	* libmp3lame/encoder.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/vbrquantize.c:
	  frame analyzer info will now be set up in main encoding
	  routine only, for the whole frame. problem with SCFSI fixed

2000-10-20 21:46  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  Added lame alpha, beta version and if MMX is enabled to the
	  version interface,
	  should be backward compatible

2000-10-20 21:43  afaber

	* frontend/brhist.c:
	  Fixed histgramming display for Win32, added fflush

2000-10-20 21:11  afaber

	* analyzer, analyzer/README.Win32, analyzer/README.Win32~:
	  Small fixes to the Win32 installation instructions, Added Win32
	  type to the .cvsignore file

2000-10-20 20:50  afaber

	* frontend/brhist.c, frontend/brhist.h:
	  Resolved const mistach beween BRHIST function declartions

2000-10-20 20:49  afaber

	* libmp3lame/libmp3lame.dsp:
	  Added MPGLIB define

2000-10-20 19:28  robert

	* misc, misc/abx.c, misc/ath.c:
	  initial checkin

2000-10-20 13:04  robert

	* libmp3lame/quantize.c, libmp3lame/vbrquantize.c:
	  I think I fixed a BUG in subblock gain code

2000-10-19 21:10  robert

	* frontend/parse.c, include/lame.h, libmp3lame/lame.c,
	  libmp3lame/psymodel.c, libmp3lame/quantize.c:
	  fixed BUG in atQ table, removed --raise-smr stuff

2000-10-18 17:34  robert

	* analyzer/mp3x.c, libmp3lame/lame-analysis.h:
	  no way to disable analysis for lame-lib

2000-10-18 17:04  robert

	* analyzer/gtkanal.h:
	  first commit

2000-10-18 16:47  robert

	* analyzer/gtkanal.c, analyzer/mp3x.c, libmp3lame/lame-analysis.h,
	  mpglib/layer3.c, mpglib/mpglib.h:
	  fixed some ambiguities
	  define NOANALYSIS as suggested in INSTALL if you don't want
	  to collect some analysis data

2000-10-18 11:18  robert

	* frontend/brhist.c, include/lame.h, libmp3lame/lame.c:
	  some tweaks

2000-10-18 11:02  shibatch

	* libmp3lame/psymodel.c:
	  Rewrote double to FLOAT8.

2000-10-18 10:55  shibatch

	* libmp3lame/psymodel.c:
	  Added short block pre-echo control to --nspsytune.
	  This improves castanets.wav.

2000-10-17 22:59  pfk

	* STYLEGUIDE, frontend/brhist.c, frontend/brhist.h,
	  frontend/get_audio.h, frontend/main.c, frontend/parse.c,
	  frontend/parse.h, frontend/timestatus.c, include/lame.h,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/psymodel.c, libmp3lame/util.h:
	  * small changes to prepare to add a lame float interface
	  * other stuff

2000-10-17 00:13  pfk

	* Makefile.unix, frontend/brhist.c, frontend/brhist.h,
	  frontend/get_audio.c, frontend/timestatus.c, include/lame.h,
	  libmp3lame/VbrTag.c, libmp3lame/bitstream.c, libmp3lame/fft.c,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/version.c,
	  libmp3lame/version.h:
	  removed some shorts.
	  two version of version report
	  some minor changes

2000-10-16 22:19  robert

	* frontend/brhist.c, frontend/main.c:
	  cosmetics

2000-10-16 19:55  florian

	* Makefile.in, analyzer/Makefile.in, configure, configure.in,
	  libmp3lame/Makefile.in, libmp3lame/lame.c, mpglib/Makefile.in:
	  integrated Alexanders patch, minor bug fixes

2000-10-16 19:29  robert

	* libmp3lame/tables.c, libmp3lame/tables.h, libmp3lame/takehiro.c:
	  takehiro.c now thread safer

2000-10-16 19:27  robert

	* frontend/brhist.h:
	  missing include files

2000-10-16 10:03  robert

	* Makefile.unix:
	  some convenience

2000-10-16 02:40  markt

	* include/lame.h, libmp3lame/VbrTag.c:
	  *** empty log message ***

2000-10-16 02:29  markt

	* include/lame.h:
	  added some prototypes. no code yet

2000-10-16 02:29  markt

	* API:
	  added some lame_set_variable prototypes in lame.h

2000-10-16 02:17  markt

	* frontend/mp3rtp.c, include/lame.h, libmp3lame/fft.h,
	  libmp3lame/lame.c:
	  fixed fft.h defines

2000-10-16 02:12  markt

	* API, Makefile.in, analyzer/mp3x.c, configure, configure.in:
	  broken frame analyzer fixed - -DANALYSIS was missing from default
	  library build

2000-10-16 01:38  markt

	* frontend/get_audio.c, frontend/main.c, frontend/mp3rtp.c:
	  fixed broken fskip()

2000-10-16 01:37  markt

	* Dll/BladeMP3EncDLL.c, STYLEGUIDE:
	  fixed Frank's broken fskip()

2000-10-16 00:44  markt

	* libmp3lame/Makefile.in, libmp3lame/mpglib_interface.c,
	  libmp3lame/util.c, libmp3lame/util.h:
	  made data for resampling dynamically allocated

2000-10-16 00:28  pfk

	* STYLEGUIDE, frontend/brhist.c, frontend/brhist.h,
	  frontend/get_audio.c, frontend/parse.c, frontend/timestatus.c,
	  frontend/timestatus.h:
	  Most changes must be switched on via KLEMM_0[0-9].
	  Also some minor changes.
	  Start of short documenation of the source code.
	  
	  _-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_
	  
	  ****************
	  *** #defines ***
	  ****************
	  
	  KLEMM_01: Use my experimental data to calculate ATH
	  But masking for >16 kHz seems to be buggy.
	  KLEMM_02: lame_set_stream_binary_mode() and lame_get_file_size()
	  as
	  lib function, still searching a better place for that
	  KLEMM_03: not more used
	  KLEMM_04: faster FIR filter with less rounding errors
	  KLEMM_05: sparse VBR bitrate display
	  KLEMM_06: fskip() uses fseek() if possible
	  KLEMM_07: use of timestatus_klemm.c which only uses timestatus.c
	  functions,
	  some code clarification and documentation

2000-10-15 22:46  markt

	* libmp3lame/util.c, libmp3lame/util.h:
	  lowpass filtering data structurs made dynamic

2000-10-15 21:23  markt

	* USAGE:
	  *** empty log message ***

2000-10-15 21:17  markt

	* frontend/get_audio.c, libmp3lame/VbrTag.c:
	  Fixed bug parsing Xing VBR header: lame will now compute
	  correct number of frames when decoding

2000-10-15 20:54  markt

	* Makefile.unix, frontend/get_audio.c, frontend/main.c:
	  playback bugfix

2000-10-15 20:12  markt

	* Dll/BladeMP3EncDLL.c:
	  msv6 dll project files updated

2000-10-15 20:05  markt

	* Dll/MP3EncDll.dsp, Dll/MP3EncDll.dsw:
	  reverted back to MSVC5 project files.
	  accidently changed them to my MSVC6 versions.
	  
	  Albert, when are you goint to upgrade!!!!

2000-10-15 20:02  markt

	* Dll/MP3EncDll.dsp, Dll/MP3EncDll.dsw:
	  msv6 dll project files updated

2000-10-15 19:51  markt

	* libmp3lame/lame.c:
	  fixed typo

2000-10-15 19:33  markt

	* USAGE, include/lame.h, libmp3lame/lame.c:
	  updated USAGE

2000-10-15 19:10  markt

	* API, Dll/BladeMP3EncDLL.c, USAGE:
	  edited API, USAGE

2000-10-14 19:18  robert

	* libmp3lame/i386/cpu_feat.nas:
	  BUG in 3DNow detection fixed, was my faulty porting to NASM

2000-10-14 11:42  robert

	* libmp3lame/takehiro.c, libmp3lame/util.h:
	  takehiro.c now thread safe ?

2000-10-14 10:08  pfk

	* frontend/brhist.c, libmp3lame/psymodel.c:
	  psymodel: FIR filter changed, trigger with KLEMM_04
	  brhist: display changed back to the old display, new display must
	  be
	  triggered by KLEMM_05

2000-10-13 22:58  pfk

	* frontend/brhist.c, frontend/brhist.h, frontend/main.c,
	  frontend/timestatus.c, include/lame.h, libmp3lame/util.c:
	  Changes are commented in brhist.c or must be enabled with
	  KLEMM_xx.

2000-10-13 16:43  robert

	* frontend/brhist.c, frontend/brhist.h, frontend/main.c,
	  include/lame.h, libmp3lame/lame.c:
	  bitrate histogram is working again

2000-10-13 15:24  robert

	* Makefile.unix:
	  small fixes

2000-10-13 05:37  florian

	* frontend/Makefile.in, frontend/Makefile.in.old,
	  libmp3lame/Makefile.in, libmp3lame/Makefile.in.old:
	  prepare for automake: new vars, system dependent checks to
	  configure.in

2000-10-13 05:17  florian

	* Makefile.in, Makefile.in.old, analyzer/Makefile.in,
	  analyzer/Makefile.in.old, configure, configure.in,
	  configure.in.old:
	  prepare for automake: new vars, system dependent checks to
	  configure.in

2000-10-12 19:19  florian

	* acinclude.m4, aclocal.m4, configure.in:
	  Preparing for automake

2000-10-12 17:48  robert

	* Makefile.MSVC, Makefile.unix:
	  NASM specific fixes

2000-10-12 17:19  robert

	* libmp3lame/i386/cpu_feat.nas:
	  CPU feature code using NASM

2000-10-12 17:18  robert

	* Makefile.MSVC, libmp3lame/Makefile.in, libmp3lame/util.c:
	  CPU feature detection code now using NASM

2000-10-12 08:59  shibatch

	* libmp3lame/psymodel.c, libmp3lame/quantize.c,
	  libmp3lame/quantize_pvt.c, libmp3lame/util.h:
	  Add two new features to --nspsytune.
	  1. New block type selecting algorithm.
	  2. Improved noise shaping at high bitrate CBR. Turned on by -q1
	  --nspsytune.

2000-10-12 00:50  robert

	* Makefile.MSVC:
	  it's getting better :-)

2000-10-12 00:49  robert

	* frontend/parse.c:
	  BUG fix in case LAME compiled without GTK, then called with -g

2000-10-11 23:08  robert

	* libmp3lame/Makefile.in, libmp3lame/quantize_pvt.c,
	  libmp3lame/util.c:
	  fixes for Frank's ATH

2000-10-11 18:45  robert

	* include/lame.h, libmp3lame/lame.c:
	  added alternative to lame_encode_finish:
	  - lame_encode_flush, does the same as lame_encode_finish
	  but will not free lame internal buffers
	  - lame_close, will free lame internal buffers
	  
	  So the following is equivalent to lame_encode_finish
	  lame_encode_flush()
	  <do some stuff like statistics>
	  lame_close()
	  
	  But this adds the possibility to get some more
	  analyzing after all frames are encoded.
	  
	  Suggested to use with
	  lame_bitrate_hist()
	  lame_stereo_mode_hist()

2000-10-11 17:54  robert

	* libmp3lame/encoder.c, libmp3lame/lame.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  simple statistics (histogram) added,
	  problem: lame_encode_finish() also frees internal buffer
	  solution: two new functions splitting lame_encode_finish()
	  into lame_encode_flush() and lame_close()

2000-10-11 01:03  florian

	* analyzer:
	  *** empty log message ***

2000-10-10 23:18  pfk

	* libmp3lame/util.c:
	  Additional ATH calculation (enabled by KLEMM_01).

2000-10-10 20:26  robert

	* libmp3lame/lame.c, libmp3lame/takehiro.c:
	  use of MMX should be safe on non MMX CPUs now,
	  but this will not work on Microsoft platforms.
	  We will have to adapt the assembler routines
	  for the MS assembler or for NASM.

2000-10-10 19:36  robert

	* Makefile.MSVC, libmp3lame/util.c:
	  Makefile work

2000-10-10 18:34  robert

	* libmp3lame/lame.c:
	  cut'n'paste devil :(

2000-10-10 18:28  robert

	* libmp3lame/lame.c, libmp3lame/util.c, libmp3lame/util.h:
	  Frank's CPU feature detection code added

2000-10-10 17:59  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h, frontend/brhist.c,
	  libmp3lame/VbrTag.c, libmp3lame/libmp3lame.dsp:
	  MSVC Win32 updates, added Alpha Beta version numbers to DLL
	  version structure
	  (should be backward compatible)

2000-10-10 05:31  markt

	* TODO:
	  notes about what needs to be done for configure

2000-10-09 19:55  markt

	* libmp3lame/encoder.h, libmp3lame/lame-analysis.h:
	  *** empty log message ***

2000-10-09 19:55  markt

	* TODO, analyzer/Makefile.in, analyzer/gpkplotting.c,
	  analyzer/gtkanal.c, analyzer/mp3x.c, frontend/timestatus.c,
	  include/lame.h:
	  generic cleanup

2000-10-09 19:42  markt

	* TODO:
	  frame analyzer now works with ./configure, at least for linux

2000-10-09 19:29  markt

	* Makefile.B32, Makefile.DJGPP, Makefile.MSVC, TODO, configure,
	  configure.in, frontend/get_audio.c, frontend/get_audio.h:
	  attept to fix Makefiles.B32, DJGPP, MSVC.

2000-10-09 16:27  markt

	* libmp3lame/lame-analysis.h, mp3x:
	  *** empty log message ***

2000-10-09 16:24  markt

	* mp3x/Makefile.in, mp3x/README.Win32, mp3x/aclocal.m4,
	  mp3x/config.status, mp3x/configure, mp3x/configure.in,
	  mp3x/gpkplotting.c, mp3x/gpkplotting.h, mp3x/gtkanal.c,
	  mp3x/mp3x.c, mp3x/mp3x.dsp, mp3x/mp3x.dsw:
	  removing mp3x

2000-10-09 16:23  markt

	* Makefile.in, Makefile.unix, aclocal.m4, analyzer,
	  analyzer/Makefile.in, analyzer/README.Win32,
	  analyzer/README.Win32~, analyzer/gpkplotting.c,
	  analyzer/gpkplotting.h, analyzer/gtkanal.c, analyzer/mp3x.c,
	  analyzer/mp3x.dsp, analyzer/mp3x.dsw, configure, configure.in,
	  frontend/Makefile.in, frontend/main.h,
	  libmp3lame/lame-analysis.h:
	  moved mp3x -> analyzer (name was clashing with executable name)
	  work on configure to get mp3x to compile (not yet done)
	  but gtk stuff in configure.in now working

2000-10-09 14:38  markt

	* libmp3lame/lame.c:
	  added a comment about downsampling

2000-10-08 20:08  markt

	* frontend/main.c, frontend/timestatus.c, include/lame.h,
	  libmp3lame/VbrTag.c, libmp3lame/VbrTag.h, libmp3lame/encoder.c,
	  libmp3lame/lame.c:
	  cleaned up VBR tag stuff

2000-10-08 19:43  markt

	* frontend/get_audio.c, frontend/get_audio.h, frontend/main.c,
	  frontend/main.h, frontend/mp3rtp.c, frontend/parse.c,
	  frontend/parse.h, frontend/timestatus.c, frontend/timestatus.h,
	  include/lame.h, libmp3lame/lame.c, libmp3lame/version.c,
	  libmp3lame/version.h, mp3x/gtkanal.c, mp3x/mp3x.c:
	  lots of changes, just trying to get things to compile.

2000-10-08 17:48  markt

	* libmp3lame/id3tag.h:
	  Trying to add id3tag.h again.

2000-10-07 08:25  robert

	* lame4dos.bat:
	  bug fixes by Alexander Stumpf

2000-10-07 05:04  markt

	* libmp3lame/id3tag.c, libmp3lame/util.h:
	  missing id3tag.h include files for util.h, id3tag.c

2000-10-07 05:03  markt

	* mp3x/gtkanal.c:
	  id3tag struct in id3tag.h

2000-10-07 05:00  markt

	* Dll/LameDLLInterface.htm, INSTALL, Makefile.unix,
	  frontend/main.c, frontend/main.h, include/lame.h,
	  libmp3lame/id3tag.c, libmp3lame/lame.c, libmp3lame/util.c,
	  libmp3lame/util.h:
	  id3tag_spec made it back into lame.h. deleted.
	  
	  frameNum has to be in the library - the front end has no way
	  of correctly counting frames unless it parses the mp3 output
	  headers. totalframes should also probably be in the library.

2000-10-06 21:16  florian

	* Makefile.in, configure, configure.in, frontend/Makefile.in,
	  libmp3lame/Makefile.in:
	  minor changes for install and uninstall

2000-10-06 13:19  takehiro

	* libmp3lame/id3tag.c, libmp3lame/lame.c, libmp3lame/util.h:
	  quick fix, maybe work but not checked hard

2000-10-06 13:19  takehiro

	* frontend/main.c, include/lame.h:
	  quick fix, I don't check this

2000-10-06 00:55  markt

	* libmp3lame/id3tag.c, libmp3lame/lame.c, libmp3lame/util.h:
	  id3tag cleanup

2000-10-06 00:54  markt

	* frontend/parse.c, include/lame.h:
	  id3 stuff cleaned up. id3 related strucs made internal to the
	  library

2000-10-05 19:53  cisc

	* frontend/get_audio.c, frontend/get_audio.h:
	  small correction

2000-10-05 05:18  markt

	* TODO, frontend/Makefile.in:
	  *** empty log message ***

2000-10-05 05:13  markt

	* Makefile.in, configure, configure.in, libmp3lame/Makefile.in:
	  more configure work.

2000-10-05 04:50  markt

	* testcase.mp3:
	  updated testcase

2000-10-05 04:49  markt

	* Makefile.in, libmp3lame/psymodel.c:
	  added comment about Naoki's blocktype bug fix.
	  also, for JSTEREO but with the '-d' option, L and R channels
	  can have different block types, so we cant just look at the
	  block type for channel 0

2000-10-05 04:31  markt

	* Makefile.in, frontend/Makefile.in:
	  forgot to add Makefile.in

2000-10-04 22:37  robert

	* Lame.vbs:
	  needs Windows Script >= 5.1, donated by Ralf Kempkens
	  <Ralf.Kempkens@epost.de>

2000-10-04 22:04  afaber

	* mp3x/mp3x.dsp:
	  MSVC project file update

2000-10-04 17:13  afaber

	* Dll/MP3EncDll.dsp, libmp3lame/libmp3lame.dsp, mpglib/mpglib.dsp:
	  MSVC project update(s) to new layout of the libmp3lame files

2000-10-04 17:13  afaber

	* Dll/Example.dsw, Dll/MP3EncDll.dsp:
	  MSVC project update(s) to new layout of the libmp3lame files

2000-10-04 12:48  shibatch

	* libmp3lame/psymodel.c:
	  Bugfix of long block pre-echo control.

2000-10-04 12:13  takehiro

	* i386, libmp3lame/i386, libmp3lame/i386/choose_table.nas,
	  libmp3lame/i386/fft.nas, libmp3lame/i386/fft3dn.nas,
	  libmp3lame/i386/fftfpu.nas, libmp3lame/i386/fftsse.nas,
	  libmp3lame/i386/ffttbl.nas, libmp3lame/i386/nasm.h:
	  moved i386 directory

2000-10-04 04:26  markt

	* libmp3lame/Makefile.in:
	  *** empty log message ***

2000-10-04 04:23  markt

	* configure, configure.in:
	  make still does not work. some kind of infinite loop

2000-10-04 04:00  markt

	* Makefile.unix:
	  more configure work

2000-10-04 03:54  markt

	* Makefile.in, Makefile.unix, VbrTag.c, VbrTag.h, amiga_mpega.c,
	  bitstream.c, bitstream.h, configure, configure.in,
	  debugscalefac.c, encoder.c, encoder.h, fft.c, fft.h, id3tag.c,
	  l3side.h, lame-analysis.h, lame.c, libmp3lame/Makefile.am,
	  libmp3lame/Makefile.in, libmp3lame/VbrTag.c, libmp3lame/VbrTag.h,
	  libmp3lame/amiga_mpega.c, libmp3lame/bitstream.c,
	  libmp3lame/bitstream.h, libmp3lame/debugscalefac.c,
	  libmp3lame/encoder.c, libmp3lame/encoder.h, libmp3lame/fft.c,
	  libmp3lame/fft.h, libmp3lame/id3tag.c, libmp3lame/l3side.h,
	  libmp3lame/lame-analysis.h, libmp3lame/lame.c,
	  libmp3lame/machine.h, libmp3lame/mpglib_interface.c,
	  libmp3lame/newmdct.c, libmp3lame/newmdct.h,
	  libmp3lame/psymodel.c, libmp3lame/psymodel.h,
	  libmp3lame/quantize.c, libmp3lame/quantize.h,
	  libmp3lame/quantize_pvt.c, libmp3lame/quantize_pvt.h,
	  libmp3lame/reservoir.c, libmp3lame/reservoir.h,
	  libmp3lame/tables.c, libmp3lame/tables.h, libmp3lame/takehiro.c,
	  libmp3lame/util.c, libmp3lame/util.h, libmp3lame/vbrquantize.c,
	  libmp3lame/version.c, libmp3lame/version.h,
	  libmp3lame/vorbis_interface.c, machine.h, mpglib_interface.c,
	  newmdct.c, newmdct.h, psymodel.c, psymodel.h, quantize.c,
	  quantize.h, quantize_pvt.c, quantize_pvt.h, reservoir.c,
	  reservoir.h, tables.c, tables.h, takehiro.c, util.c, util.h,
	  vbrquantize.c, version.c, version.h, vorbis_interface.c:
	  Moved everything to libmp3lame. worked on configure, but
	  not quite working.

2000-10-04 03:02  markt

	* frontend/aclocal.m4, frontend/configure, frontend/configure.in,
	  mpglib/configure, mpglib/configure.in:
	  removed more configure files in subdirectories.
	  There should only be one ./configure, in the main directory

2000-10-04 02:57  markt

	* configure, configure.in, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/timestatus.c,
	  frontend/timestatus.h:
	  more configure work

2000-10-04 01:07  markt

	* configure, configure.in, frontend/Makefile.in:
	  more configure patches for frontend.

2000-10-04 01:02  markt

	* TODO, configure, configure.in, frontend/configure:
	  new configure that doesn't run configure in all the
	  subdirectories!

2000-10-03 22:22  afaber

	* mp3x/gtkanal.c:
	  Fixes to MP3x app, seems like frame number was never increased,
	  caused assert error
	  and mp3done was not set when file/quit menu option was selected
	  (caused assert error also)

2000-10-03 22:09  afaber

	* README.WINGTK, frontend/lame.dsp, frontend/lame.dsw, lamelib.dsp,
	  libmp3lame/libmp3lame.dsp, mp3x, mp3x/README.Win32, mp3x/mp3x.c,
	  mp3x/mp3x.dsp, mp3x/mp3x.dsw, mpglib/mpglib.dsp:
	  Re-organized some of the MSVC project files, fixes to compile
	  mp3x again on Win32 platform

2000-10-03 17:56  robert

	* fft.c:
	  removed unused includes

2000-10-03 10:32  robert

	* fft.c, fft.h, lame.c, psymodel.c, quantize.c, quantize_pvt.c,
	  util.h:
	  fft.c now thread safe !?

2000-10-03 09:44  robert

	* encoder.c, newmdct.c, newmdct.h:
	  removed redundancies

2000-10-03 09:43  robert

	* testcase.mp3:
	  updated testcase for LAME 3.88 alpha 1

2000-10-03 04:07  markt

	* INSTALL:
	  Some configure options are for the library, some are for
	  the front end. How do we pass front end options to
	  frontend/configure?

2000-10-03 03:57  markt

	* Makefile.in:
	  *** empty log message ***

2000-10-03 03:52  markt

	* mp3x/Makefile.in, mp3x/aclocal.m4, mp3x/configure,
	  mp3x/configure.in:
	  better gtk check for mp3x

2000-10-02 10:15  markt

	* Makefile.in:
	  make -> $(MAKE)

2000-10-02 09:00  markt

	* mp3x/gtkanal.c, quantize_pvt.h:
	  Fix for winamp bug. (commented out for now)

2000-10-02 06:20  markt

	* aclocal.m4, configure, configure.in, frontend/configure,
	  frontend/configure.in, mp3x/config.log:
	  configure updates

2000-10-01 19:05  afaber

	* frontend/lame.dsp, lamelib.dsp, mpglib/mpglib.dsp:
	  Adapted project file in order to compile it again with the MSVC
	  compiler

2000-10-01 18:34  markt

	* Makefile.in:
	  *** empty log message ***

2000-10-01 18:25  markt

	* INSTALL, Makefile.in, frontend/Makefile.in, frontend/configure,
	  frontend/configure.in, frontend/main.c, frontend/parse.c,
	  id3tag.c, include/lame-analysis.h, include/lame-id3tag.h,
	  include/lame.h, lame-analysis.h, lame.c, mp3x/Makefile.in:
	  trying to move decoding back into library

2000-10-01 17:58  afaber

	* frontend/get_audio.c:
	  Make it compile again with MSVC compiler

2000-10-01 17:17  markt

	* frontend/gpkplotting.c, frontend/gpkplotting.h,
	  frontend/gtkanal.c, frontend/main.c, frontend/mp3x.c, mp3x,
	  mp3x/Makefile.in, mp3x/aclocal.m4, mp3x/config.log,
	  mp3x/config.status, mp3x/configure, mp3x/configure.in,
	  mp3x/gpkplotting.c, mp3x/gpkplotting.h, mp3x/gtkanal.c,
	  mp3x/mp3x.c:
	  moved GTK frame analyzer stuff into mp3x

2000-10-01 15:30  takehiro

	* Makefile.in, aclocal.m4, configure, configure.in,
	  frontend/Makefile.in, frontend/get_audio.c, frontend/main.c,
	  frontend/vorbis_interface.c, mpglib/Makefile.in,
	  mpglib/configure, mpglib/configure.in, mpglib/main.c,
	  mpglib_interface.c, vorbis_interface.c:
	  more modularization work.
	  VORBIS and mpglib interface is now in the libmp3lame library.
	  functions in these files are wrapper of mpglib/libvorbis.
	  I think it is ok to include them in libmp3lame.

2000-10-01 15:19  takehiro

	* VbrTag.c:
	  merged VbrTag.h into lame.h

2000-10-01 15:03  takehiro

	* Dll/BladeMP3EncDLL.c, encoder.c, frontend/get_audio.c,
	  include/lame.h, lame.c, util.h:
	  merged VbrTag.h into lame.h

2000-10-01 11:28  afaber

	* mpglib/common.c, mpglib/common.h, mpglib/dct64_i386.c,
	  mpglib/dct64_i386.h, mpglib/decode_i386.c, mpglib/decode_i386.h,
	  mpglib/interface.c, mpglib/interface.h, mpglib/layer1.c,
	  mpglib/layer1.h, mpglib/layer2.c, mpglib/layer2.h,
	  mpglib/layer3.c, mpglib/layer3.h, mpglib/main.c, mpglib/mpg123.h,
	  mpglib/mpglib.dsp, mpglib/mpglib.h:
	  More MPGLIB cleanup, got rid of the global mpstr struct gmp (in
	  order to make lib thread safe)

2000-10-01 10:40  afaber

	* mpglib/common.c, mpglib/common.h, mpglib/decode_i386.c,
	  mpglib/interface.c, mpglib/layer3.c, mpglib/mpg123.h:
	  Cleanup of mpg123.h file (still need to get rid of the dirty
	  extern gmp hack)

2000-10-01 10:05  afaber

	* mpglib/layer1.c, mpglib/mpg123.h:
	  Removed unused stuff in MPG123.h file, removed stuff that was
	  commented out in Layer1.c

2000-10-01 10:00  afaber

	* mpglib/mpg123.h, mpglib/tabinit.c:
	  Removed some double/float warnings, removed disable warnings
	  pragma for MSVC compiler

2000-10-01 09:52  takehiro

	* frontend/main.c, lame.c:
	  moved main_crc_init from frontend to library

2000-10-01 09:51  afaber

	* mpglib/common.c, mpglib/common.h, mpglib/dct64_i386.c,
	  mpglib/decode_i386.c, mpglib/interface.c, mpglib/layer2.c,
	  mpglib/layer3.h, mpglib/mpg123.h, mpglib/mpglib.dsp,
	  mpglib/tabinit.c, mpglib/tabinit.h:
	  Reorganized some of the function prototypes, removed non existing
	  function prototypes from common.h

2000-10-01 09:32  takehiro

	* libmp3lame:
	  initial check in for libmp3lame

2000-10-01 09:29  takehiro

	* libmp3lame, libmp3lame/Makefile.am:
	  check in for the feauture...

2000-10-01 09:10  afaber

	* bitstream.h, frontend/main.c:
	  Added void main_CRC_init (void) function prototype in bitstream.h

2000-10-01 06:24  takehiro

	* frontend/Makefile.in, frontend/mp3x.c:
	  making frameanalyzer problem fixed

2000-10-01 04:56  markt

	* util.c:
	  added Frank's normalization of Blackman filter coefficients.
	  This had no effect on any of my resample test cases.

2000-09-30 15:32  markt

	* STYLEGUIDE:
	  more notes on identation and int in STYLEGUIDE

2000-09-30 15:09  afaber

	* id3tag.c:
	  fixed signed/unsigned mismatch

2000-09-30 14:50  markt

	* HACKING, bitstream.c, bitstream.h, util.c, util.h:
	  added stuff in HACKING on how to handle global data
	  moved freegfc() into util.c

2000-09-30 14:28  markt

	* util.c:
	  number of pre-computed convolution windows determined
	  dynamically.
	  (storage still needs to malloc'd and free'd in
	  lame_encode_finish).

2000-09-30 14:17  robert

	* Makefile.in:
	  small fix

2000-09-30 13:14  afaber

	* Dll/BladeMP3EncDLL.c, Dll/MP3EncDll.dsp:
	  Changes in order to get it compile on a Win32 platform

2000-09-30 12:49  afaber

	* VbrTag.h, frontend/brhist.c, frontend/get_audio.c,
	  frontend/lame.dsp, frontend/lame.dsw, frontend/timestatus.c,
	  lame.dsp, lame.dsw, lamelib.dsp, mpglib/mpglib.dsp:
	  Changes in order to get it compile on a Win32 platform

2000-09-30 08:29  takehiro

	* frontend/parse.c:
	  renamed some functions

2000-09-30 08:21  takehiro

	* doc/html/contributors.html, doc/html/examples.html,
	  doc/html/history.html, doc/html/id3.html, doc/html/index.html,
	  doc/html/lame.css, doc/html/modes.html, doc/html/node6.html,
	  doc/html/switchs.html:
	  new document for 3.87 from Gaby.

2000-09-30 08:15  takehiro

	* frontend/configure, frontend/configure.in:
	  debug about gtk detection

2000-09-30 07:55  takehiro

	* frontend/get_audio.c, frontend/main.c, frontend/main.h,
	  frontend/mp3rtp.c, frontend/parse.c, include/lame.h, lame.c:
	  removed outpath from library.

2000-09-30 06:58  takehiro

	* bitstream.c, bitstream.h, util.c, util.h:
	  moved some code from util to bitstream

2000-09-30 04:40  markt

	* VbrTag.c, frontend/vorbis_interface.c, lame.c:
	  re-enabled id3v2 tag and Xing vbr header

2000-09-29 18:16  takehiro

	* frontend/get_audio.c, frontend/main.c, frontend/main.h,
	  frontend/parse.c, frontend/timestatus.c, include/lame.h, lame.c:
	  "silent" and "brhist" is moved to frontend

2000-09-29 18:07  takehiro

	* frontend/main.c, frontend/main.h, frontend/parse.c,
	  frontend/timestatus.c, include/lame.h, lame.c:
	  "update_interval" is moved to frontend

2000-09-29 17:50  takehiro

	* encoder.c, frontend/main.c, frontend/mp3x.c, frontend/parse.c,
	  include/lame-analysis.h, include/lame.h, lame.c, psymodel.c,
	  quantize.c, vbrquantize.c:
	  changed from "gtkflag" to "analysis", because it is not only for
	  gtk.

2000-09-29 17:48  takehiro

	* frontend/timestatus.c:
	  oops, timestatus not correctly displayed

2000-09-29 17:44  takehiro

	* frontend/gtkanal.c, frontend/main.c, frontend/mp3x.c,
	  frontend/parse.c, id3tag.c, include/analysis.h, include/id3tag.h,
	  include/lame-analysis.h, include/lame-id3tag.h, include/lame.h,
	  lame.c, mpglib/layer3.c, mpglib/mpglib.h, quantize.c,
	  quantize_pvt.c, util.h, vbrquantize.c:
	  OK, Florian, I renamed include file avoid confusing.
	  (it's so generic and not likely to be unique)

2000-09-29 17:31  takehiro

	* Makefile.in, encoder.c, frontend/Makefile.in, frontend/brhist.c,
	  frontend/get_audio.c, frontend/get_audio.h, frontend/gtkanal.c,
	  frontend/main.c, frontend/main.h, frontend/parse.c,
	  frontend/parse.h, frontend/timestatus.c,
	  frontend/vorbis_interface.c, lame.c, parse.c, psymodel.c, util.c,
	  util.h, vbrquantize.c:
	  1. timestatus things are moved to frontend(maybe complete ?)
	  2. input file handling is moved to frontend(so the file input API
	  is removed)

2000-09-29 04:51  markt

	* frontend/brhist.c, frontend/get_audio.c, frontend/get_audio.h,
	  mpglib/main.c:
	  removed analysis.h from get_audio.c
	  typo in brhist.c

2000-09-28 23:31  robert

	* lame4dos.bat:
	  improved lame.bat for 4DOS users, donated by Alexander Stumpf
	  <alestrix@unm.edu>

2000-09-28 16:37  takehiro

	* Makefile.in, VbrTag.c, brhist.c, brhist.h, configure,
	  configure.in, encoder.c, frontend, frontend/Makefile.in,
	  frontend/aclocal.m4, frontend/brhist.c, frontend/brhist.h,
	  frontend/configure, frontend/configure.in, frontend/get_audio.c,
	  frontend/get_audio.h, frontend/gpkplotting.c,
	  frontend/gpkplotting.h, frontend/gtkanal.c, frontend/ieeefloat.c,
	  frontend/ieeefloat.h, frontend/lametime.c, frontend/lametime.h,
	  frontend/main.c, frontend/mp3rtp.c, frontend/mp3x.c,
	  frontend/parse.c, frontend/portableio.c, frontend/portableio.h,
	  frontend/rtp.c, frontend/rtp.h, frontend/timestatus.c,
	  frontend/timestatus.h, frontend/vorbis_interface.c, get_audio.c,
	  get_audio.h, gpkplotting.c, gpkplotting.h, gtkanal.c, id3tag.h,
	  ieeefloat.c, ieeefloat.h, include/id3tag.h, include/lame.h,
	  lame.c, lametime.c, lametime.h, main.c, mp3rtp.c, mp3x.c,
	  mpglib/Makefile.in, mpglib/layer3.c, mpglib/mpglib.h, parse.c,
	  portableio.c, portableio.h, psymodel.c, quantize.c,
	  quantize_pvt.c, rtp.c, rtp.h, tables.c, timestatus.c,
	  timestatus.h, util.c, util.h, vbrquantize.c, version.c,
	  version.h, vorbis_interface.c:
	  moved frontend staffs into frontend/
	  Need to debug vorbis/mpglib/analyzer/bitrate histgram.
	  still long way to go...
	  
	  HAVEGTK is changed ANALYSIS(library side) and HAVEGTK(frontend
	  side)
	  
	  BRHIST is deleted from library. all the bitrate histogram works
	  are
	  now in frontend(but not works properly, yet).
	  
	  timestatus things are also moved to frontend.
	  
	  parse.c is now out of library.

2000-09-27 12:06  takehiro

	* gtkanal.h, include/analysis.h:
	  moved gtkanal.h to include/analysis.h

2000-09-27 04:17  florian

	* Makefile.in:
	  fixed make install -> move of lame.h broke it.

2000-09-26 17:47  takehiro

	* Makefile.in, gtkanal.h:
	  more clean up and debug "make clean"

2000-09-26 17:35  takehiro

	* Makefile.in:
	  moved more staff out of libmp3lame

2000-09-26 17:28  takehiro

	* Makefile.in:
	  removed brhist.o etc from library

2000-09-26 17:25  takehiro

	* include, include/lame.h:
	  lame.h moved to include

2000-09-26 17:24  takehiro

	* timestatus.c:
	  little clean up

2000-09-26 17:03  takehiro

	* Makefile.in, configure, configure.in, lame.h:
	  1 moved lame.h into include/
	  2 modularized work with mpglib

2000-09-26 16:59  cisc

	* amiga_mpega.c:
	  Mark, you broke lame_decode_initfile(), what were you thinking?
	  ;) .. anyways, fixed, and added new mp3data struct variables...

2000-09-26 15:48  takehiro

	* mpglib, mpglib/Makefile.in, mpglib/common.c, mpglib/configure,
	  mpglib/configure.in, mpglib/dct64_i386.c, mpglib/decode_i386.c,
	  mpglib/interface.c, mpglib/layer3.c, mpglib/main.c,
	  mpglib/tabinit.c:
	  configure support and more modularized work

2000-09-26 15:47  takehiro

	* mpglib/Makefile:
	  now this is generated by configure

2000-09-26 05:22  markt

	* version.h:
	  CVS is now 3.88 alpha 1

2000-09-26 04:49  markt

	* version.h:
	  updated to 3.87 beta

2000-09-26 04:48  markt

	* get_audio.c, get_audio.h, gtkanal.c, mpglib/main.c, timestatus.c:
	  removed all references to gfc-> from get_audio.c
	  this was needed if we are going to move get_audio.c out of
	  libmp3lame and into the front end code.

2000-09-26 01:00  markt

	* HACKING, STYLEGUIDE:
	  *** empty log message ***

2000-09-26 00:55  markt

	* CodingStyle.c, lametype.h:
	  Removed two files

2000-09-26 00:09  markt

	* amiga_mpega.c, get_audio.c, lame.c, lame.h, timestatus.c,
	  timestatus.h, util.h:
	  when input file was a mp3, it was trashing a bunch of
	  variables (mode, bitrate, etc..)

2000-09-25 21:37  robert

	* Makefile.MSVC:
	  MMX choose table support, needs NASM (can be disabled)

2000-09-25 21:23  florian

	* configure, configure.in:
	  changed default installation prefix to /usr (/usr/local/lib
	  doesn't seem to be in default linker path)

2000-09-25 20:58  florian

	* Makefile.in:
	  fixed shared lib (runtime linker information)

2000-09-25 15:20  markt

	* Makefile.in:
	  removed ?= from makefile

2000-09-25 14:36  shibatch

	* psymodel.c:
	  A small bugfix of --nspsytune.

2000-09-25 02:13  florian

	* Makefile.in, configure, configure.in:
	  cleaner version handling of libmp3lame.so, uninstall target in
	  Makefile

2000-09-25 00:30  markt

	* id3tag.c:
	  fixed genre bug in id3tag.c

2000-09-24 22:59  markt

	* id3tag.c, vbrquantize.c:
	  removed yet another unsigned loop counter

2000-09-24 21:53  robert

	* Makefile.MSVC:
	  enabled Takehiro's IEEE hack

2000-09-24 14:59  robert

	* configure, configure.in:
	  BUG fixed, which prevented using Layer1/2 decoding abilities

2000-09-24 14:34  robert

	* encoder.c, lame.c, quantize.c, quantize_pvt.c, quantize_pvt.h,
	  util.h, vbrquantize.c:
	  code cleanup and very little speedup

2000-09-24 12:38  robert

	* quantize.c:
	  minor RH_AMP tweak, and once again, do not use tabulator within
	  this file

2000-09-24 12:29  robert

	* quantize.c:
	  please do not use tabulators within this file (better in no
	  files)

2000-09-24 12:22  robert

	* VbrTag.c:
	  minor BUG fix for MPEG-2(.5)

2000-09-24 11:39  takehiro

	* quantize.c:
	  sorry, scalefac_scale did not have a bug. that was my compiler
	  bug...

2000-09-24 10:38  afaber

	* mpglib/interface.c:
	  Fixed MP1/2 decoding problem, for files that were encoded with a
	  CRC

2000-09-24 01:32  markt

	* get_audio.c:
	  minor fix to mp3 header parsing

2000-09-23 23:36  afaber

	* bitstream.c:
	  removed unused variable bits

2000-09-23 23:36  afaber

	* id3tag.c:
	  fixed signed/unsigned mismatch, added bitstream.h include file

2000-09-23 22:31  markt

	* VbrTag.c, bitstream.c, bitstream.h, id3tag.c, id3tag.h, lame.c,
	  lame.h, main.c, parse.c:
	  id3v1 and id3v2 tags now get written directly into
	  the bitstream instead of into the output file.
	  
	  The only file I/O left in libmp3lame will be the VBR tag.

2000-09-23 21:13  markt

	* get_audio.c, lame.h, mpglib/main.c:
	  removed lame_decode_fromfile() calls from the API,
	  moved them into get_audio.c

2000-09-23 20:08  markt

	* lame.h, mpglib/main.c:
	  some mpglib restructering.
	  All mp3 routines which work with files now only make
	  calls to lame_decode_* that are in the API.

2000-09-23 14:48  markt

	* VbrTag.c:
	  bug in Xing header for MPEG2.5 fixed

2000-09-23 09:07  takehiro

	* quantize.c, quantize_pvt.c, util.c, util.h:
	  workaround for subblock_gain and scalefac_scale artifact.
	  I hope this will fix -q1 problems.

2000-09-23 02:11  markt

	* version.c:
	  *** empty log message ***

2000-09-23 01:23  markt

	* mpglib/main.c:
	  *** empty log message ***

2000-09-23 01:23  markt

	* VbrTag.c, VbrTag.h, lame.h, parse.c, util.c, util.h:
	  bug fixed in decoding: if syncword found in Xing VBR header toc,
	  it would confuse things.
	  
	  Also fixed overflow in vbrtag.c, and changed size of xing vbr
	  header to 128kbs. This will make non-vbr aware players
	  give a more reasonable playing time estimate.

2000-09-23 00:51  robert

	* psymodel.c:
	  minor fix

2000-09-22 22:54  shibatch

	* quantize.c, quantize_pvt.c:
	  Minor bugfixes of --nspsytune. Frame analyzer works correctly
	  with --nspsytune.

2000-09-21 23:55  shibatch

	* parse.c, psymodel.c, quantize.c:
	  Tweaks to --nspsytune. This improves CBR quality.

2000-09-21 16:50  afaber

	* Dll/LameDLLInterface.htm:
	  Updated lame_enc.dll documentation file

2000-09-20 22:56  takehiro

	* Makefile.in, configure, configure.in, encoder.c, lame.c,
	  timestatus.c:
	  more configure updates.
	  not compiling/linking BRHIST routine when it is disabled.

2000-09-20 20:50  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h, Dll/Example.cpp:
	  Added bNoRes option in LHV1 structure, updated example to have
	  default settings as in testcase.wav test case

2000-09-20 19:08  afaber

	* brhist.c, lame.h:
	  Changed BRHST un-signed variables to signed variables

2000-09-20 18:50  afaber

	* Dll/BladeMP3EncDLL.c:
	  High quality was not enabled

2000-09-20 18:49  afaber

	* Dll/MP3EncDll.dsp:
	  Added TAKEHIRO_IEEE754_HACK define to project file

2000-09-20 18:41  afaber

	* lame.dsp:
	  Added TAKEHIRO_IEEE754_HACK define to project file

2000-09-20 16:40  robert

	* quantize.c:
	  VBR old quality tuning for --raise-smr

2000-09-19 14:40  takehiro

	* Makefile.in:
	  update glibc inline math error detection

2000-09-19 14:32  takehiro

	* configure, configure.in:
	  oops, needless gabage removed

2000-09-19 14:26  takehiro

	* configure:
	  newer configure script made by configure.in

2000-09-19 14:12  takehiro

	* Makefile.in, configure.in:
	  not compiling/linking mpglib code when it disabled.
	  not compiling/linking gtk related code when it disabled.
	  
	  control LAYER1 and LAYER2 decoding function with
	  --enable-decode-layer1 and
	  --enable-decode-layer2
	  
	  autodetect math inline bug of glibc < 2.1.3

2000-09-19 14:11  robert

	* quantize_pvt.c:
	  modified --ATHlower to not touch sfb21

2000-09-19 14:06  takehiro

	* .:
	  ignore files related configure script and library itself

2000-09-19 05:33  markt

	* mpglib/main.c:
	  got rid of a size_t in main.c

2000-09-19 04:54  markt

	* mpglib/main.c:
	  removed unsigned int being used as a loop counter.

2000-09-19 04:18  markt

	* encoder.h, lame.c, util.c, util.h:
	  updated filter settings

2000-09-18 22:29  afaber

	* Dll/Example.cpp:
	  Bug fixes in example file

2000-09-18 21:57  markt

	* testcase.mp3, util.c, util.h:
	  changed BPC to 160 on Frank's advice.
	  This means pre-compute 160 filter-sinc-windows, instead of just
	  16.
	  Should also fix the aliasing problem shown in mp3.com bulliten
	  board

2000-09-18 21:30  markt

	* Makefile.B32, Makefile.DJGPP, Makefile.in, Makefile.unix:
	  Makefiles updated to reflect that all the ASM code has been
	  removed and replaced with IEEE754_HACK. I'm glad to see
	  it gone - but I didn't remove it!

2000-09-18 21:28  markt

	* STYLEGUIDE, VbrTag.c, brhist.c, get_audio.c, machine.h,
	  quantize.c, quantize_pvt.c, takehiro.c, vorbis_interface.c:
	  minor code cleanup.

2000-09-18 20:53  afaber

	* lame.dsp:
	  Fixed library problem in ReleaseGTK mode

2000-09-18 20:09  markt

	* config.log:
	  removed config.log from repository

2000-09-18 19:43  afaber

	* brhist.c:
	  Fixes for Cygwin compiler

2000-09-18 18:52  markt

	* INSTALL, Makefile.in, Makefile.unix, config.log:
	  more configure updates

2000-09-18 18:34  markt

	* INSTALL, INSTALL.configure, Makefile, Makefile.in, Makefile.unix,
	  aclocal.m4, confdefs.h, config.guess, config.log, config.sub,
	  configure, configure.in, install-sh:
	  added configure stuff from Florian!
	  
	  original Makefile is now in Makefile.unix

2000-09-18 01:48  cisc

	* util.c, vbrquantize.c, version.h:
	  warning fixes & bumped alpha version

2000-09-17 23:25  markt

	* Makefile, lame.h:
	  added note about shared library

2000-09-17 22:25  markt

	* USAGE, encoder.c, lame.c, lame.h, parse.c:
	  --scale option added

2000-09-17 21:47  afaber

	* bitstream.c, get_audio.c, quantize.c:
	  removed unecessary (int) casts

2000-09-17 21:21  markt

	* HACKING, STYLEGUIDE, lame.h:
	  some editing

2000-09-17 21:04  afaber

	* encoder.c, lame.dsp:
	  Enabled BRHIST in MSVC project file, and changed calls to brhist
	  functions in encoder.c

2000-09-17 20:55  markt

	* encoder.c, takehiro.c:
	  forgot to add encoder.c to cvs

2000-09-17 20:53  markt

	* bitstream.c, machine.h, util.c, util.h:
	  more 'unsigned's removed

2000-09-17 20:52  afaber

	* brhist.c, brhist.h, lame.c, lame.h, timestatus.c:
	  Moved BRHIST variables to lame_global_flags structure, so a
	  library (like lame_enc.dll) can have access to the histogramming
	  data

2000-09-17 20:20  markt

	* lame.c, testcase.mp3, version.c:
	  resampling nolonger needs to quantize back to integers
	  since internal representation of PCM samples is now floating
	  point

2000-09-17 19:50  markt

	* Makefile, fft.c, gtkanal.c, lame.c, machine.h, mpglib/main.c,
	  newmdct.c, newmdct.h, psymodel.c, psymodel.h, testcase.mp3,
	  util.c, util.h:
	  sample_t changes updated, sample_t = FLOAT

2000-09-17 18:52  markt

	* Dll/MP3EncDll.dsp, INSTALL, Makefile, Makefile.B32,
	  Makefile.DJGPP, Makefile.MSVC, encoder.h, lame.c, lame.dsp,
	  lametime.h, main.c, timestatus.h:
	  added encoder.c which has the core encoding function.
	  fixed sample_t stuff

2000-09-17 17:55  robert

	* Makefile, l3side.h, quantize.c, quantize_pvt.c, takehiro.c:
	  not everyone has access to Vorbis in Mark's home directory ;-)

2000-09-17 17:54  afaber

	* mpglib/main.c:
	  Changed return value from zero to -1 when EOF has been reached

2000-09-17 16:19  takehiro

	* quantize.c:
	  added #include <math.h>

2000-09-17 10:07  takehiro

	* get_audio.c:
	  trivial warning fix

2000-09-17 09:32  takehiro

	* quantize_pvt.c:
	  removed old assembler code

2000-09-17 09:30  takehiro

	* main.c:
	  quick hack for "sample_t is not defined"

2000-09-17 05:00  cisc

	* get_audio.h, version.c:
	  updated some libsndfile related stuff

2000-09-17 04:19  cisc

	* VbrTag.h, bitstream.h, brhist.h, encoder.h, fft.h, get_audio.h,
	  gpkplotting.h, gtkanal.h, id3tag.h, ieeefloat.h, l3side.h,
	  lame.h, lametime.h, lametype.h, machine.h, newmdct.h,
	  portableio.h, psymodel.h, quantize.h, quantize_pvt.h,
	  reservoir.h, rtp.h, tables.h, timestatus.h, util.h, version.h:
	  conformed all this-is-included-defines to match
	  'project_file_name' style

2000-09-16 22:52  afaber

	* lame.dsp, reservoir.c, reservoir.h, tables.c, tables.h,
	  timestatus.c, timestatus.h, util.c, util.h, vbrquantize.c:
	  Changed lame header/source to LGPL license where necessary, added
	  LGPL header to files when applicable

2000-09-16 22:39  afaber

	* Dll/MP3EncDll.dsp, encoder.h, fft.h, get_audio.c, get_audio.h,
	  gpkplotting.c, gpkplotting.h, gtkanal.c, gtkanal.h, l3side.h,
	  lametime.c, lametime.h, newmdct.h, parse.c, psymodel.h,
	  quantize.c, quantize.h, quantize_pvt.c, quantize_pvt.h:
	  Changed lame header/source to LGPL license where necessary, added
	  LGPL header to files when applicable

2000-09-16 22:16  afaber

	* fft.c, machine.h, psymodel.c:
	  Removed the float/double compiler warnings

2000-09-16 22:08  afaber

	* Dll/BladeMP3EncDLL.h, Dll/MP3EncDll.dsp:
	  Small changes to the lame_enc project settings

2000-09-16 21:52  afaber

	* brhist.c, brhist.h, get_audio.c, gtkanal.c, lametime.h, main.c,
	  mpglib/layer2.c, mpglib/layer3.c, mpglib/main.c, psymodel.c,
	  quantize.c, takehiro.c, timestatus.c, vbrquantize.c:
	  Removed a bunch of signed/unsigned compiler warnings
	  and removed const/no const assignements compiler warnings
	  welcome to const hell Frank!

2000-09-16 21:07  markt

	* Makefile, quantize_pvt.c:
	  shared lib support, IEEE stuff the default on i386

2000-09-16 21:07  cisc

	* amiga_mpega.c:
	  added mode & mode_ext support to mp3data struct, like
	  mpglib/main.c

2000-09-16 21:06  afaber

	* lame.dsp:
	  Added lametime.c and lametime.h files to project

2000-09-16 19:03  markt

	* HACKING, STYLEGUIDE:
	  style updates

2000-09-16 18:54  markt

	* STYLEGUIDE:
	  adding STYLEGUIDE

2000-09-16 18:54  markt

	* CodingStyle.c, LICENSE, USAGE, bitstream.c, get_audio.c,
	  gtkanal.c, id3tag.c, l3side.h, lame.c, lame.h, lametime.c,
	  machine.h, mpglib/common.h, mpglib/interface.h, mpglib/main.c,
	  parse.c, quantize.c, quantize_pvt.c, quantize_pvt.h, tables.c,
	  tables.h, takehiro.c, testcase.mp3, timestatus.c, util.h,
	  vbrquantize.c, version.c:
	  got rid of a bunch of unsigned variables.

2000-09-16 13:59  afaber

	* brhist.c, psymodel.c, quantize_pvt.c, takehiro.c, vbrquantize.c:
	  Removed the signed/unsigned mismatch compiler warnings

2000-09-16 13:30  afaber

	* bitstream.c:
	  Removed the signed/unsigned mismatch compiler warnings

2000-09-16 12:53  afaber

	* mpglib/layer1.c:
	  Switched include from mpg123.h to common.h to avoid compiler
	  warnings

2000-09-16 00:43  cisc

	* psymodel.c, timestatus.c:
	  warning fixes

2000-09-15 16:36  shibatch

	* psymodel.c:
	  Tweaks to --nspsytune.

2000-09-15 15:01  takehiro

	* quantize_pvt.c:
	  oops, foolish bug ...

2000-09-15 13:03  takehiro

	* quantize_pvt.c:
	  TAKEHIRO_IEEE754_HACK now C99 aliasing rule compliant.
	  it runs correctly with "-fstrict-aliasing" (GCC)

2000-09-14 03:32  markt

	* brhist.c, mpglib/mpg123.h:
	  fixed VBR display

2000-09-13 23:31  markt

	* Makefile, quantize.c, reservoir.c:
	  NORES_TEST turned back of

2000-09-13 18:46  afaber

	* Dll/BladeMP3EncDLL.c:
	  Fixed bug in lame_enc.dll when using preset option, and return
	  proper frame size when using MPEG-2

2000-09-13 18:24  afaber

	* mpglib/common.c, mpglib/common.h, mpglib/interface.c,
	  mpglib/interface.h, mpglib/layer2.c, mpglib/layer3.c,
	  mpglib/main.c, mpglib/mpg123.h, mpglib/mpglib.dsp,
	  mpglib/mpglib.h:
	  Added common.h and interface.h, and added the functions
	  prototypes that should be in these files, instead of mpg123.h

2000-09-13 18:22  markt

	* util.c:
	  precompute for filter turned back on

2000-09-13 18:19  markt

	* lame.c, util.c, util.h:
	  replaced upsampling filter with 19 point blackman filter

2000-09-13 11:18  markt

	* parse.c:
	  typo in documentation

2000-09-13 11:03  markt

	* Makefile, parse.c, testcase.mp3, vorbis_interface.c:
	  fixed vorbis_interface.c so it would compile

2000-09-13 10:59  markt

	* Makefile, bitstream.c, fft.c, get_audio.c, lame.c, lame.h,
	  psymodel.c:
	  fixed some bugs introduced by frank.
	  get_audio.c must count samples for .wav files.
	  lame.c: disabling ATH is not the way to solve the problem Frank
	  wants to solve

2000-09-13 09:23  robert

	* quantize.c:
	  quality adjustment for VBR old to better use --raise-smr

2000-09-12 22:36  pfk

	* brhist.c, gtkanal.c, id3tag.c, lame.c, lametype.h, main.c,
	  mpglib/common.c, mpglib/huffman.h, mpglib/l2tables.h,
	  mpglib/layer2.c, mpglib/layer3.c, mpglib/main.c, mpglib/mpg123.h,
	  mpglib/tabinit.c, psymodel.c, quantize.c, quantize_pvt.c,
	  tables.c, tables.h, takehiro.c, util.c, util.h, vbrquantize.c,
	  vorbis_interface.c:
	  added const to all constant tables, so that they are placed in
	  write
	  protected RAM or in ROM (standalone DSP version).
	  
	  The SEGFAULT be with you (instead of debugging for hours).
	  
	  Also some remarks instead of changes instead of code changes.
	  
	  Remove them if they are fully useless.

2000-09-12 20:47  robert

	* Makefile, lame.c, lame.h, parse.c, psymodel.c, quantize_pvt.c:
	  added --raise-smr <0..1> a new toy to control quality

2000-09-12 18:20  pfk

	* Makefile, Makefile.B32, Makefile.DJGPP, Makefile.MSVC,
	  bitstream.c, get_audio.c, lame.c, lame.h, lametype.h,
	  mpglib/main.c, parse.c, util.h, version.h, vorbis_interface.c:
	  Some bug fixes
	  Activated lametime.c
	  Two types of display update now possible (old and new, depends on
	  Makefile)

2000-09-12 06:26  pfk

	* brhist.c, lame.c, lametype.h, version.c:
	  brhist.c Changed display of percentages between 0.1 and 1 %
	  
	  lame.c: ???
	  version.h: include as less as possible
	  
	  lametype.h: First try to disentangle the #include jungle of lame
	  (this style would be the immediate death for larger
	  projects)

2000-09-12 00:47  markt

	* main.c, mpglib/l2tables.h, mpglib/layer2.h:
	  al_table struct renamed al_table2

2000-09-11 20:33  shibatch

	* psymodel.c:
	  --nspsytune uses additive masking.

2000-09-11 20:05  robert

	* psymodel.c:
	  disabled "additive masking" in subband calculation

2000-09-11 20:04  robert

	* lame.c, lame.h, parse.c, quantize.c, quantize.h, quantize_pvt.c:
	  experimental mix of VBR new and old available with --vbr-mtrh

2000-09-10 23:24  markt

	* HACKING:
	  added rule #1 LAME style guide

2000-09-10 22:13  markt

	* doc/html/modes.html:
	  modes.html was not in the repository

2000-09-10 21:45  markt

	* testcase.mp3, testcase.wav:
	  made the test case a little longer
	  (100k .wav file)

2000-09-10 21:27  markt

	* lame.h, mpglib/huffman.h, mpglib/l2tables.h, mpglib/layer2.h,
	  parse.c:
	  reverted mpglib routines back to original.
	  please keep mpglib as close as possible to the mpg123/mpglib
	  code that it is based on.
	  
	  Went back to display every 100 frames, which I prefer.

2000-09-10 20:54  markt

	* Makefile, lame.c, timestatus.c:
	  went back to display every 100 frames.

2000-09-10 19:54  cisc

	* brhist.c:
	  small adjustment

2000-09-10 19:52  cisc

	* version.c:
	  correct includes for libsndfile prototypes

2000-09-10 18:05  pfk

	* Makefile, USAGE, brhist.c, brhist.h, fft.c, fft.h, lame.c,
	  lame.h, mpglib/huffman.h, version.h:
	  some minor changes. Start to introduce sample_t.
	  
	  Changed VBR display.

2000-09-10 13:11  pfk

	* fft.c, mpglib/l2tables.h, mpglib/layer2.h, parse.c, psymodel.c,
	  version.h:
	  fft.c: loop variables short => size_t/int
	  layer2.h
	  l2tables.h: formated, structure definition changed for (possible)
	  speedup
	  parse.c optimized sfb21 usage for presets

2000-09-09 23:00  pfk

	* get_audio.c, gtkanal.c, lametime.c, lametime.h, mpglib/common.c,
	  mpglib/l2tables.h, mpglib/layer2.c, mpglib/layer2.h,
	  mpglib/mpg123.h:
	  struct al_table {} <=>
	  typedef struct {} al_table conflict removed
	  
	  lame/gtkanal.c: typecast of functions ptr's

2000-09-09 21:11  pfk

	* CodingStyle.c, Makefile, USAGE, lametime.c, lametime.h, mlame,
	  mlame_corr.c, mpglib/layer1.c, mpglib/layer2.c, quantize_pvt.c,
	  quantize_pvt.h, vbrquantize.c, version.c, version.h:
	  Makefile: added option -pedantic
	  USAGE: updated some remarks,
	  added some remarks,
	  kbs => kbps, HZ/hz => Hz, space between number and unit
	  mlame: -mf => -mj
	  uses mlame_corr
	  mlame_corr: First try of a program which analyzes the total file
	  and recommend flags
	  quantize_pvt.*: made some arrays const
	  asm => __asm__
	  version.* copied vom pfk1 tree
	  mpglib/*.c added newline at the end of the file
	  
	  CodingStyle.c start of an CodingStyle Guide, pre-alpha
	  lametime.* heavy system specific code should moved to this
	  file, should not mixed with other code

2000-09-09 21:04  cisc

	* amiga_mpega.c:
	  warning fix

2000-09-09 14:41  robert

	* bitstream.c:
	  fixed debugging code

2000-09-09 11:40  takehiro

	* newmdct.c:
	  sorry, back to old...

2000-09-09 11:22  robert

	* quantize.c:
	  BUG fix for possible endless loop in RH_AMP code

2000-09-09 10:59  takehiro

	* newmdct.c:
	  bit faster mdct.
	  there's some difference caused by rounding problem, but i think
	  this is OK.

2000-09-09 10:34  takehiro

	* bitstream.c:
	  bit faster putbits

2000-09-09 10:28  takehiro

	* bitstream.c:
	  small fix of debugging code

2000-09-08 19:28  robert

	* quantize.c:
	  to combine VBR_rh with VBR_mt define RH_VBR_MTRH at compile time

2000-09-08 16:05  robert

	* quantize.c, quantize_pvt.c, quantize_pvt.h:
	  merging branches pfk1 with main

2000-09-08 14:55  robert

	* gtkanal.h, quantize.c, quantize_pvt.c, quantize_pvt.h,
	  vbrquantize.c:
	  set_pinfo calcs noise and masking of its own

2000-09-08 12:17  robert

	* parse.c, quantize.c, quantize_pvt.c, quantize_pvt.h, takehiro.c,
	  vbrquantize.c:
	  substitution of lame_global_flags with lame_internal_flags

2000-09-08 11:22  robert

	* parse.c:
	  Bug fix for automatic file type recognition back into MAIN branch

2000-09-08 10:04  shibatch

	* psymodel.c:
	  A minor bugfix of mask_add().

2000-09-07 23:37  pfk

	* mlame_corr.c:
	  file to use with mlame to select different modes by pre analysing
	  wav files.

2000-09-07 18:06  afaber

	* mpglib/mpglib.dsp:
	  Cleanup project file a bit

2000-09-07 18:05  afaber

	* README.WINGTK:
	  Small modification to the instruction, in order to compile/link
	  with latest GTK/GDK libs

2000-09-07 18:04  afaber

	* lame.dsp:
	  Changed project file to work with latest GTK/GDK libs

2000-09-07 17:40  afaber

	* Dll/BladeMP3EncDLL.c:
	  Fixed problem in VBR WriteTag function (it did not compile, MP3
	  file stream was not closed properly, wrong file name was used)

2000-09-07 12:26  shibatch

	* psymodel.c, quantize_pvt.c:
	  Tweaks to --nspsytune.

2000-09-06 07:50  shibatch

	* psymodel.c, quantize_pvt.c:
	  updated --nspsytune.

2000-09-06 07:15  markt

	* Makefile, amiga_mpega.c, bitstream.c, brhist.c, fft.c, fft.h,
	  get_audio.c, get_audio.h, gtkanal.c, gtkanal.h, lame.c, lame.h,
	  main.c, mp3rtp.c, mpglib/common.c, mpglib/decode_i386.c,
	  mpglib/huffman.h, mpglib/interface.c, mpglib/l2tables.h,
	  mpglib/layer1.c, mpglib/layer2.c, mpglib/layer3.c, mpglib/main.c,
	  mpglib/mpg123.h, newmdct.c, newmdct.h, parse.c, psymodel.c,
	  psymodel.h, quantize.c, quantize_pvt.c, quantize_pvt.h,
	  reservoir.c, tables.c, tables.h, takehiro.c, timestatus.c,
	  util.c, util.h, vbrquantize.c, version.c, version.h,
	  vorbis_interface.c:
	  LAME CVS reverted back to Aug 30 version.
	  
	  code before this can be checked out with:
	  cvs update -r pfk1

2000-09-05 23:55  pfk

	* HACKING, Makefile, TODO, get_audio.c, lame.c, lame.h, mlame,
	  mpglib/README, mpglib/huffman.h, mpglib/layer1.c,
	  mpglib/layer2.c, parse.c, version.h:
	  HACKING: add type proposals
	  Makefile: layer 1 enabled
	  mlame: flaw removed
	  get_audio: C-Linkage problem solved
	  layer1.c: bug fixed
	  disabled decoding enabled, works for C/C++
	  huffman.c: read protect/scope protect tables using
	  'const'/'static'
	  lame.h: Starting RPC support, 3 macros IN/OUT/INOUT defined
	  Starting supported for szip compressed wave/aiff
	  parse.c: detection of raw/wav/aiff input (proposal of RH)
	  boolean bug fixed !=||!=||!=
	  layer2.c: common layer1/layer2 buffer moved from layer2.c to
	  layer1.c
	  lame.c: bug fixed: -V9 switched to mono
	  Auto switch to mono reported wrong compression ratio
	  report of forced joint stereo usage
	  
	  *: Some spelling errors removed

2000-09-05 22:20  robert

	* quantize.c:
	  more modularization and clean-up, some few more to come

2000-09-05 20:39  shibatch

	* psymodel.c:
	  Tweaks to --nspsytune(table optimization).

2000-09-05 16:36  cisc

	* mpglib/common.c, mpglib/interface.c, mpglib/l2tables.h,
	  mpglib/layer1.c, mpglib/layer2.c, mpglib/layer3.c, parse.c:
	  quick fix to make mpglib compile, prolly still broken though.
	  fixed mono bug in layer1.c?

2000-09-04 23:17  shibatch

	* psymodel.c:
	  Tweaks to --nspsytune.

2000-09-04 19:56  robert

	* quantize.c:
	  more cleanup in VBR_rh

2000-09-04 16:05  robert

	* quantize.c:
	  more modularization work
	  
	  it is now possible to combine VBR_RH with VBR_MT

2000-09-03 22:33  cisc

	* brhist.c:
	  small change

2000-09-03 19:10  robert

	* quantize.c:
	  updated remarks

2000-09-03 17:21  pfk

	* get_audio.c, gtkanal.c, lame.h, main.c, mp3rtp.c, parse.c,
	  util.c, util.h, version.h:
	  some bugfixes, move some functionality to two simple functions.
	  
	  Added support of 24 and 32 bit BE and LE AIFF and WAV files
	  (simple support, only using upper 16 bit)
	  
	  marking 3 areas which seems to be buggy.

2000-09-03 17:19  robert

	* quantize.c, quantize_pvt.c, quantize_pvt.h, takehiro.c,
	  vbrquantize.c:
	  code cleanup in quantize_pvt.c

2000-09-03 15:06  robert

	* lame.c, quantize.c, quantize_pvt.c, quantize_pvt.h, takehiro.c,
	  vbrquantize.c:
	  takehiro.c now free of lame_global_flags

2000-09-03 14:05  robert

	* quantize.c, quantize_pvt.c, quantize_pvt.h:
	  code clean up

2000-09-03 13:06  pfk

	* Makefile, brhist.c, get_audio.c, get_audio.h, gtkanal.c, lame.c,
	  lame.h, main.c, version.c, version.h:
	  minor fixes
	  
	  get_audio: (simple) support of 24 bit PCM
	  Replaced the different types of tabs by 3, 4 or 8 spaces

2000-09-03 12:42  shibatch

	* psymodel.c:
	  Tweaks to --nspsytune.

2000-09-03 11:47  shibatch

	* psymodel.c, quantize_pvt.c:
	  Tweaks to --nspsytune. Noise calculation using tonality is
	  canceled.

2000-09-03 09:54  robert

	* tables.c:
	  fix for i386/choose_table.nas, could not be linked

2000-09-03 06:50  cisc

	* lame.c:
	  removed *extremely* annoying debug. fixed related bug? should be
	  checked if mono encoding still works!

2000-09-03 06:48  cisc

	* get_audio.c:
	  fixed conflicting prototype when LIBSNDFILE is defined

2000-09-02 18:06  robert

	* quantize_pvt.c:
	  small fix to previous check in

2000-09-02 18:03  robert

	* gtkanal.h, quantize.c, quantize_pvt.c, quantize_pvt.h,
	  vbrquantize.c:
	  set_pinfo does now the noise calculation for the frame analyzer
	  code cleanup

2000-09-02 15:36  pfk

	* mlame:
	  extended functionality and feature fix

2000-09-02 13:36  pfk

	* main.c:
	  bug fix.

2000-09-02 13:28  pfk

	* get_audio.c, main.c:
	  Bug fix in final fwrite().

2000-09-02 06:38  cisc

	* util.h:
	  warning and error fixes

2000-09-02 05:42  cisc

	* mpglib/mpg123.h:
	  It's important to remember the PARENT_IS_SLASH define when
	  including files from previous dir.

2000-09-02 05:29  cisc

	* amiga_mpega.c, lame.c, util.c:
	  warning and error fixes

2000-09-02 03:13  pfk

	* Makefile, get_audio.c, gtkanal.c, lame.c, lame.spec,
	  mpglib/l2tables.h, mpglib/layer2.c, mpglib/mpg123.h, psymodel.h,
	  timestatus.c, util.c, util.h, vbrquantize.c, version.c,
	  version.h:
	  Quick and Dirty: Now compilable with g++.
	  
	  Layer 1 and 2 decoding makes problems.
	  
	  structs and typedef structs with the same name.
	  Must be patched.

2000-09-02 01:03  cisc

	* psymodel.c, quantize_pvt.c, timestatus.c:
	  warning fixes

2000-09-01 23:28  pfk

	* Makefile, lame.c, lame.h, mpglib/layer1.c, mpglib/layer2.c,
	  util.c, util.h:
	  Bugfix for 9 kbps error

2000-09-01 23:28  robert

	* get_audio.c:
	  small typo in remark

2000-09-01 20:21  pfk

	* bitstream.c, get_audio.c, gtkanal.c, gtkanal.h, lame.c, lame.h,
	  mpglib/common.c, mpglib/interface.c, mpglib/layer1.c,
	  mpglib/layer2.c, mpglib/layer3.c, mpglib/main.c, mpglib/mpg123.h,
	  newmdct.c, psymodel.c, quantize.c, quantize_pvt.c, reservoir.c,
	  tables.c, tables.h, util.c, util.h, vbrquantize.c, version.h:
	  * A lot of minor changes
	  
	  * Changed the name of structure element from 'stereo' to
	  'channels'
	  if it has the meaning of 'channels'. This is so confusing, so
	  that
	  there are some bugs in lame.
	  
	  Rest read in 'Bad and worse programming styles' and 'job
	  security'
	  
	  --
	  Frank Klemm

2000-09-01 16:19  robert

	* lame.c:
	  small fix, lame -v x.wav did not work anymore

2000-09-01 04:04  cisc

	* amiga_mpega.c:
	  some warning and error fixes

2000-09-01 04:03  cisc

	* LICENSE:
	  libsndfile is now LGPL

2000-09-01 00:05  pfk

	* brhist.c, lame.c, parse.c, quantize.c, util.c, util.h, version.c:
	  presets modified
	  changed name of bitrate_table to index_to_bitrate
	  added table bitrate_to_index, speedup of functions using it.

2000-08-31 21:33  pfk

	* Makefile, amiga_mpega.c, fft.c, fft.h, get_audio.c, get_audio.h,
	  gtkanal.c, lame.c, lame.h, main.c, mp3rtp.c,
	  mpglib/decode_i386.c, mpglib/main.c, newmdct.c, newmdct.h,
	  psymodel.c, psymodel.h, quantize_pvt.c, util.c, util.h,
	  version.c, version.h, vorbis_interface.c:
	  Introduction of the type sample_t for PCM samples.
	  'short' should not be used in the future.

2000-08-31 11:06  shibatch

	* psymodel.c:
	  minor bugfix of mask_add().

2000-08-30 23:29  pfk

	* LICENSE, PRESETS.draft, bitstream.c, lame.c, parse.c,
	  timestatus.c, util.h, version.c, version.h:
	  *** empty log message ***

2000-08-30 07:18  shibatch

	* quantize_pvt.c:
	  Tweaks to --nspsytune.

2000-08-29 22:35  pfk

	* debugscalefac.c, doc/html/switchs.html, get_audio.c, id3tag.c,
	  lame.c, lame.h, parse.c, util.c, util.h, version.c, version.h:
	  *** empty log message ***

2000-08-29 10:14  shibatch

	* psymodel.c, quantize_pvt.c:
	  Tweakings to --nspsytune.
	  Tonality table is optimized a little.
	  Noise calculation now uses tonality.
	  NSATHSCALE is decreased from 103 to 100.

2000-08-28 20:18  robert

	* lame.c:
	  small tweak to RH_VALIDATE_MS

2000-08-28 19:55  robert

	* lame.h, parse.c, psymodel.c:
	  small fixes

2000-08-28 19:27  shibatch

	* psymodel.c:
	  Tweaks to the tonality table. vbrtest.wav is now encoded
	  correctly with --nspsytune.

2000-08-28 18:37  pfk

	* lame.h, parse.c, version.c, version.h:
	  parse.c: presets are taken from a table
	  version.c: version incremented (should be done more often)

2000-08-28 10:04  shibatch

	* parse.c, psymodel.c, quantize_pvt.c:
	  Changes to --nspsytune.

2000-08-27 11:51  robert

	* lame.c:
	  tweak to validate MS switching criterion (to be enabled with
	  RH_VALIDATE_MS)

2000-08-27 04:34  markt

	* parse.c:
	  look for .mp3 or .MP3 when parsing input filename

2000-08-26 04:07  markt

	* Makefile:
	  openBSD tweak

2000-08-26 04:03  markt

	* Dll/BladeMP3EncDLL.h:
	  added include sys/types to BlaceMP3EncDLL.h
	  Borland compiler needs definitions for BOOL and DWORD.

2000-08-24 20:34  robert

	* quantize.c:
	  little tweak to VBR-old, side channel gets now a little lower bit
	  skeleton

2000-08-23 23:08  robert

	* Makefile, lame.c, quantize.c, quantize_pvt.c:
	  sfb21 kludge, only VBR tries to get an undistorted sfb21
	  at lower sample frequencies it makes more trouble than
	  we can gain from that, so now sfb21 will be ignored in
	  VBR mode for MPEG-2 LSF.
	  
	  first tweak (not enabled) for M/S switching criterion

2000-08-23 00:02  markt

	* lame.c, vorbis_interface.c:
	  messages about filters, vorbis modes

2000-08-22 18:29  markt

	* lame.c, util.h:
	  moved last_time into lame_internal_flags struct

2000-08-22 16:30  markt

	* get_audio.c:
	  byte swapping code now only used for 16 bit input files.

2000-08-22 03:19  markt

	* lame.c:
	  minor gtkflag changes

2000-08-22 00:08  pfk

	* get_audio.c, parse.c:
	  get_audio.c:
	  Support of reporting all supported input file types.

2000-08-21 20:16  robert

	* lame.c:
	  pe_MS was not initialized in case we don't use our psychoacoustic
	  model

2000-08-21 16:02  robert

	* lame.c, mpglib/main.c:
	  LayerI+II LSF corrections

2000-08-21 15:47  markt

	* doc/html/contributors.html, doc/html/history.html,
	  doc/html/id3.html, doc/html/index.html, doc/html/switchs.html:
	  updated docs from Gabriel

2000-08-21 05:32  markt

	* USAGE, vorbis_interface.c:
	  added support for the four new vorbis modes. (higher bitrates)
	  use with the -b option.

2000-08-21 04:52  markt

	* Dll/BladeMP3EncDLL.c, main.c, mp3rtp.c:
	  Fixed problem with writing VBR tag if user renames mp3 output
	  file during the encoding.

2000-08-21 00:40  markt

	* lame.c, lame.h, main.c, mp3rtp.c:
	  more prep work for vbr tag problem

2000-08-21 00:28  markt

	* VbrTag.c, VbrTag.h, lame.c:
	  prep work to fix Vbrtag when user renames mp3 file during
	  encodiing.

2000-08-21 00:05  markt

	* main.c, testcase.mp3:
	  fixes to main.c: it is supposed to be example code for how to
	  use the LAME library - it should only include lame.h, all other
	  .h files are supposed to be private to the encoding library.

2000-08-21 00:00  markt

	* Dll/MP3EncDll.dsp, INSTALL, Makefile, Makefile.B32,
	  Makefile.DJGPP, Makefile.MSVC, README.B32, TODO, bitstream.c,
	  gtkanal.c, lame.c, lame.dsp, quantize-pvt.c, quantize-pvt.h,
	  quantize.c, quantize_pvt.c, quantize_pvt.h, takehiro.c,
	  vbrquantize.c:
	  changed quantize-pvt.c to quantize_pvt.c
	  Added Boland stuff_

2000-08-20 23:21  markt

	* Makefile, version.h:
	  Amiga stuff for Makefile, updated version.h to 3.87alpha

2000-08-20 21:22  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  enhanced lame_enc dll debugging

2000-08-20 20:50  robert

	* lame.c:
	  fixed bug for LSF,
	  ms_ratio will not be calculated for the second granule
	  because at lower sample rates there is no second granule
	  but the M/S switching criterion relies ms_ratio on it

2000-08-19 10:54  robert

	* lame.c, mpglib/main.c, parse.c:
	  total frames of LayerI files was not correctly estimated for
	  decoding

2000-08-18 18:07  robert

	* USAGE, get_audio.c, gtkanal.c, lame.c, lame.h, mpglib/main.c,
	  parse.c:
	  LayerI+II decoding patches

2000-08-18 01:40  pfk

	* bitstream.c, brhist.c, lame.c, main.c, parse.c, timestatus.c,
	  timestatus.h:
	  bitstream.c: Table based CRC calculation code added (must be
	  enabled to use)
	  brhist.c: Enlighted design of the bit rate history output
	  main.c: init of CRC Table in bitstream.c
	  parse.c: Some parameters can be name in Hz/bps and in kHz/kbps
	  timestatus.c: Enlighted design of the time counters
	  lame.c: Changed screen update frequency from 50/100 frames to 2
	  seconds

2000-08-17 14:34  robert

	* Makefile, Makefile.DJGPP, Makefile.MSVC, mpglib/common.c:
	  Layer1/2 related update

2000-08-16 17:05  robert

	* Makefile.MSVC:
	  added GTK support (frame analyzer)

2000-08-16 01:22  afaber

	* mpglib/common.c, mpglib/interface.c, mpglib/l2tables.h,
	  mpglib/layer1.c, mpglib/layer1.h, mpglib/layer2.c,
	  mpglib/layer2.h, mpglib/mpg123.h, mpglib/mpglib.dsp, parse.c:
	  Added support for Layer 1 and Layer 2 decoding

2000-08-15 19:31  robert

	* Makefile, lame.c, takehiro.c, util.h:
	  patch to prevent core dump in case big value = 576 in
	  count_bits_long()

2000-08-13 13:44  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h, Dll/Example.cpp,
	  Dll/LameDLLInterface.htm:
	  Added Lame presets to the DLL interface + added HTM document how
	  to use the lame_enc.dll interface

2000-08-13 02:50  afaber

	* Dll/Example.cpp, Dll/Example.dsp, Dll/Example.dsw,
	  Dll/MP3EncDll.dsp:
	  Added DLL Example

2000-08-12 15:42  robert

	* quantize.c:
	  major speed increase for vbr-old

2000-08-11 20:53  robert

	* psymodel.c:
	  RH_AMP tweak, improves vbrtest.wav

2000-08-08 21:08  afaber

	* Dll/BladeMP3EncDLL.c:
	  Add new RecordItem array, not completely finsihed yet, but it
	  compiles

2000-08-08 20:11  cisc

	* psymodel.c, takehiro.c:
	  code cleanup

2000-08-08 04:01  markt

	* Makefile:
	  Dec Alpha makefile fix

2000-08-07 18:24  robert

	* Makefile, lame.c, quantize-pvt.c, quantize.c:
	  code cleanup

2000-08-07 05:53  markt

	* takehiro.c:
	  code cleanup in count_bits_long()

2000-08-07 00:00  markt

	* doc/html/history.html, mpglib/interface.c, mpglib/layer3.c:
	  bug fixed in mpglib error recovery from corrupt frames

2000-08-06 22:58  markt

	* get_audio.c, testcase.mp3:
	  oops, one bug in 3.86beta. try and change this before
	  anyone downloads source from the web site :-)

2000-08-06 22:34  markt

	* doc/html/history.html, get_audio.c, version.h:
	  lame 3.86 beta release

2000-08-06 21:34  markt

	* lame.c, lame.h, testcase.mp3:
	  disabled scalefac_scale (except if -q1 is used)
	  and make old vbr mode the default

2000-08-06 20:58  markt

	* quantize-pvt.c, testcase.mp3:
	  Reverted back to original definition of over_noise and tot_noise:
	  
	  tot_noise = is really the average over each sfb of:
	  [noise(db) - allowed_noise(db)]
	  
	  and over_noise is the same average, only over only the
	  bands with noise > allowed_noise.

2000-08-06 20:32  markt

	* mpglib/main.c:
	  Album ID tag reading bug fixes

2000-08-06 20:21  markt

	* Dll/MP3export.pas, Dll/README, INSTALL, USAGE, lame.c, lame.h,
	  parse.c, quantize-pvt.c, takehiro.c, timestatus.c, timestatus.h,
	  util.h, vbrquantize.c:
	  added the "--athlower n" option which lowers the ATH by n db.
	  
	  Added possible fix for region0_count and region1_count getting
	  set to illegal negative values.

2000-08-03 20:07  robert

	* Makefile, lame.c, lame.spec, quantize-pvt.c, quantize.c:
	  don't worry Mark! my noise calculation wrapped by RH_NOISE_CALC
	  define RH_NOISE_CALC at compile time to get my version of noise
	  calculation
	  
	  special:
	  -Y amp_scalefac_bands will amplify only the maximum distorted
	  band

2000-08-01 06:36  markt

	* INSTALL, bitstream.c:
	  added a few assert's

2000-08-01 04:53  markt

	* quantize.c:
	  Fixed amp_scalefac_bands (RH found this)

2000-08-01 04:38  markt

	* get_audio.c, mpglib/main.c, util.c, util.h:
	  updated lame --decode to read Roel's AID stuff

2000-07-31 19:17  markt

	* INSTALL, Makefile.DJGPP, lame.spec:
	  updated debug makefile

2000-07-31 19:00  markt

	* Makefile, TODO, USAGE:
	  open bsd updates to makefile

2000-07-31 18:46  markt

	* quantize.c, takehiro.c, testcase.mp3:
	  Takehiro's best_huffman_divide turned back on
	  for MPEG1 only

2000-07-31 18:38  markt

	* quantize-pvt.c, quantize-pvt.h, quantize.c, takehiro.c:
	  many commits were made over the weekend. But:
	  
	  1. MPEG2 encoding was broken
	  2. all my test cases failed.
	  
	  #11 is bad. #2 can be caused be general improvements, but
	  coupled with #1 is a bad sign.
	  
	  I spent all of sunday tracking down the many changes, and trying
	  to find what what the cause of what. I finally ran out of time
	  and am going to revert some changes back to last working copy.
	  
	  Here is a summary:
	  
	  takehiro.c: newest version, except:
	  count_bits_long uses old code (new code commented out).
	  This code was breaking MPEG2
	  
	  scfsi: uses old slen1_n, slen2_n. I will change this
	  with the next commit. (in about 10min)
	  
	  best_huffman_divide: short block code breaks MPEG2.
	  short block code is still there, but disabled.
	  
	  quantize-pvt.c: Newest version, except:
	  
	  all noice calculations reverted back to db.
	  tot_noise, over_noise, ave_noise, etc. are given in db
	  and will always be given in db.
	  
	  distort[], is left in units of energy, since this saves
	  many log10() function calls.
	  
	  
	  quantize.c : Reverted back to last working version, except
	  have newest version of VBR_iteration_loop().

2000-07-31 00:36  robert

	* psymodel.c:
	  revert back to previous version,
	  it creept in by accident

2000-07-30 23:48  robert

	* lame.c, psymodel.c, quantize.c:
	  minor tweaks

2000-07-29 22:37  robert

	* lame.bat:
	  long file names on in for, thanks to Ihárosi Wiktor

2000-07-27 16:42  takehiro

	* bitstream.c:
	  minor change

2000-07-27 16:41  takehiro

	* quantize.c, vbrquantize.c:
	  make the short block huffman coding more efficient

2000-07-27 16:39  takehiro

	* machine.h:
	  oops MMX code couldn't compile...

2000-07-27 16:38  takehiro

	* fft.c:
	  removed unused code

2000-07-27 16:38  takehiro

	* takehiro.c:
	  restore old algorithm temporary...

2000-07-26 13:56  takehiro

	* newmdct.c:
	  minor coding hack and prepare for GOGO's subband filtering code

2000-07-26 13:47  takehiro

	* quantize-pvt.c, quantize-pvt.h, takehiro.c, util.h:
	  debug for region0/1 which kzmi reported

2000-07-26 13:25  takehiro

	* i386/nasm.h, machine.h, quantize-pvt.h, takehiro.c:
	  i hope this fix make intel native asm code available on VC

2000-07-25 22:28  robert

	* quantize.c:
	  possible endless loop for different -X modes fixed,
	  some modes do not minimize "over" and would never
	  reach over == 0 in some rare cases

2000-07-25 20:39  robert

	* quantize.c:
	  small "fast encode" fix

2000-07-25 20:09  robert

	* Makefile:
	  fixed typo

2000-07-25 19:24  robert

	* quantize-pvt.c, quantize-pvt.h, quantize.c:
	  fixed somehow messed up noise calculation

2000-07-25 15:42  robert

	* quantize.c:
	  small bug fix

2000-07-24 23:50  cisc

	* get_audio.c:
	  small fix

2000-07-24 23:32  cisc

	* USAGE, get_audio.c:
	  Changed lame_decoder() to write native endian format when not
	  writing WAV header (-x will byteswap).

2000-07-24 20:31  markt

	* quantize.c:
	  removed some print statements

2000-07-24 20:30  markt

	* quantize-pvt.c, quantize.c, takehiro.c:
	  fixed rare bug with -X0 getting stuck in a loop when over>0 and
	  over and over_noise didn't change from one iteration to the next.

2000-07-24 05:51  markt

	* parse.c:
	  updated docs

2000-07-24 05:42  markt

	* USAGE, get_audio.c, gtkanal.c, lame.h, parse.c:
	  -t: disables Xing header for encoding,
	  disables WAV header for decoding

2000-07-21 17:39  markt

	* vorbis_interface.c:
	  vorbis interface fixes

2000-07-21 17:09  markt

	* INSTALL, Makefile.DJGPP, README.DJGPP:
	  updated some DJGPP stuff

2000-07-21 16:50  markt

	* VbrTag.c, encoder.h, get_audio.c:
	  added some comments about decoder/encoder delay,

2000-07-20 04:29  kzmi

	* machine.h:
	  add mingw32 support. using windows.h.

2000-07-20 04:28  kzmi

	* Makefile:
	  add non-UNIX environment support. (NOUNIXCMD)
	  specifying UNAME,ARCH,PGM on command line is available.

2000-07-19 23:19  robert

	* vbrquantize.c:
	  oops, fixed typo

2000-07-19 22:42  robert

	* vbrquantize.c:
	  fixed problem with Roel's clips.wav file,
	  "pseudo endless loop" problem now gone
	  maybe the annoying spikes too?

2000-07-18 22:57  robert

	* vbrquantize.c:
	  moved calc_xmin out of VBR_noise_shaping
	  moved xr34 calculation out of VBR_noise_shaping
	  changed analog silence detection from granule based to frame
	  based
	  hacked in a first digital silence treatment per granule (can be
	  optimized)
	  
	  round about 8 percent faster on my Pentium 200 machine

2000-07-18 00:09  markt

	* lame.c:
	  moved some stuff into lame_init()

2000-07-17 23:36  robert

	* vbrquantize.c:
	  to make someone happy ;)

2000-07-17 23:28  markt

	* lame.c:
	  Changes from Kimmo: split lame_init_params() into 3 smaller
	  routines. This somehow fixes some crashes on PPC (which OS?).

2000-07-17 23:10  markt

	* USAGE, lame.c, machine.h, main.c, mpglib/common.c,
	  mpglib/mpg123.h, quantize.c, util.c:
	  Mac patches. (in an #ifdef macintosh)

2000-07-17 15:38  robert

	* vbrquantize.c:
	  analog silence treatment like in old VBR

2000-07-16 02:16  robert

	* quantize.c:
	  minor tweaks

2000-07-15 19:20  robert

	* quantize.c:
	  OK found what was wrong, should work now

2000-07-15 17:55  markt

	* lame.c, quantize.c:
	  Reverted back to older version of quantize.c becuase
	  version 1.137 broke some of my CBR test cases.
	  
	  made scalefac_scale always on for all vbr modes

2000-07-15 16:49  robert

	* lame.c, quantize.c:
	  Takehiro's scalefac_scale now default for old-VBR, no more -q1
	  necessary

2000-07-15 04:55  markt

	* Makefile:
	  updated makefile for Dec Alpha Linux

2000-07-14 20:06  markt

	* USAGE, lame.h:
	  c++ name mangling fix, type in USAGE

2000-07-12 18:03  robert

	* util.c, util.h:
	  critical band width formula added

2000-07-11 02:25  markt

	* vbrquantize.c:
	  tuning for vbr_mt to fix chirp from John Dalton

2000-07-11 00:47  markt

	* quantize.c:
	  better tuning for ABR mode

2000-07-11 00:31  markt

	* gtkanal.c, mpglib/main.c, quantize-pvt.c, quantize.c,
	  vbrquantize.c:
	  bug in frame analyzer (causes hangs near eof) fixed
	  better resyncing during mp3 file initialization in mpglib

2000-07-10 23:03  markt

	* lame.c, main.c, parse.c, reservoir.c, testcase.mp3,
	  vbrquantize.c:
	  tuned new VBR so Roel will be happy :-)

2000-07-10 00:24  markt

	* timestatus.c:
	  timestatus overflow fix

2000-07-09 20:05  markt

	* Makefile.DJGPP, README.DJGPP, quantize-pvt.c, quantize.c:
	  more updated for DJGPP

2000-07-09 13:19  shibatch

	* fft.h:
	  I forgot to commit fft.h. (by Naoki Shibata)

2000-07-09 13:16  shibatch

	* fft.c, lame.c, lame.h, parse.c, psymodel.c, quantize-pvt.c:
	  Added --nspsytune command line option. This should solve vbrtest
	  problem. (By Naoki Shibata)

2000-07-09 03:22  gramps

	* parse.c:
	  additional help for ID3-style comments in Ogg Vorbis output from
	  a patch by Ralph Giles

2000-07-09 03:20  gramps

	* lame.c:
	  don't add ID3 tags to Ogg Vorbis output

2000-07-09 03:18  gramps

	* vorbis_interface.c:
	  partial support for ID3-style comments from a patch by Ralph
	  Giles

2000-07-08 15:08  robert

	* quantize.c:
	  long time BUG in old VBR fixed, preventing LAME from using more
	  than 2500 bits per channel

2000-07-08 00:39  markt

	* README.DJGPP, get_audio.c:
	  8bit .wav files supported.

2000-07-07 21:42  robert

	* quantize-pvt.h, quantize.c:
	  little VBR-old speed up (upto 10%)

2000-07-07 20:09  markt

	* Makefile.DJGPP, README.DJGPP:
	  updated DJGPP stuff from Chris Wise

2000-07-07 00:34  robert

	* quantize.c:
	  more "old-VBR" tuning

2000-07-06 22:20  cisc

	* amiga_mpega.c:
	  Should now work for GCC.

2000-07-06 16:54  robert

	* Dll/BladeMP3EncDLL.c:
	  DLL uses now the same VBR routine by default as the EXE

2000-07-05 04:57  markt

	* quantize-pvt.c, timestatus.c:
	  minor changes

2000-07-04 23:52  cisc

	* brhist.c, quantize-pvt.c, takehiro.c:
	  code cleanup

2000-07-04 23:52  gramps

	* id3tag.h, lame.h:
	  make id3tag.h dependent on lame.h instead of the other way around

2000-07-04 22:16  markt

	* lame.h, testcase.mp3, util.h, version.h:
	  updated version to 3.86 alpha

2000-07-04 15:40  robert

	* lame.c, quantize.c:
	  old VBR silence bug fixed

2000-07-04 14:46  robert

	* VbrTag.c:
	  oops, 1 bit forgotten

2000-07-04 03:32  kzmi

	* quantize-pvt.c, quantize-pvt.h, quantize.c, vbrquantize.c:
	  old tot_noise and over_noise are obsolete.
	  tot_avg_noise and over_avg_noise were renamed as tot_noise
	  and over_noise.

2000-07-04 03:28  gramps

	* USAGE, VbrTag.c, doc/html/id3.html, id3tag.c, id3tag.h, lame.c,
	  lame.h, main.c, mp3rtp.c, parse.c:
	  ID3 version 2 tag support

2000-07-03 17:10  robert

	* VbrTag.c:
	  keep CRC bit for additional Xing-VBR frame

2000-07-03 16:21  markt

	* doc/html/history.html:
	  *** empty log message ***

2000-07-03 16:20  markt

	* README.DJGPP, lame.h, psymodel.c, quantize-pvt.c, vbrquantize.c,
	  version.h:
	  lame 3.85 beta release
	  (old VBR mode back to being the default)

2000-07-03 00:58  markt

	* machine.h, psymodel.c:
	  added note about FLOAT8 = float breaking certain features

2000-07-03 00:57  markt

	* Makefile:
	  *** empty log message ***

2000-07-03 00:48  markt

	* psymodel.c, testcase.mp3:
	  updated mid/side demasking thresholds with takehiro's formula

2000-07-03 00:25  markt

	* psymodel.c, vbrquantize.c:
	  more tweakes to ATH/PE calculation. psymodel.c uses additive
	  masking
	  so ATH needed a *= numlines.

2000-07-02 23:51  markt

	* psymodel.c:
	  added ATH threshold to PE calculation.

2000-07-02 23:39  markt

	* psymodel.c, quantize-pvt.c, quantize-pvt.h, util.c, util.h,
	  vbrquantize.c:
	  code so that ATH can be used for PE formula

2000-07-02 21:44  markt

	* Makefile:
	  Makefile updated so float8 = double

2000-07-02 18:40  markt

	* lame.c, quantize-pvt.c, quantize-pvt.h, quantize.c, takehiro.c,
	  util.c, vbrquantize.c:
	  short block re-ording complete!
	  all routines updated, but not cleaned up: it is now possible
	  to merge most short block and long block loops

2000-07-01 22:12  robert

	* lame.c:
	  code cleanup

2000-07-01 21:48  robert

	* util.c:
	  integer based padding routine

2000-07-01 20:37  cisc

	* lame.c, quantize-pvt.h, util.h:
	  code cleanup

2000-07-01 18:12  markt

	* bitstream.c, lame.c, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  takehiro.c, testcase.mp3, util.c, util.h, vbrquantize.c:
	  re-order short blocks data storage. in preperation for takehiro's
	  asm routines

2000-07-01 15:23  robert

	* lame.c, quantize-pvt.c, quantize.c:
	  old VBR no ATH bug fixed

2000-07-01 13:52  robert

	* lame.c, quantize-pvt.h, quantize.c, util.c, util.h:
	  code cleanup

2000-07-01 11:26  robert

	* lame.c, lame.h, parse.c, quantize-pvt.c, quantize.c:
	  made switch between default VBR modes easier, look into lame.h

2000-07-01 02:35  markt

	* lame.c, parse.c, version.h:
	  3.85 alpha (CVS) version: new vbr mode is the default

2000-07-01 02:32  markt

	* parse.c:
	  still bugs trying to make 3.84beta release

2000-07-01 02:16  markt

	* lame.c, parse.c, quantize-pvt.c:
	  i screwed up the 3.84 beta release, lets try again.
	  making old vbr mode the defailt, temporarily

2000-07-01 02:11  markt

	* version.h:
	  *** empty log message ***

2000-06-30 23:32  markt

	* Makefile:
	  makefile: -DFLOAT8 is float

2000-06-30 23:30  markt

	* lame.c, parse.c, version.h:
	  ok, back to new vbr mode = default

2000-06-30 23:29  markt

	* doc/html/history.html:
	  *** empty log message ***

2000-06-30 23:27  markt

	* Makefile, lame.c, parse.c, version.h:
	  old vbr mode made the default. --vbr-old and --vbr-new options
	  can specify which mode. This is just for the next 10min
	  while I put out lame 3.84. CVS will then revert back to
	  default = new vbr mode

2000-06-30 23:15  markt

	* get_audio.c:
	  8 bit input support (maybe works?)

2000-06-30 14:25  kzmi

	* VbrTag.c, bitstream.c, brhist.c, debugscalefac.c, gtkanal.c,
	  psymodel.c, quantize-pvt.c, vbrquantize.c, vorbis_interface.c:
	  replaced fprintf/printf/exit() functions with macro

2000-06-27 13:58  kzmi

	* get_audio.c, lame.c, main.c, timestatus.c, timestatus.h:
	  replace printf/fprintf/exit() with macro.
	  decode_progress() and decode_progress_finish() are added
	  in timestatus.c. these functions are used in lame_decode().

2000-06-27 02:23  markt

	* INSTALL, TODO:
	  minor updates

2000-06-27 01:18  markt

	* mpglib/layer3.c:
	  Sergey's bug fix for layer3.c MPEG2 tables

2000-06-26 20:47  afaber

	* Dll/MP3EncDll.dsp:
	  Version 1.30 beta 1

2000-06-26 16:08  markt

	* quantize.c:
	  put Roberts ABR -V n settings back in.

2000-06-26 05:56  markt

	* TODO, lame.c, quantize.c, vbrquantize.c:
	  fixed one bug in VBR handling of vbrquantize.c
	  changed ABR mode: it should not be adjusting masking thresholds
	  based on quality settings. It doesn't make sense to change the
	  maskings if the number of bits is fixed.

2000-06-25 19:07  robert

	* quantize-pvt.h, quantize.c:
	  little improvements for experimentalX modes

2000-06-24 07:49  kzmi

	* parse.c, util.h:
	  replace printf/fprintf and exit() in parse.c with macro

2000-06-24 06:00  kzmi

	* util.h:
	  abolished LAME_ASSERT().

2000-06-24 05:57  kzmi

	* util.c:
	  LAME_ASSERT() macro was removed. It was replaced assert().

2000-06-23 14:45  kzmi

	* util.h:
	  FLUSH_ERR() --> FLUSH_ERROR()

2000-06-23 13:59  kzmi

	* util.h:
	  new macros used instead of fprintf(), assert(), and exit().

2000-06-23 13:53  kzmi

	* util.c:
	  Alternative printing function lame_errorf() was added.
	  It will be called instead of fprintf() by using new macro
	  "ERRORF"

2000-06-22 20:41  robert

	* lame.c, parse.c:
	  minor bugs

2000-06-22 18:42  robert

	* USAGE, lame.c, parse.c, quantize.c:
	  Marks new VBR now the default, old still under there, use
	  --vbr-old

2000-06-22 17:16  robert

	* lame.c, util.h:
	  padding as in "MPEG-Layer3/Bitstream Syntax and Decoding"

2000-06-22 16:06  robert

	* mpglib/layer3.c:
	  table correction, thanks to Sergey Sapelin

2000-06-21 15:38  afaber

	* machine.h:
	  Update project file (added FLOAT8_is_float define)

2000-06-21 05:04  markt

	* lame.c, lame.h, psymodel.c, psymodel.h, quantize-pvt.c, util.h,
	  vorbis_interface.c:
	  added some return codes, removed some exits

2000-06-21 01:14  markt

	* brhist.c:
	  added brhist display for windows, from mremondini@racine.ra.it

2000-06-21 00:05  markt

	* machine.h:
	  added default settings for FLOAT8

2000-06-20 23:34  markt

	* psymodel.c, quantize-pvt.c, util.h:
	  added code to compute numlines and bo,bu arrays instead of using
	  table data

2000-06-20 13:28  takehiro

	* Makefile, machine.h:
	  prepare for architecture depending optimization

2000-06-19 05:08  markt

	* main.c, parse.c, vorbis_interface.c:
	  more vorbis bugs fixed

2000-06-19 01:10  markt

	* lame.c:
	  oops, if (gfc->filter_type=0) statement fixed

2000-06-18 14:39  markt

	* mpglib/interface.c:
	  mpglib: better resyncing code

2000-06-18 04:34  cisc

	* version.c:
	  minor change

2000-06-18 04:15  cisc

	* takehiro.c:
	  Removed redundant return.

2000-06-18 04:13  markt

	* vorbis_interface.c:
	  more ogg updates

2000-06-18 04:09  markt

	* lame.c:
	  ogg stuff

2000-06-18 04:08  cisc

	* version.c:
	  Included get_audio.h for prototypes...

2000-06-18 03:59  cisc

	* version.c:
	  Added version-check for libsndfile...

2000-06-18 03:43  markt

	* lame.c, version.c:
	  ogg updates

2000-06-17 14:00  robert

	* parse.c, quantize-pvt.c:
	  typo in parse.c was disabling -Xn settings

2000-06-17 04:57  takehiro

	* TODO:
	  something added

2000-06-17 00:35  markt

	* lame.c, psymodel.c, util.h:
	  code to compute numlines_[] mostly debugged, but not in use.
	  only remaining psycho acoustic table data: minval (for long
	  blocks) and SNR for short blocks.

2000-06-16 18:17  markt

	* encoder.h, psymodel.c, testcase.mp3, util.h:
	  psymodel cleanup, bval[] table lookup replaced with
	  formulas. Next step: replace numlines with formulas

2000-06-16 18:03  robert

	* TODO:
	  point added

2000-06-16 17:31  markt

	* psymodel.c, quantize-pvt.c, vbrquantize.c:
	  code re-arrangment in psymode.c. In preperation to replace
	  bval[] from table lookup with formulas.

2000-06-15 16:45  markt

	* vbrquantize.c:
	  bug fix for -Y VBR mode: needed to check for return code
	  for count_bits for errors.

2000-06-15 15:56  markt

	* mpglib/interface.c, parse.c, vbrquantize.c, vorbis_interface.c:
	  mpglib bug fix from Naoki

2000-06-15 00:06  markt

	* lame.c:
	  scalefac_scale disabled for VBR modes.

2000-06-14 21:13  markt

	* quantize-pvt.c, quantize-pvt.h, quantize.c, vbrquantize.c:
	  A new noise moded(-X 7) and a new structure to keep track
	  of all the different noise measures, from Iwasa Kazmi

2000-06-14 21:04  markt

	* lame.c, lame.h, main.c, parse.c, util.c, util.h:
	  new FindNearestBitrate routine, return codes for
	  lame_init_params.

2000-06-14 20:03  markt

	* USAGE, gpkplotting.c:
	  font in frame analyzer, updates to USAGE

2000-06-14 03:15  markt

	* Makefile.MSVC, vorbis_interface.c:
	  updated vorbis support (mono, different samplerates)

2000-06-13 12:42  takehiro

	* bitstream.c, tables.h, takehiro.c:
	  minor optimization and debug

2000-06-13 11:27  takehiro

	* i386/choose_table.nas, tables.c, takehiro.c:
	  completely MMXed choose table routine. faster!

2000-06-12 05:29  markt

	* psymodel.c, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  vbrquantize.c:
	  added a different FFT->MDCT normalization, disabled.

2000-06-10 09:38  robert

	* Makefile.MSVC, machine.h:
	  vorbis support

2000-06-10 03:39  cisc

	* quantize-pvt.c, vbrquantize.c:
	  Aaargh, and now it suddenly doesn't work anymore, just gives me
	  tons and tons of bitreservoir errors. :P

2000-06-10 02:43  cisc

	* vbrquantize.c:
	  Ooops, fixed redefinition...

2000-06-10 01:56  cisc

	* quantize-pvt.c, vbrquantize.c:
	  TAKEHIRO_IEEE754_HACK works great now, let's put it as default.

2000-06-10 00:26  markt

	* quantize-pvt.c:
	  Mat's ROUNDFAC fix. If #define TAKEHIRO_IEEE754_HACK is
	  used, ROUNDFAC should be set to -0.0946

2000-06-10 00:10  markt

	* README, USAGE, lame.c, parse.c, quantize.c:
	  trivial changes

2000-06-09 22:54  robert

	* USAGE:
	  average bitrate coding (--abr) usage added

2000-06-09 22:18  robert

	* get_audio.c, lame.c, lame.h, parse.c, quantize.c, quantize.h:
	  Safe VBR mode available as --abr x, targetting x kbits
	  you can use -b and -B switches as usual with VBR

2000-06-09 20:34  markt

	* get_audio.c, get_audio.h, quantize-pvt.c:
	  minor changes

2000-06-09 01:00  markt

	* tables.c:
	  fixed bug in tables.c

2000-06-09 00:28  markt

	* lame.c:
	  fix for when compiling without #define HAVEVORBIS

2000-06-09 00:22  markt

	* Dll/MP3EncDll.dsp, INSTALL, Makefile:
	  updated makefiles for VORBIS option

2000-06-08 23:19  markt

	* USAGE, lame.c, parse.c, vorbis_interface.c:
	  .ogg encoding now works.

2000-06-08 21:15  markt

	* lame.c, vorbis_interface.c:
	  more .ogg encoding stuff. still not working

2000-06-08 20:46  markt

	* get_audio.c, gtkanal.c, lame.c, lame.h, parse.c,
	  vorbis_interface.c:
	  --ogg option to produce .ogg files. not yet working

2000-06-08 06:04  markt

	* API:
	  updated API to include HAVEVORBIS info

2000-06-08 05:59  markt

	* Makefile:
	  makefile for libvorbis support

2000-06-08 05:58  markt

	* USAGE, get_audio.c, lame.h, mpglib/main.c, parse.c,
	  vorbis_interface.c:
	  added .ogg file decoding.
	  Re-encode all your Vorbis files to MP3 !!!
	  
	  (or, lame --decode input.ogg output.wav might be usefull)

2000-06-07 23:36  markt

	* INSTALL, get_audio.c, lame.h, reservoir.c:
	  started to add .ogg decoding

2000-06-07 22:56  sbellon

	* VbrTag.c, get_audio.c, ieeefloat.c, ieeefloat.h, machine.h,
	  mpglib/decode_i386.c, mpglib/mpg123.h, portableio.c, reservoir.c:
	  added support for FPA10 hardware (RISC OS only)

2000-06-07 22:54  sbellon

	* main.c:
	  changed RISC OS file typing code

2000-06-07 13:26  takehiro

	* lame.spec:
	  spec file for RPM. contributed by ramsy@linux.or.jp

2000-06-05 18:44  markt

	* gpkplotting.c, gtkanal.c:
	  updated X windows font selection (thanks Iwasa)

2000-06-05 17:38  markt

	* testcase.mp3:
	  updated testcase

2000-06-03 05:39  takehiro

	* takehiro.c:
	  more aggressive scfsi using and little bit memory optimization

2000-06-03 02:46  cisc

	* get_audio.c:
	  code cleanup

2000-06-01 18:30  cisc

	* amiga_mpega.c:
	  code cleanup

2000-05-31 21:00  markt

	* quantize.c:
	  SAFE_VBR mode fixes

2000-05-31 20:40  markt

	* USAGE, lame.c, parse.c, psymodel.c, quantize.c:
	  SAFE_VBR mode added. not enabled yet

2000-05-31 19:52  cisc

	* vbrquantize.c:
	  code cleanup

2000-05-31 03:37  cisc

	* timestatus.c:
	  code cleanup

2000-05-31 00:25  robert

	* Makefile:
	  *** empty log message ***

2000-05-31 00:24  robert

	* mpglib/interface.c, mpglib/layer3.c, mpglib/main.c,
	  mpglib/mpg123.h:
	  code cleanup

2000-05-30 23:47  markt

	* lame.c:
	  -f "fast mode" is now a little slower.
	  
	  It now *will* compute psycho acoustics, used for total bits
	  and pre-echo detection. But it will not spend any time
	  trying to find optimal scalefactors.

2000-05-30 23:37  markt

	* TODO, bitstream.c, lame.c, takehiro.c, vbrquantize.c:
	  fixed scfsi bug in vbrquantize.c

2000-05-30 22:29  cisc

	* VbrTag.c, bitstream.c, get_audio.c, l3side.h, main.c,
	  quantize-pvt.c, quantize-pvt.h, util.c:
	  code cleanup

2000-05-30 22:01  cisc

	* takehiro.c, timestatus.c:
	  code cleanup

2000-05-30 20:02  robert

	* Makefile, VbrTag.c, gtkanal.c, lame.c, newmdct.c, psymodel.c,
	  quantize-pvt.c, reservoir.c, takehiro.c, timestatus.c, util.c,
	  util.h, vbrquantize.c:
	  code cleanup

2000-05-30 16:01  robert

	* fft.c, get_audio.c, main.c:
	  code cleanup

2000-05-30 15:16  robert

	* bitstream.c, l3side.h, quantize.c, quantize.h, util.h:
	  signed/unsigned fixes

2000-05-30 12:00  takehiro

	* i386/nasm.h:
	  some difinition added

2000-05-30 00:20  markt

	* Makefile:
	  updated Makefile comment about Sun OS

2000-05-30 00:12  markt

	* bitstream.c, psymodel.c, reservoir.c, takehiro.c, vbrquantize.c:
	  psymodel.c: yet another loop check added
	  
	  vbrquantize.c: best_scalefac_store can change scalefac values.
	  scalefactors must be saved and restored before each call.
	  
	  bitstream.c: added part2_length assert check

2000-05-29 22:11  cisc

	* get_audio.c:
	  Included portableio.h for prototypes.

2000-05-29 21:49  cisc

	* amiga_mpega.c:
	  Skip bad frames.

2000-05-29 19:59  robert

	* VbrTag.c, get_audio.c, quantize.c, vbrquantize.c:
	  compiler warnings fixed

2000-05-29 17:15  afaber

	* VbrTag.c:
	  Private bit was not retained in VBR header, changes masking bit
	  from 0x0c to 0x0d

2000-05-29 17:14  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h:
	  Some minor fixes to the DLL inteface, added Write VBR tag
	  prototype
	  Added LGPL header to *.h file

2000-05-29 17:12  afaber

	* lame.dsp:
	  Set proper defines in the debug mode

2000-05-29 16:50  markt

	* mpglib/layer3.c, mpglib/main.c, mpglib/mpglib.h:
	  parent_is_slash #define, for the Amiga

2000-05-29 14:39  markt

	* brhist.c, machine.h, mpglib/interface.c, psymodel.c:
	  fixed loop problems with large --cwlimit

2000-05-28 23:47  markt

	* Makefile:
	  makefile now -O3 (linux) by default

2000-05-28 23:46  markt

	* Makefile, USAGE, mpglib/interface.c, psymodel.c, quantize.c,
	  reservoir.c, vbrquantize.c:
	  code cleanup

2000-05-28 06:56  takehiro

	* quantize-pvt.c:
	  restored old GCC asm routine

2000-05-28 05:01  takehiro

	* Makefile, i386/choose_table.nas, i386/nasm.h, takehiro.c:
	  initial MMX implementation attempt..
	  use MMX to huffman coding.

2000-05-27 07:41  takehiro

	* takehiro.c:
	  more pseudo SIMD code

2000-05-27 07:23  takehiro

	* testcase.mp3:
	  testcase update for new scalefac_scale

2000-05-27 07:22  takehiro

	* quantize-pvt.c, quantize-pvt.h, takehiro.c:
	  pseude SIMD countbit routine preparing for MMX

2000-05-27 07:21  takehiro

	* quantize.c:
	  oops, foolish typemiss...

2000-05-27 03:22  takehiro

	* lame.c, quantize.c, version.h:
	  scalefac_scale debug and use it as default
	  new subblock_gain algorithm
	  (initial release, -Z to enable it, but not works fine)

2000-05-27 03:08  takehiro

	* newmdct.c:
	  minor changing

2000-05-27 03:07  takehiro

	* quantize-pvt.c, quantize-pvt.h, quantize.c:
	  scalefac_scale debug and use it as default
	  new subblock_gain algorithm
	  (initial release, -Z to enable it, but not works fine)

2000-05-26 19:03  markt

	* gtkanal.c, lame.c, vbrquantize.c:
	  vbrquantize.c work

2000-05-26 01:49  cisc

	* amiga_mpega.c, get_audio.c, lame.h:
	  Fixed a bunch of errors that must have been committed whilst
	  sleepwalking. ;)

2000-05-26 00:49  markt

	* get_audio.c, lame.h, main.c:
	  moved WAV writing routines and lame_decoder() into get_audio.c

2000-05-26 00:31  markt

	* main.c:
	  another bug in WriteWav

2000-05-26 00:02  markt

	* main.c, quantize-pvt.c:
	  tweaked .wav file output. Fixed small bug in short block bit
	  allocation (thanks Iwasa!)

2000-05-25 15:27  markt

	* amiga_mpega.c, get_audio.c, gtkanal.c, lame.h, main.c,
	  mpglib/interface.c, mpglib/main.c, vbrquantize.c:
	  updated frame analyzer to show difference between
	  original/decoded
	  updated --decode to return bitrate, samplerate, numchannels.
	  if samplerate or numchannels changes in mp3 stream, --decode will
	  quite with an error message

2000-05-24 16:53  markt

	* gtkanal.c, gtkanal.h, mpglib/layer3.c, quantize-pvt.c,
	  vbrquantize.c:
	  tweeks to frame analyzer: added preflag display

2000-05-24 00:24  cisc

	* get_audio.c:
	  Fixed silly bug preventing mp3input working together with
	  libsndfile.

2000-05-23 04:12  markt

	* bitstream.c, gtkanal.c, quantize-pvt.c, quantize.c,
	  vbrquantize.c:
	  fixed a bug in the frame analyzer: was rescaling axis between
	  plotting
	  energy, masking and noise

2000-05-22 18:20  shibatch

	* mpglib/tabinit.c:
	  Changed decwin to more precise one.

2000-05-22 17:11  markt

	* quantize-pvt.c:
	  make IEEE stuff the default on GNU i386

2000-05-22 13:14  cisc

	* main.c:
	  fwrite() returns number of objects actually written

2000-05-22 05:39  markt

	* main.c:
	  fixed bug for mono .wav headers for lame --decode

2000-05-22 05:27  markt

	* quantize-pvt.c:
	  GNU C ASM code is temporarily disabled.
	  
	  It looks like Takehiro's changes to 1.107 have broken this code.

2000-05-22 05:07  markt

	* mpglib/common.c, mpglib/interface.c:
	  more robust frame resyncing

2000-05-22 05:07  markt

	* quantize.c, vbrquantize.c:
	  more robust mpglib decoding (frame re-syncing)

2000-05-21 20:54  cisc

	* quantize-pvt.c:
	  TAKEHIRO_IEEE754_HACK should not be default

2000-05-21 20:31  robert

	* parse.c:
	  PRESET tunings:
	  + after we dropped the ISO 7680 bit buffer limit
	  allow more often 320 kbits frames
	  + for MPEG2.5 allow more often short blocks
	  only 8 kHz streams have now problems with
	  short blocks, so don't use them then
	  - dropped all highpass filters, cos the
	  results did not satisfy me
	  (in my opinion they actually do more harm than good)
	  - tuned some lowpass filters cos of the sfb21
	  noise calculation in VBR

2000-05-21 19:33  markt

	* l3side.h, parse.c, psymodel.c, quantize-pvt.c, quantize-pvt.h,
	  quantize.c, quantize.h, reservoir.c, reservoir.h:
	  comments and other code cleanup

2000-05-21 16:42  takehiro

	* quantize.c:
	  new scalefac_scale algorithm option changed -Y -> -Z

2000-05-21 16:25  takehiro

	* quantize.c:
	  new scalefac_scale algorithm

2000-05-21 16:19  takehiro

	* quantize-pvt.h, quantize.c:
	  new scalefac_scale algorithm. enables with -Y option.

2000-05-21 13:15  takehiro

	* quantize-pvt.h:
	  VBR bug related sfb22 fixed

2000-05-21 12:52  takehiro

	* quantize-pvt.c:
	  VBR bug related sfb22 fixed

2000-05-20 23:49  robert

	* quantize-pvt.c, quantize.c:
	  minor tweaks

2000-05-20 23:33  robert

	* quantize.c:
	  VBR hack

2000-05-20 21:05  markt

	* bitstream.c, main.c, mpglib/main.c:
	  totbit buffer overflow fix

2000-05-20 20:57  cisc

	* takehiro.c:
	  choose_table() argument is int *, not unsigned *

2000-05-20 20:36  markt

	* bitstream.c, main.c, mpglib/main.c, util.h:
	  mpglib will not decode Xing VBR header.
	  write_timing changed to unsigned long to allow encoding more
	  than 4 hours at 128kbs.

2000-05-20 15:13  markt

	* main.c, takehiro.c:
	  .wav output with --decode was putting wrong filesize in
	  .wav header.

2000-05-19 23:01  markt

	* quantize-pvt.c, testcase.mp3, version.h:
	  updated version.h lame 3.84 alpha (CVS version)

2000-05-19 20:07  markt

	* doc/html/history.html:
	  updated history for 3.83beta

2000-05-19 20:06  markt

	* version.h:
	  lame 3.83beta

2000-05-19 18:43  markt

	* encoder.h, lame.c:
	  possible serious buffering problem fixed. LAME buffers input data
	  until
	  it has at least 1904 samples. It then starts processing them.
	  
	  The old MDCT routines processed 1728 samples and had a delay of
	  528.
	  (introduced in 3.54) But the new MDCT routines, because the delay
	  is
	  so small (48 samples) actually need 2014 samples to be in the
	  buffer!
	  
	  If you use the lame input routines, for MPEG1, this bug would
	  not be found because the buffer always has well over 2014 samples
	  in it. But it was possible to trigger it in MPEG2 encodings.
	  It would also be possible to trigger this bug in programs
	  that use lame_enc.dll and pass LAME data that is not in
	  chunks of 1152 samples.

2000-05-19 17:23  markt

	* bitstream.c, lame.c, lame.h, parse.c, psymodel.c, testcase.mp3,
	  vbrquantize.c:
	  tweaks to psymodel initialization
	  more work on vbrquantize.c

2000-05-19 01:38  markt

	* bitstream.c, gtkanal.c, l3side.h, lame.c, main.c, psymodel.c,
	  quantize-pvt.c, takehiro.c, testcase.mp3, vbrquantize.c:
	  more work on vbrquantize.c

2000-05-18 18:50  markt

	* main.c:
	  updated decoder to remove initial delay

2000-05-18 05:38  markt

	* main.c, parse.c:
	  lame --decode now outputs .wav files
	  file length in header is set to 2^32-1, since we dont
	  really know ahead of time how many samples are in the mp3 file.

2000-05-17 22:25  markt

	* quantize.c:
	  ath lower bug in quantize.c fixed

2000-05-17 22:08  markt

	* quantize-pvt.c:
	  calc_noise1 will compute noise (and related info) for
	  last scalefactor band

2000-05-17 21:33  markt

	* quantize-pvt.c, takehiro.c:
	  frame analyzer displays more data in scalefactor band 22

2000-05-17 19:50  markt

	* gtkanal.h, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  vbrquantize.c:
	  modified frame analyzer to show last scalefactor band,
	  energy and masking (from the ATH). distortion is not yet
	  calculated in this band.

2000-05-17 18:08  markt

	* gtkanal.c, l3side.h, quantize-pvt.c, quantize.c, util.h:
	  Possible fix for recent VBR/sweep bug:
	  min bitrate not strictly enforced with -F, and
	  analog silence was only being computed up th 16khz.

2000-05-17 17:16  markt

	* USAGE, brhist.c, get_audio.c, lame.c, lame.h, main.c, quantize.c,
	  reservoir.c, util.c, util.h, vbrquantize.c:
	  moved framenum, totalframes into lame.h (so programs which
	  use libmp3lame can compute statis info.)

2000-05-15 14:34  takehiro

	* newmdct.c:
	  more coding hack.

2000-05-14 23:38  cisc

	* newmdct.c:
	  Changed M_PI to PI

2000-05-14 08:12  takehiro

	* newmdct.c:
	  Naoki's short block mdct
	  and more my coding hack for mdct

2000-05-14 05:51  takehiro

	* newmdct.c, util.h:
	  mdct_long coding hack

2000-05-12 11:59  robert

	* lame.c, newmdct.c, util.h:
	  code cleanup

2000-05-12 09:51  takehiro

	* doc/html/history.html, newmdct.c:
	  more work in mdct_long and window_filtering

2000-05-12 05:03  markt

	* testcase.mp3:
	  updated testcase.mp3

2000-05-12 05:02  markt

	* version.h:
	  updated version to 3.83 alpha

2000-05-12 05:01  markt

	* API, README, doc/html/history.html, version.h:
	  version 3.83

2000-05-10 13:55  takehiro

	* quantize-pvt.c:
	  dirty hack for IEEE 754 FPU

2000-05-09 16:31  afaber

	* Dll/BladeMP3EncDLL.c:
	  Fixed problem with build date string, temp string was not
	  initialized properly (was missing a terminator character)

2000-05-09 11:30  takehiro

	* newmdct.c:
	  Naoki's mdct_long with my coding hack.

2000-05-09 02:45  cisc

	* API, brhist.c:
	  Added NOTERMCAP define and changed brhist.c accordingly.

2000-05-08 22:48  markt

	* testcase.mp3:
	  *** empty log message ***

2000-05-08 22:41  markt

	* doc/html/history.html, quantize-pvt.c:
	  updated side channel 320kbs fix

2000-05-08 22:30  robert

	* Makefile, quantize-pvt.c:
	  M/S reduce_side bug fixed,
	  nonISO buffer mode should work now for
	  320 kbits MS streams too

2000-05-08 19:55  afaber

	* Dll/BladeMP3EncDLL.c:
	  Bumped up version number of DLL version library, fixed some of
	  the debugging code

2000-05-08 17:46  markt

	* doc/html/contributors.html, doc/html/examples.html,
	  doc/html/id3.html, doc/html/index.html, doc/html/lame.css,
	  doc/html/node5.html, doc/html/node6.html, doc/html/switchs.html:
	  updated docs from Gaby

2000-05-08 17:40  markt

	* version.h:
	  3.82 alpha(CVS) version

2000-05-08 17:38  markt

	* doc/html/history.html, version.h:
	  3.81beta release

2000-05-08 17:31  markt

	* Makefile, USAGE, VbrTag.c, brhist.c, lame.h, parse.c,
	  reservoir.c, util.c:
	  removed ISO buffer limitation, added --strictly-enforce-ISO
	  option

2000-05-08 00:06  markt

	* quantize-pvt.c, reservoir.c, vbrquantize.c:
	  started work on vbrquantize.c

2000-05-07 11:09  robert

	* Makefile, Makefile.MSVC, lame.c, parse.c:
	  -F option now working,
	  updated Makefile for DOS

2000-05-07 03:59  takehiro

	* newmdct.c, util.h:
	  new IDCT32 routine from Naoki and optimized by me.
	  and more thread safe coding.

2000-05-06 22:20  markt

	* testcase.mp3:
	  updated testcases

2000-05-06 22:20  markt

	* Dll/MP3EncDll.dsp, l3bitstream-pvt.h, lame.dsp:
	  updated project files (for removal of ISO code)

2000-05-06 22:18  markt

	* Makefile, VbrTag.c, VbrTag.h, formatBitstream.c,
	  formatBitstream.h, l3bitstream.c, l3bitstream.h, lame.c,
	  quantize.c, util.c, util.h, vbrquantize.c:
	  removed all ISO code

2000-05-06 22:09  markt

	* version.h:
	  updated version to 3.81 alpha 1 (CVS version)

2000-05-06 22:07  markt

	* Makefile:
	  updated makefile for beta release

2000-05-06 22:05  markt

	* doc/html/history.html, version.h:
	  updated version for 3.80 beta release

2000-05-06 22:04  markt

	* USAGE, lame.c, lame.h, main.c, parse.c, quantize.c,
	  vbrquantize.c:
	  added -F option and compression ratio options

2000-05-06 19:16  markt

	* USAGE, lame.c, lame.h, parse.c, util.c:
	  --freeformat option added. Allows lame to write free format
	  bitstreams

2000-05-06 18:45  markt

	* bitstream.c, gtkanal.c, lame.c, mpglib/interface.c,
	  mpglib/main.c, parse.c, psymodel.c, quantize.c, util.h:
	  fixed buffer overflow in bitstream.c,
	  some improvements to mpglib mp3 playback

2000-05-06 09:27  afaber

	* Dll/MP3EncDll.dsp:
	  Update workspaces

2000-05-05 20:48  markt

	* get_audio.c, mpglib/common.c, mpglib/interface.c, mpglib/main.c,
	  mpglib/mpg123.h, mpglib/mpglib.h:
	  free format mp3 decoding now supported, not fully tested

2000-05-04 00:07  takehiro

	* takehiro.c:
	  new count bits routine, speed up by coding hack.

2000-05-03 07:23  markt

	* psymodel.c:
	  updated comments

2000-05-03 07:21  markt

	* psymodel.c, testcase.mp3:
	  minor change to long block pre-echo control

2000-05-03 07:18  markt

	* psymodel.c, testcase.mp3:
	  minor improvement to long block pre-echo control

2000-05-03 03:03  markt

	* mpglib/interface.c:
	  one last tweak to interface.c

2000-05-03 02:37  markt

	* bitstream.c, mpglib/common.c, mpglib/interface.c,
	  mpglib/layer3.c, mpglib/mpg123.h, mpglib/mpglib.h:
	  worked on mpglib/interface.c to make it handle free format
	  bitstreams. (not yet working)

2000-05-02 18:31  afaber

	* Dll/BladeMP3EncDLL.c, Dll/MP3EncDll.dsp:
	  Update lame version, bug fixes to play list file generation

2000-05-02 02:53  markt

	* reservoir.c:
	  for CBR encoding, 320kbs, dont allow main_data_begin > 0

2000-05-02 01:24  markt

	* gtkanal.c, lame.c, psymodel.c, quantize-pvt.c, quantize.c,
	  testcase.mp3:
	  tweaked a bunch of bit allocation on PE settings,
	  based on tunings of testsignal2.wav and castanets.wav

2000-05-01 18:18  markt

	* Dll/MP3EncDll.dsp, bitstream.c, brhist.c, brhist.h, lame.c,
	  lame.dsp, lame.h, parse.c, quantize-pvt.c, reservoir.c, util.h:
	  bunch of vbr histogram tweaks, and added --athshort option for
	  internal testing only

2000-04-30 17:54  robert

	* lame.bat:
	  DOS batch file to put on Win9x Desktop
	  you can drag and drop waves on it and
	  lame will encode them to mp3

2000-04-29 23:41  markt

	* bitstream.c, reservoir.c, testcase.mp3:
	  moved where main_data_begin was corrected when
	  padding into ancillary data

2000-04-26 18:14  markt

	* USAGE, lame.h, main.c, parse.c:
	  added the '--decode' option

2000-04-26 00:54  markt

	* machine.h, main.c, mpglib/decode_i386.c, mpglib/layer3.c:
	  moved calls to open the output file until after all
	  arguments and other conditions are checked.

2000-04-24 22:22  markt

	* reservoir.c, testcase.mp3:
	  disabled acillary drain (pre-frame) because of bugs at 320kbs
	  Could be player or encoder bug - have to investigate

2000-04-19 21:35  robert

	* psymodel.c:
	  MPEG2.5 related bug fixed

2000-04-19 19:35  markt

	* lame.c, lame.h, parse.c, quantize-pvt.c:
	  fixed bug allowing more than 4096 bits in a granule

2000-04-18 22:57  markt

	* lame.c, lame.h, psymodel.c, reservoir.c, takehiro.c, util.h:
	  lame is now (*maybe*) thread safe!

2000-04-18 08:35  markt

	* Dll/BladeMP3EncDLL.c, VbrTag.c, VbrTag.h, bitstream.c, brhist.c,
	  get_audio.c, get_audio.h, id3tag.c, id3tag.h, l3bitstream.c,
	  lame.c, lame.h, parse.c, psymodel.c, quantize-pvt.c, quantize.c,
	  reservoir.c, reservoir.h, testcase.mp3, util.c, util.h,
	  vbrquantize.c:
	  more thread-safe work, and id3/vbr tag bug discovered by
	  Iwasa Kasmi

2000-04-17 23:29  markt

	* lame.c, psymodel.c, quantize-pvt.c, quantize.c, util.h,
	  vbrquantize.c:
	  tweaks to vbrquantize.c

2000-04-17 06:12  markt

	* lame.c, testcase.mp3, vbrquantize.c:
	  updated test cases

2000-04-17 05:16  markt

	* bitstream.c, l3bitstream.c, l3side.h, lame.c, mpglib/main.c,
	  quantize-pvt.c, reservoir.c:
	  fixed VBR stuffing bits problem.
	  if VBR uses a large frame, then ResvMax may become small, and
	  we have to make sure that main_data_begin doesn't create a buffer
	  which violates this condition.
	  
	  The old code wouldn't check the above, but it would pad the
	  current
	  frame with 0's (wasting bits) to make sure the *next* frame
	  didn't
	  violate the ResvMax condition. But this results in wasted bits
	  because in VBR there is a change the *next* frame will be at
	  a lower bitrate and thus have a larger ResvMax.

2000-04-17 02:25  markt

	* API, Makefile, get_audio.c, gtkanal.c, lame.c, lame.h, main.c,
	  mp3x.c, mpglib/layer3.c, mpglib/main.c, mpglib/mpglib.h, parse.c,
	  psymodel.c, quantize-pvt.c, quantize.c, reservoir.c, util.h,
	  vbrquantize.c:
	  Removed gtk specific code from libmp3lame.a: libmp3lame.a can now
	  be compiled to support the mp3 frame analyzer but does not
	  require
	  gtk to be installed. (and thus, HAVEGTK does not occure in any
	  of the library routines). To compile the frame analyzer (mp3x)
	  you
	  of course still need gtk installed.

2000-04-17 00:38  markt

	* VbrTag.c, mpglib/main.c:
	  another attempt at fixing mpeg2.5 xing header

2000-04-16 20:56  markt

	* VbrTag.c:
	  VBR header for MPEG2.5 fixed.

2000-04-14 00:37  markt

	* lame.c, main.c, parse.c, quantize-pvt.c, reservoir.c,
	  reservoir.h, vbrquantize.c:
	  tweaks to vbrquantize.c

2000-04-13 01:19  markt

	* lame.c, quantize-pvt.c, quantize-pvt.h, quantize.c, util.h,
	  vbrquantize.c:
	  vrbquantize bounds checking

2000-04-12 18:44  markt

	* lame.c, psymodel.c, quantize.c, testcase.mp3, util.h,
	  vbrquantize.c:
	  removed gfc->masking_lower

2000-04-12 18:29  markt

	* Makefile, quantize-pvt.c, quantize.c:
	  Robert's VBR quality control mode now the default!

2000-04-12 18:19  markt

	* lame.c, quantize-pvt.c, quantize-pvt.h, quantize.c, util.h,
	  vbrquantize.c:
	  removed static masking_lower

2000-04-12 18:10  markt

	* quantize-pvt.c, quantize.c:
	  removed reduce_sidechannel static

2000-04-12 18:08  markt

	* quantize-pvt.c, quantize.c, vbrquantize.c:
	  removed convert_mdct static variable

2000-04-12 18:06  markt

	* quantize-pvt.c, util.h:
	  removed static variables ATH_l, ATH_s

2000-04-12 15:47  markt

	* doc/man/lame.1, lame.c, psymodel.c, quantize.c, vbrquantize.c:
	  removed qthr_l, qthr_s from psymodel.c
	  fixed man pages.
	  enabled new VBR code with -Y
	  disabled -Z option: it was broken since subblock gains were not
	  scaled into xrpow

2000-04-12 01:37  markt

	* mpglib/layer3.c:
	  gtk changes

2000-04-12 01:36  markt

	* gtkanal.c, lame.c, psymodel.c, quantize-pvt.c, quantize-pvt.h,
	  quantize.h, testcase.mp3, vbrquantize.c:
	  new vbr quantize code mostly done!

2000-04-11 18:24  markt

	* Makefile, get_audio.c, get_audio.h, gtkanal.c, gtkanal.h, lame.c,
	  lame.h, quantize.c, util.h, vbrquantize.c:
	  removed all static variables from get_audio.c

2000-04-11 15:59  markt

	* HACKING:
	  updated HACKING

2000-04-11 15:55  markt

	* API, l3side.h, lame.c, quantize-pvt.c, quantize-pvt.h, util.h,
	  vbrquantize.c:
	  updated API

2000-04-11 00:33  markt

	* lame.c, vbrquantize.c:
	  work on vbrquattize.c

2000-04-10 20:24  markt

	* lame.c, main.c, quantize-pvt.c, quantize.c, util.c, util.h:
	  bug fixed: mfbuf overflow, and check for overflow was
	  being done, but after the overflow and in this case
	  the overflow was causing the check to pass.

2000-04-10 16:41  markt

	* lame.c, quantize.h, util.c, util.h, vbrquantize.c:
	  precompute 16 windows for downsampling. much faster.

2000-04-09 21:53  markt

	* bitstream.c, mpglib/layer3.c, mpglib/main.c, quantize-pvt.c,
	  quantize.c, reservoir.c, util.c:
	  MPEG2.5 8khz sampling rate fixes

2000-04-09 02:08  markt

	* encoder.h, lame.c, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  reservoir.c, testcase.mp3:
	  tweaked bit reservoir growth, bits on pe allocation

2000-04-08 23:33  markt

	* lame.c, quantize-pvt.c, takehiro.c, testcase.mp3:
	  tweaks to on_pe, side channel bit allocation

2000-04-08 20:33  markt

	* encoder.h, testcase.mp3:
	  switched to encoder delay of 576

2000-04-08 20:31  markt

	* bitstream.c, tables.c:
	  *** empty log message ***

2000-04-08 20:22  markt

	* bitstream.c, quantize.c, takehiro.c, testcase.mp3, util.c,
	  util.h:
	  Takehiro's more efficient short block huffman coding.

2000-04-08 18:08  markt

	* lame.c:
	  VBR compression ratio calculation

2000-04-08 01:26  cisc

	* bitstream.c:
	  Use INLINE dammit! ;)

2000-04-07 00:11  markt

	* util.c:
	  fixed M_PI problem

2000-04-06 21:08  robert

	* bitstream.c:
	  scalefac_band compile fix

2000-04-06 18:30  markt

	* gtkanal.c, l3bitstream-pvt.h, l3bitstream.c, l3side.h, lame.c,
	  quantize-pvt.c, quantize-pvt.h, quantize.c, takehiro.c,
	  testcase.mp3, util.h, vbrquantize.c:
	  made scalefac_band non-local

2000-04-06 18:23  markt

	* bitstream.c, bitstream.h:
	  *** empty log message ***

2000-04-06 18:18  markt

	* bitstream.c:
	  opps, forgot to add this...

2000-04-06 17:31  markt

	* Dll/MP3EncDll.dsp, Makefile.MSVC, lame.dsp:
	  MSVC project files updated

2000-04-06 17:29  markt

	* VbrTag.c, VbrTag.h, lame.c, util.c, util.h:
	  takehiro bitstream.c now on by default!

2000-04-06 16:32  markt

	* mpglib/layer3.c, takehiro.c:
	  takehiro bitstream.c now working for CBR, disabled by default

2000-04-05 08:54  markt

	* Makefile, l3bitstream.c, lame.c, quantize-pvt.c, quantize-pvt.h,
	  tables.c, tables.h, takehiro.c, util.c, util.h:
	  takehiro's bitstream.c package close to being interface
	  into lame.

2000-04-05 01:38  markt

	* lame.c, util.c:
	  mode_fixed bug

2000-04-04 21:36  robert

	* parse.c:
	  *** empty log message ***

2000-04-04 19:51  markt

	* lame.c, util.c, util.h:
	  upsampling uses old interpolation code
	  downsampling uses good, FIR/blackman window code

2000-04-04 19:05  markt

	* Makefile.MSVC:
	  loopold removed from makefile.msvc

2000-04-04 17:45  markt

	* parse.c:
	  cwlimit parsing fixed

2000-04-04 16:05  markt

	* lame.c:
	  default mode will use jstereo

2000-04-04 03:20  markt

	* Dll/BladeMP3EncDLL.c:
	  more fixes

2000-04-03 19:50  markt

	* quantize-pvt.c:
	  more gfc changes

2000-04-03 19:19  markt

	* HACKING, lame.c, newmdct.c, quantize-pvt.c, quantize.c, util.c,
	  util.h:
	  more thread-safe work. (moving all static variables into
	  gfc-> struct)

2000-04-03 18:24  markt

	* brhist.c, get_audio.c, gtkanal.c, l3bitstream.c, l3bitstream.h,
	  lame.c, lame.h, newmdct.c, parse.c, psymodel.c, quantize-pvt.c,
	  quantize.c, quantize.h, reservoir.c, takehiro.c, util.c, util.h,
	  vbrquantize.c:
	  more massive internal changes in the goal of making
	  lame thread safe.
	  
	  gfp-> user controlled varialbes
	  gfc-> all other internal parameters. eventually all global
	  variables
	  which very from encode to encode need to go in here.

2000-04-03 16:00  markt

	* Dll/BladeMP3EncDLL.c:
	  dll finally fixed!

2000-04-03 15:38  markt

	* Dll/BladeMP3EncDLL.c:
	  bug in lame_enc.dll finally fixed?

2000-04-03 01:19  markt

	* lame.c:
	  tweaked output display

2000-04-03 01:07  markt

	* Dll/BladeMP3EncDLL.c, lame.c:
	  new resample routine. very slow, but much higher
	  quality.

2000-04-02 19:38  robert

	* parse.c:
	  overhauled presets:
	  + redesign of help page
	  + phone uses now MPEG2.5 8kHz
	  + sw new, for short wave radio @ 11.025 kHz 24kbits
	  + am new, for am radio @ 16 kHz 32 kbits
	  + fm changed, 22.05 kHz 64 kbits
	  + radio new, the old fm at 112 kbits
	  rest minor tunings

2000-04-02 02:44  markt

	* Dll/BladeMP3EncDLL.c:
	  *** empty log message ***

2000-04-02 01:55  markt

	* Dll/BladeMP3EncDLL.c:
	  more DLL updates

2000-04-02 01:39  markt

	* parse.c, util.c:
	  MPEG2.5 docs

2000-04-02 01:35  markt

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h, l3bitstream.c,
	  lame.c, lame.h, psymodel.c, quantize-pvt.c, quantize-pvt.h,
	  util.c, util.h:
	  MPEG2.5 support

2000-04-01 11:47  robert

	* lame.c, parse.c:
	  + added --version and --license switch
	  + changed --help to give only most important options
	  so that it fits on a 80x24 terminal display
	  + added --longhelp switch to get the full options list
	  + changed -v and -Vx, now they switch quality to 2
	  you can go down below 2 by adding -f or -q5
	  + changed presets:
	  turned off highpass filters except for "phone"
	  voice sample freq now 22.05 kHz, due to mpg123 bug

2000-03-31 23:53  markt

	* Dll/BladeMP3EncDLL.c, lame.c, vbrquantize.c:
	  mono encoding bug in DLL fixed

2000-03-31 22:47  robert

	* Makefile, quantize-pvt.c, quantize-pvt.h:
	  removed RH_ATH/RH_SIDE_CBR code

2000-03-31 22:45  robert

	* quantize.c:
	  minor frame analyzer fix for VBR, removed RH_ATH/RH_SIDE_CBR code

2000-03-30 23:38  markt

	* get_audio.c, gtkanal.c, lame.c, main.c, parse.c, quantize.c:
	  added -q option for internal testing. sets value of gfp->quality

2000-03-30 19:51  markt

	* testcase.mp3:
	  updated test case for new CVS version

2000-03-30 09:37  markt

	* INSTALL:
	  wingtk doc

2000-03-30 08:24  markt

	* README:
	  updated docs

2000-03-30 08:04  markt

	* Dll/MP3EncDll.dsp, README.WINGTK, lame.dsp:
	  new directions for compiling mp3x under windows

2000-03-30 07:55  markt

	* README.WINGTK:
	  updated project files

2000-03-30 05:39  markt

	* gtkanal.c, gtkanal.h, mpglib/layer3.c:
	  added big_values display to mp3x

2000-03-30 05:38  markt

	* VbrTag.c, l3bitstream.c, lame.c, machine.h, quantize.c,
	  takehiro.c, vbrquantize.c:
	  fixed 2 bugs:
	  
	  l3bitstream.c:
	  1. The quadruples were encoded with the wrong sign: Any non-zero
	  coefficient in the count1 block was marked as negative!
	  
	  takehiro.c:
	  2. short blocks were using a cod_info->big_values=288
	  (which should be 576, because big_values was multiplied by 2
	  in the new version of takehiro.c)

2000-03-29 20:49  markt

	* doc/html/contributors.html, doc/html/examples.html,
	  doc/html/history.html, doc/html/id3.html, doc/html/index.html,
	  doc/html/lame.css, doc/html/node6.html, doc/html/switchs.html:
	  new docs from Gabriel

2000-03-29 02:00  markt

	* brhist.c, fft.c, get_audio.c, globalflags.h, gtkanal.c,
	  gtkanal.h, l3bitstream.c, lame.c, newmdct.c, parse.c, psymodel.c,
	  quantize-pvt.c, quantize.c, reservoir.c, takehiro.c, util.c,
	  vbrquantize.c:
	  removed globalflags.h

2000-03-29 01:55  markt

	* vbrquantize.c:
	  removed "globalflags.h"

2000-03-29 01:06  markt

	* lame.c:
	  missing free() calls in lame_encode_buffer_interleave

2000-03-28 16:41  takehiro

	* takehiro.c:
	  oops, this may remove any seg.fault bug...

2000-03-28 15:39  takehiro

	* takehiro.c:
	  Oops, I mistake count1 for big_values

2000-03-28 13:08  cisc

	* amiga_mpega.c:
	  Fixed overflow problem with nsamp calculation.

2000-03-27 22:09  markt

	* lame.c:
	  fixes to lame_encode_buffer_interleave()

2000-03-27 21:46  robert

	* takehiro.c:
	  added assertion, cos here is a bug

2000-03-27 21:09  markt

	* lame.c:
	  minor change with ms_ener_ratio_ave

2000-03-27 21:06  markt

	* lame.c:
	  ms_ener_ratio_ave M/S override switch turned off

2000-03-27 20:58  markt

	* version.h:
	  changed version number. 3.80 is the next beta version,
	  3.70 is the stable version

2000-03-27 19:42  markt

	* version.h:
	  this is new the future version 3.81beta

2000-03-27 19:26  markt

	* takehiro.c:
	  gfp->use_scfsi replaced by mode_gr==2

2000-03-27 17:53  markt

	* get_audio.c, lame.c, main.c, util.c, version.h:
	  Ingo -f fast-math rounding problem fix

2000-03-27 15:59  takehiro

	* takehiro.c:
	  debugged seg.fault when count1 is 576

2000-03-26 21:21  markt

	* quantize.c:
	  opps, i screwed up quantize.c! fixed version

2000-03-26 10:05  markt

	* Makefile, gtkanal.c, gtkanal.h, lame.c, mpglib/layer3.c,
	  quantize.c:
	  disabled noise_shaping_stop=1, fixed jstereo default mode
	  problem,
	  updated frame analyzer

2000-03-25 18:16  cisc

	* takehiro.c:
	  Use INLINE define instead of inline.

2000-03-25 16:18  takehiro

	* lame.c:
	  enable noise_shaping==1 when high quality mode even without -X

2000-03-25 16:17  takehiro

	* takehiro.c:
	  debug seg.fault when big_values==0

2000-03-25 15:01  takehiro

	* l3bitstream.c, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  tables.c, takehiro.c:
	  best huffman divide now "more best" huffman coding.

2000-03-25 03:35  takehiro

	* Makefile, l3bitstream.c, l3side.h, lame.c, psymodel.c,
	  quantize-pvt.c, tables.c, takehiro.c:
	  new faster lossless coding routine(especially with -h)

2000-03-25 00:06  markt

	* get_audio.c, main.c, quantize-pvt.h, quantize.c, takehiro.c,
	  testcase.mp3:
	  added some code to best_scalefac_store:
	  if ix=0 in a scalefactor band, set the scalefactor to 0 also,
	  to save a few bits.

2000-03-24 23:02  markt

	* get_audio.c, machine.h:
	  _setmode stuff for windows

2000-03-24 22:37  markt

	* lame.h, mpglib/layer3.c, quantize-pvt.c:
	  fixed MPEG2 tables in mpglib,
	  added more accurate mp3buffer size bound in lame.h

2000-03-24 18:21  markt

	* quantize-pvt.c, quantize-pvt.h, quantize.c:
	  *** empty log message ***

2000-03-24 18:08  markt

	* Dll/MP3EncDll.dsp, lame.dsp, loopold.c:
	  removed loopold.c

2000-03-24 18:07  markt

	* Makefile, gtkanal.c, lame.c, lame.h, psymodel.c, psymodel.h,
	  quantize-pvt.c, quantize.c:
	  Removed old M/S stereo mode which uses L/R masking functions.
	  removed loopold.c and all that other "dual channel" stuff

2000-03-24 02:00  markt

	* VbrTag.c, lame.c:
	  MPEG2 VBR tag should now work

2000-03-24 01:38  markt

	* Dll/BladeMP3EncDLL.c, VbrTag.c, VbrTag.h, lame.c:
	  Xing MPEG2 headers work

2000-03-23 22:13  markt

	* Makefile:
	  uncommented options in makefile

2000-03-23 22:12  markt

	* API, Makefile, lame.c, lame.h, mpglib/common.c, util.c:
	  strange bug with -f fast-math option under gcc: it would compute
	  the mp3 framesize wrong, due to incorrect integer truncation.
	  Replaced implicit integer truncation with floor() to fix this.

2000-03-23 20:30  markt

	* Makefile, get_audio.c, lame.c, lame.h:
	  autoconvert moved into lame_encode_buffer

2000-03-23 20:02  markt

	* Makefile, lame.c, lame.h:
	  lame_encode_buffer_interleaved added to API

2000-03-22 17:14  markt

	* vbrquantize.c:
	  minor fixes

2000-03-22 04:15  markt

	* doc/html/history.html, version.h:
	  lame 3.66

2000-03-22 02:03  markt

	* USAGE, gtkanal.c, lame.c, machine.h, main.c, parse.c,
	  quantize-pvt.c:
	  float/double version of ASM quantize routines

2000-03-21 23:02  markt

	* brhist.c, brhist.h, gtkanal.c, gtkanal.h, l3bitstream-pvt.h,
	  l3bitstream.c, l3bitstream.h, lame.c, loopold.c, mpglib/layer3.c,
	  newmdct.c, newmdct.h, psymodel.c, psymodel.h, quantize-pvt.c,
	  quantize-pvt.h, quantize.c, quantize.h, reservoir.c, reservoir.h,
	  takehiro.c, util.c, util.h, vbrquantize.c:
	  replaced all "gf." by gfp->

2000-03-21 20:58  markt

	* lame.c:
	  fixed bug in lame_encode_finish: if user informs
	  lame that mp3buffer has size 0, lame will not check to make
	  sure the mp3buffer is big enough to store the encoded data

2000-03-19 22:29  afaber

	* Dll/BladeMP3EncDLL.c, lame.c:
	  Fixes to the lame DLL encoder:
	  If last frame is too short, set number of output bits to 0
	  instead of -1
	  Don't disable the bWriteVBRTag flag if input file is null

2000-03-19 20:07  cisc

	* get_audio.c:
	  Byteswapping (-x) didn't work when using LIBSNDFILE

2000-03-18 00:33  cisc

	* parse.c:
	  Fixed typo in define.

2000-03-18 00:32  cisc

	* amiga_mpega.c:
	  Erk, didn't spot the array/pointer relation, restored to previous
	  version.

2000-03-17 22:09  cisc

	* amiga_mpega.c:
	  b[] is a pointer, so supply addresses of pcm_l&r.

2000-03-17 19:29  markt

	* lame.c:
	  VBR was displaying MPEG2 bitrates - fixed.

2000-03-17 04:31  markt

	* machine.h:
	  added note to machine.h

2000-03-17 04:30  markt

	* doc/html/history.html, version.h:
	  uncommented gcc ASM code

2000-03-17 04:27  markt

	* quantize-pvt.c:
	  uncommended the GCC ASM code for quantize_xrpow()

2000-03-17 00:07  markt

	* doc/html/history.html:
	  updated history

2000-03-17 00:05  markt

	* amiga_mpega.c, get_audio.c, lame.h, mpglib/main.c,
	  mpglib/mpglib.dsp, version.h:
	  some tweaks to API for lame_decode_fromfile

2000-03-16 23:59  markt

	* Dll/BladeMP3EncDLL.c:
	  project file tweaks

2000-03-16 23:36  markt

	* get_audio.c, lame.dsp, lame.h, parse.c:
	  updated project files MSVC5

2000-03-16 23:19  markt

	* main.c:
	  updated MSVC6 branch project files

2000-03-16 19:42  markt

	* Dll/BladeMP3EncDLL.c:
	  update to bladedll

2000-03-16 19:38  markt

	* gtkanal.c, lame.h, mpglib/main.c:
	  one last change to lame_decode()...

2000-03-16 19:05  markt

	* get_audio.c, get_audio.h, gtkanal.c, gtkanal.h, lame.c, lame.h,
	  main.c, mp3rtp.c, mp3x.c, parse.c, util.c, util.h:
	  more and these really should be the last changes to the API
	  'gf' is no substantiated by the calling program, and passed
	  to all lame_routines(). This is so lame can eventually be
	  made re-entrant by putting all static vars in the gf struct.

2000-03-16 16:44  markt

	* VbrTag.c, loopold.c, machine.h, takehiro.c, timestatus.c, util.h:
	  strict ansi patches from Heiko.Eissfeld

2000-03-16 05:33  markt

	* USAGE, main.c:
	  updated USAGE

2000-03-16 05:28  markt

	* mpglib/mpglib.dsp:
	  revert back to MSVC5

2000-03-16 05:26  markt

	* main.c, mpglib/mpglib.dsp:
	  MSVC changes to compile lame.exe

2000-03-16 00:24  markt

	* timestatus.c:
	  fixed CPU/playing time display bug

2000-03-15 20:37  markt

	* encoder.h, l3bitstream.c, lame.h, parse.c, psymodel.c,
	  takehiro.c:
	  minor changes

2000-03-15 01:34  markt

	* filters.c, filters.h:
	  removed filters.c, filters.h

2000-03-15 01:20  markt

	* Dll/MP3EncDll.dsp, Dll/MP3EncDll.dsw:
	  reverting back to MSVC5

2000-03-15 01:14  markt

	* Dll/BladeMP3EncDLL.c, Dll/MP3EncDll.dsp, Dll/MP3EncDll.dsw,
	  parse.c:
	  updated bladeenc dll

2000-03-14 21:01  markt

	* lame.c, loopold.c, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  quantize.h, reservoir.c, reservoir.h, util.c, util.h,
	  vbrquantize.c:
	  removed fr_ps struct

2000-03-14 20:45  markt

	* brhist.c, brhist.h, l3bitstream.c, l3bitstream.h, lame.c, lame.h,
	  psymodel.c, psymodel.h, quantize-pvt.c, quantize-pvt.h,
	  quantize.c, reservoir.c, timestatus.h, util.c, util.h,
	  vbrquantize.c:
	  removed "info" sturct. Removing fr_ps struct

2000-03-14 07:21  markt

	* brhist.c, l3bitstream.c, lame.c, lame.h, loopold.c, psymodel.c,
	  quantize-pvt.c, quantize.c, reservoir.c, timestatus.c,
	  timestatus.h, util.c, util.h:
	  removed more info-> variables

2000-03-14 06:36  markt

	* lame.c, lame.h:
	  removing some of the info-> variables

2000-03-10 23:21  cisc

	* amiga_mpega.c:
	  Had to make a dummy break-function to make atexit() work.

2000-03-10 23:18  markt

	* loopold.c:
	  segfault fixed in loopold.c when count[] was zero for one channel
	  and non-zero for the other

2000-03-10 20:56  markt

	* API, Makefile, gtkanal.c, lame.c, lame.h, version.c:
	  added lame_version to the API

2000-03-10 20:13  markt

	* Dll/BladeMP3EncDLL.c:
	  removed some unnecessary #includes

2000-03-10 19:48  markt

	* Dll/MP3EncDll.dsp:
	  added #define LAMEPARSE to Blade project files

2000-03-10 19:38  markt

	* Dll/BladeMP3EncDLL.c, Dll/MP3EncDll.dsp:
	  Blade dll updated to handle new interface
	  Albert: Sorry if I broke anything, the old version is still in
	  CVS :-)

2000-03-10 16:01  markt

	* Dll/BladeMP3EncDLL.c:
	  BladeMP3 changes

2000-03-10 01:41  markt

	* Dll/BladeMP3EncDLL.c:
	  some changes to update BaldeMP3encdll

2000-03-09 19:42  markt

	* quantize-pvt.c, takehiro.c:
	  quantize_xrpow_ISO also uses tablelookup, so it also needed the
	  "xr too big" check

2000-03-09 18:05  markt

	* HACKING, get_audio.c, gtkanal.c, lame.c, lame.h, main.c, parse.c:
	  *** empty log message ***

2000-03-08 00:16  markt

	* Makefile, parse.c:
	  Robert's presets restored. polyphase filters now except
	  filterwidth
	  arguments.

2000-03-08 00:05  markt

	* lame.c, lame.h, newmdct.c:
	  polyphase filter now supports lowpasswidth and highpasswidth
	  option

2000-03-07 23:10  cisc

	* API:
	  Added AMIGA_MPEGA #define, and fixed a few minor spelling
	  errors... ;)

2000-03-07 22:56  cisc

	* amiga_mpega.c:
	  Made sure lame_decode_fromfile() won't return an empty frame
	  (unless there's an error).

2000-03-07 18:33  markt

	* API, Makefile, get_audio.c, lame.c, main.c, parse.c:
	  minor tweeks to compile w/o decoding library

2000-03-07 08:01  sbellon

	* lame.c:
	  fixed RISC OS FPE support

2000-03-06 23:28  markt

	* API:
	  API description

2000-03-06 19:55  markt

	* lame.dsp:
	  *** empty log message ***

2000-03-06 19:53  markt

	* Dll/MP3EncDll.dsp, Makefile, Makefile.MSVC, get_audio.c,
	  get_audio.h, id3tag.h, lame.c, lame.dsp, lame.h, mpglib/common.c,
	  mpglib/dct64_i386.c, mpglib/decode_i386.c, mpglib/interface.c,
	  mpglib/layer3.c, mpglib/main.c, mpglib/tabinit.c, parse.c:
	  More library work.
	  Sorry to all the non-unix developers out there, but this will
	  break all project files. (but we are very close to the end of
	  the modularization work).
	  
	  New code needs the following #defines set somewhere:
	  
	  #define HAVEMPGLIB to use mpglib's mp3 *decoding* capibility
	  (or AMEGA_MPEGA on amiga systems, and dont define HAVEMPGLIB)
	  #define BRHIST to allow the display of the VBR historgram
	  #define LIBSNDFILE to use Erik de Castro Lopo's libsndfile
	  #define LIBSNDLAME to use LAME's minimial internal sndfile I/O
	  #define LAMEPARSE to use LAME's command line parsing/option
	  setting routines
	  #define HAVEGTK to compile in support for the GTK mp3 frame
	  analyzer

2000-03-06 15:16  markt

	* filters.c, lame.c, lame.h, newmdct.c, parse.c, testcase.mp3:
	  polyphase filters are now working!
	  some tuning of polyphase fiters.
	  removed MDCT filters, sfb21 cutoff.

2000-03-06 03:59  markt

	* gtkanal.c, lame.c, newmdct.c, psymodel.c:
	  new polyphase filters

2000-03-05 19:07  markt

	* psymodel.c:
	  Gabriel: L/R block matching fix

2000-03-05 18:37  markt

	* doc/html/id3.html, id3tag.c, id3tag.h, lame.c, lame.h, newmdct.c,
	  parse.c:
	  id3v1.1 tags from Don Melton

2000-03-04 17:19  markt

	* get_audio.c, lame.c:
	  resampling moved to fill_buffer_resample() in lame.c,
	  which has linear and 3rd order resampling

2000-03-04 15:31  markt

	* amiga_mpega.c, get_audio.c, lame.c, main.c, quantize-pvt.c,
	  quantize.c, vbrquantize.c:
	  patch from kimmo to remove more c++ comments,
	  fill_buffer_linear_resample code

2000-03-04 05:13  markt

	* lame.c:
	  fill_buffer routines

2000-03-03 19:25  markt

	* encoder.h, mp3rtp.c:
	  mp3rtp uses new, simplified lame_encode_buffer interface

2000-03-03 19:01  markt

	* encoder.h, lame.c, lame.h, main.c, testcase.mp3:
	  added lame_encode_buffer, the new, prefered encoding interface

2000-02-29 21:19  robert

	* parse.c:
	  minor change, exit if wrong preset type was given

2000-02-29 19:54  robert

	* fft.c:
	  minor speed up (at least for gcc)

2000-02-29 18:00  sbellon

	* lame.c:
	  moved RISC OS floating point exception part into debugging area

2000-02-27 19:11  robert

	* quantize-pvt.c, quantize-pvt.h, quantize.c:
	  removed unused parameter

2000-02-27 18:36  robert

	* psymodel.c:
	  code clean up, giving an overall speed up of 3-4%

2000-02-25 07:00  markt

	* lame.c:
	  minor cleanup

2000-02-24 15:30  robert

	* loopold.c, quantize-pvt.c, quantize-pvt.h, quantize.c:
	  code cleanup, removed unused parameter

2000-02-23 02:40  cisc

	* amiga_mpega.c:
	  Fixed up the cleanup function.

2000-02-22 23:21  robert

	* quantize-pvt.c, testcase.mp3:
	  switched back to original bin_search_stepsize routine

2000-02-22 03:14  markt

	* quantize-pvt.c, testcase.mp3:
	  RH bin_search_stepsize now the default

2000-02-22 00:54  cisc

	* amiga_mpega.c:
	  Oh, nevermind, wasn't very efficient here, proper place would be
	  in sndfile.lib

2000-02-22 00:15  cisc

	* lame.h, main.c:
	  Minor fixes.

2000-02-21 23:58  cisc

	* amiga_mpega.c:
	  Make sure input-file is closed when CTRL-C is received.

2000-02-21 23:05  markt

	* Makefile, doc/html/history.html, get_audio.c, gtkanal.c,
	  ieeefloat.c, l3bitstream.c, main.c, mpglib/main.c, util.c:
	  some 64bit DEC Alpha patches

2000-02-21 22:40  robert

	* quantize-pvt.c:
	  speed up of bin_search_stepsize, the bin_search is about 20%
	  faster, resulting in an average speed up of 2-3%, disabled by
	  default as it breaks the testcase

2000-02-21 22:39  sbellon

	* main.c:
	  further RISC OS support added

2000-02-21 00:21  markt

	* lame.c:
	  minor

2000-02-20 23:36  markt

	* get_audio.c, lame.c, lame.h, main.c, mp3rtp.c, parse.c:
	  num_samples /eof fix, ABORTFP fix, stdout mode setting fix

2000-02-20 20:31  markt

	* mpglib/common.c:
	  better error message

2000-02-20 20:03  robert

	* get_audio.c:
	  removed warning about fill up with zeros again, it is common at
	  the end of an input file

2000-02-20 19:26  robert

	* Makefile, quantize-pvt.c, quantize-pvt.h, quantize.c:
	  code cleanup

2000-02-20 18:45  robert

	* psymodel.c:
	  code cleanup

2000-02-20 14:09  robert

	* formatBitstream.c:
	  fixed assertion failure while flushing bitstream at end of
	  encoding

2000-02-20 14:08  robert

	* get_audio.c:
	  applied patch from Jeremy Hall to fix stdin problem

2000-02-19 13:33  afaber

	* fft.c, formatBitstream.c, formatBitstream.h, get_audio.c,
	  ieeefloat.c, l3bitstream.c, machine.h, mpglib/mpg123.h,
	  quantize-pvt.c, quantize.c, takehiro.c:
	  Fixed many warning messages when compiling with MSVC

2000-02-19 12:15  afaber

	* VbrTag.c:
	  Applied patch from Mathew Hendry to avoid divide by zero

2000-02-18 16:21  robert

	* quantize-pvt.c:
	  Mathew Hendry: line 813 should read "#define USE_MSC_ASM" and not
	  "#define USE_MSC_AMS"

2000-02-18 16:20  robert

	* Makefile.MSVC:
	  Mathew Hendry: "stray space character following the \" patched

2000-02-18 16:07  robert

	* lame.c:
	  Mathew Hendry's CYGWIN gcc FPU control patch

2000-02-18 06:16  markt

	* Makefile, gtkanal.c, lame.c, testcase.mp3:
	  changed sizeof frame analyzer so it fits on a 800x600 screen.
	  make --nores also set info->padding = 0. Avoids problem with
	  small changes in frac_SpF causing different padding patterns and
	  mp3 file differences.

2000-02-17 20:12  robert

	* loopold.c, quantize-pvt.h, quantize.c:
	  code clean-up, it seems the SIGFPE exceptions are gone

2000-02-16 22:42  cisc

	* machine.h:
	  Fixed INLINE define for SAS/C.

2000-02-16 21:18  cisc

	* main.c:
	  Include machine.h so that needed prototypes get added.

2000-02-16 07:06  markt

	* quantize-pvt.c, quantize-pvt.h, quantize.c:
	  new quantize xrpow for PowerPC

2000-02-16 03:56  markt

	* lame.c, quantize-pvt.c:
	  patches for risc_os

2000-02-15 22:16  robert

	* quantize.c:
	  forgot to save plotting data for best quantization in
	  VBR_iteration_loop

2000-02-15 17:12  robert

	* quantize-pvt.h, quantize.c:
	  VBR_iteration_loop speedup, removed extra quantizations

2000-02-15 14:51  markt

	* Makefile:
	  *** empty log message ***

2000-02-15 14:40  markt

	* lame.c:
	  some non-initilized memory patches

2000-02-15 02:01  robert

	* quantize-pvt.c:
	  bug in ms_convert fixed when source and target are the same

2000-02-14 21:48  robert

	* quantize-pvt.c:
	  Mathew Hendry's quantize-xrpow gcc asm patch

2000-02-14 21:47  robert

	* quantize.c:
	  missing line added

2000-02-14 21:01  afaber

	* lame.c:
	  Intialize float arrays to avoid floating point errors when
	  ABORTFP is set
	  Added floating point trapping code for MSVC compiler

2000-02-14 20:53  afaber

	* psymodel.c:
	  Intialize tot_ener to avoid floating point errors

2000-02-14 05:50  markt

	* Makefile, loopold.c, parse.c, quantize-pvt.c, quantize-pvt.h,
	  quantize.c:
	  merged takehiro2 branch back into main branch.
	  there is still a floating point exception which I cannot track
	  down.

2000-02-14 04:53  markt

	* doc/html/contributors.html, doc/html/examples.html,
	  doc/html/id3.html, doc/html/index.html, doc/html/switchs.html,
	  psymodel.c, takehiro.c:
	  updated html from Gabriel

2000-02-13 18:52  robert

	* fft.c, fft.h, psymodel.c:
	  modified energy calculation, fast MS fft patch should work now

2000-02-11 17:21  markt

	* Makefile, quantize.c, reservoir.c:
	  fixed scalefac not-initialized problem. This would happen when
	  xr = 0 and outer_loop() was skipped.

2000-02-11 16:34  markt

	* psymodel.c:
	  m/s fft patch was not working.

2000-02-10 19:00  robert

	* parse.c:
	  added help info for --preset option and shuffled help text to
	  bring it more in line

2000-02-10 16:33  robert

	* fft.c, fft.h, psymodel.c:
	  applied Takehiro Tominaga's fast MS fft patch

2000-02-10 14:22  markt

	* doc/html/history.html:
	  updated history file from web page

2000-02-10 06:43  markt

	* version.h:
	  updated version

2000-02-10 06:15  markt

	* lame.c, parse.c, version.c, version.h:
	  added version info to help output

2000-02-10 05:42  markt

	* psymodel.c, testcase.mp3:
	  fixed bug in s3_s spreading function calculation.

2000-02-10 05:35  markt

	* loopold.c, quantize-pvt.h, quantize.c:
	  undo some of takehiro's speedup improvements until we can track
	  down floating point exceptions

2000-02-10 05:23  markt

	* Makefile, brhist.c, brhist.h, lame.c, loopold.c, psymodel.c,
	  quantize-pvt.c, quantize-pvt.h, quantize.c:
	  some bug fixes:
	  
	  psymodel partition band calculations done *before*
	  the spreading function need to use npart_l_orig. After spreading
	  function,
	  use npart_s_orig.
	  
	  3 cases of scalefac->l[i][sfb] which should be
	  scalefac->l[sbf][i]
	  
	  VBR min & max bitrate index setting problem fixed

2000-02-08 17:20  afaber

	* lame.c:
	  Guarded disp_brhist with #ifdef BRHIST, in order to compile with
	  MSVC

2000-02-06 10:58  takehiro

	* loopold.c, psymodel.c, quantize-pvt.c, quantize-pvt.h,
	  quantize.c:
	  new spread matrix calculation is now default(faster!)

2000-02-06 05:59  markt

	* gtkanal.c:
	  Iwasa's changes to gtkanal.c

2000-02-05 12:32  robert

	* lame.c, lame.h, parse.c, util.c, util.h:
	  applied Ingo Saitz command line usage patch

2000-02-05 05:16  markt

	* Makefile.MSVC, brhist.c, lame.c, loopold.c, machine.h, parse.c,
	  quantize.c:
	  minor code cleanup

2000-02-04 23:49  robert

	* parse.c:
	  added "--preset type" option to offer users some usefull
	  presettings
	  that should shorten the commandline for most cases

2000-02-04 21:42  afaber

	* lame.c, lame.dsp:
	  Fixed WIN32 compiler errors

2000-02-04 21:26  robert

	* parse.c:
	  added '--help' option to allow DOS users to scroll thru the usage
	  page with "lame --help | more"

2000-02-04 13:57  takehiro

	* loopold.c, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  takehiro.c:
	  speed up hack on the array index.

2000-02-04 13:49  takehiro

	* psymodel.c:
	  and more faster calculation

2000-02-04 02:29  markt

	* lame.h, quantize-pvt.c:
	  minor editing

2000-02-03 15:43  takehiro

	* psymodel.c:
	  some assertion bug(?) has removed

2000-02-03 15:31  takehiro

	* psymodel.c:
	  more faster calculation

2000-02-02 22:30  afaber

	* mpglib/main.c:
	  Added MPGLIB library

2000-02-02 22:25  takehiro

	* psymodel.c:
	  faster spread function calculation

2000-02-02 17:12  markt

	* get_audio.h, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  takehiro.c, vbrquantize.c:
	  minor code cleanup

2000-02-02 14:58  takehiro

	* i386/fft.nas:
	  merging interface

2000-02-02 10:31  markt

	* Makefile:
	  testcase update

2000-02-02 10:25  markt

	* Makefile, filters.c, lame.c, quantize-pvt.c, quantize.c:
	  fixed some minor bugs in new scalefac data structure code

2000-02-01 14:09  takehiro

	* l3side.h, lame.c, loopold.c, psymodel.c, psymodel.h,
	  quantize-pvt.c, quantize-pvt.h, quantize.c, quantize.h,
	  vbrquantize.c:
	  code clean up. changed definition of structure to optimize array
	  index calculation

2000-02-01 11:26  takehiro

	* Makefile, fft.c, filters.c, filters.h, gtkanal.c,
	  l3bitstream-pvt.h, l3bitstream.c, l3bitstream.h, l3side.h,
	  lame.c, loopold.c, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  quantize.h, tables.c, tables.h, takehiro.c, vbrquantize.c:
	  scalefactor's structure changed

2000-01-31 22:21  robert

	* quantize-pvt.c, quantize.c:
	  little VBR tweaks

2000-01-31 18:56  markt

	* get_audio.c, lame.c, lame.h, main.c, mp3rtp.c, mp3x.c, version.h:
	  updated comments, added lame_close_infile()

2000-01-31 18:22  markt

	* lame.c, lame.h, main.c, mp3rtp.c, mp3x.c:
	  more modularizaton, fixes to mp3rtp

2000-01-31 05:49  markt

	* mlame:
	  spelling error

2000-01-30 23:00  cisc

	* fft.c:
	  Use INLINE define instead of inline.

2000-01-30 22:53  takehiro

	* i386, i386/fft.nas, i386/fft3dn.nas, i386/fftfpu.nas,
	  i386/fftsse.nas, i386/ffttbl.nas:
	  initial check in of asm code from GOGO
	  not working :)

2000-01-30 22:19  markt

	* get_audio.c, lame.c, lame.h, parse.c:
	  renamed highq quality

2000-01-30 22:01  markt

	* mp3rtp.c:
	  added Felix's email to mp3rtp.c code :-)

2000-01-30 20:44  markt

	* lame.h:
	  minor editing

2000-01-30 20:27  markt

	* lame.c, lame.h, loopold.c, parse.c, psymodel.c, quantize-pvt.c,
	  quantize.c, takehiro.c:
	  cleaned up highq option.
	  highq=0 (best quality)... 9(worst quality)
	  values supported so far:
	  highq=9 (turned on with -f)
	  highq=5 (default)
	  highq=2 (turned on with -h)

2000-01-30 19:34  markt

	* lame.c, lame.h, machine.h, parse.c, quantize-pvt.c, takehiro.c:
	  fixed bug in overflow check in count_bits

2000-01-30 16:19  robert

	* Makefile, quantize-pvt.c, quantize-pvt.h, quantize.c:
	  introduced new RH_ATH compile switch and splitted
	  RH_masking into RH_SIDE_VBR, RH_SIDE_CBR and
	  RH_QUALITY_CONTROL=n (n=1,2)
	  
	  RH_SIDE_CBR turns off side channel reduction for CBR
	  RH_SIDE_VBR turns off side channel reduction for VBR
	  RH_QUALITY_CONTROL=1 plus above switches equals RH_masking
	  RH_QUALITY_CONTROL=2 is tuned to be used with RH_ATH
	  RH_ATH changes the definition of distortion a bit:
	  a)PSY: noise over the PSY model threshold as average
	  quantization noise over PSY model threshold
	  b)ATH: max quantization noise of a single freq line over
	  the ATH threshold of that freq line
	  the resulting distortion is the min of a) and b)

2000-01-30 10:54  markt

	* parse.c:
	  parse_args and related code moved to parse.c

2000-01-30 10:53  markt

	* brhist.c, brhist.h:
	  vbr histogram code in seperate file

2000-01-30 10:50  markt

	* Dll/BladeMP3EncDLL.c, Dll/MP3EncDll.dsp, Makefile, TODO,
	  get_audio.c, gtkanal.c, id3tag.c, id3tag.h, lame.c, lame.h,
	  main.c, mp3rtp.c, mp3x.c, mpglib/main.c, util.c, util.h:
	  more modularization work. Calling program is now responsible for
	  mp3 file output

2000-01-30 07:41  markt

	* l3side.h, loopold.c, machine.h, psymodel.c, quantize-pvt.c,
	  quantize-pvt.h, quantize.c, takehiro.c, vbrquantize.c:
	  #ifdef NOPOW removed, replaced by table lookup. You can still
	  go back to pow() function or exp() by changing the POW20 and
	  IPOW20
	  macros in machine.h

2000-01-30 06:17  takehiro

	* l3side.h, loopold.c, psymodel.c, quantize-pvt.c, quantize-pvt.h,
	  quantize.c, takehiro.c:
	  use lookup table to calculate pow(2.0, xx).(when not defined
	  NOPOW)

2000-01-30 06:09  takehiro

	* fft.c:
	  faster FFT implementation

2000-01-28 00:57  cisc

	* lame.c:
	  Fixed -m bug when bitrate>=160 (gf.mode_fixed was not set).

2000-01-27 17:16  markt

	* lame.h:
	  comments about varius quality options

2000-01-27 08:27  markt

	* quantize-pvt.c:
	  Fixed bug in scale_bitcount_lsf which was severely limiting the
	  possible scalefacters used by MPEG2. (Lionel Bonnet)

2000-01-27 07:49  markt

	* Makefile, get_audio.c, get_audio.h, gtkanal.c, machine.h:
	  minor changes

2000-01-27 05:05  markt

	* fft.c, fft.h, psymodel.c, testcase.mp3:
	  Takehiro's fft_short and fft_long now the defaults.

2000-01-27 04:35  markt

	* rtp.c:
	  more updates from Felix

2000-01-27 03:52  cisc

	* loopold.c, psymodel.c, quantize-pvt.c, quantize.c, takehiro.c,
	  util.h, vbrquantize.c:
	  Added NOPOW define, will replace certain pow()s with optimized
	  variant.

2000-01-27 02:06  cisc

	* get_audio.c:
	  Removed faulty undef of bitwidth.

2000-01-26 12:37  takehiro

	* fft.c:
	  window[] and window_s[]'s buffer overrun is corrected

2000-01-26 10:27  markt

	* filters.c, lame.c, lame.h:
	  more bugs in specifigy a lowpass filter fixed. --voice mode
	  recast in terms of lowpass filter and other options

2000-01-26 08:39  markt

	* psymodel.c:
	  removed two c++ commends (by request)

2000-01-26 08:00  markt

	* lame.c:
	  moved frameNum=0 from lame_init to lame_init_params

2000-01-26 07:44  markt

	* Dll/BladeMP3EncDLL.c, Dll/MP3EncDll.dsp, HACKING, get_audio.c,
	  get_audio.h, globalflags.h, gtkanal.c, lame.c, lame.h, main.c,
	  mp3rtp.c, mp3x.c:
	  more modularizatin work. LAME can now be used as a true library,
	  without the need to construct simulated argument lists.
	  See HACKING for some details, and mp3rtp.c for an example.

2000-01-25 17:50  markt

	* get_audio.c, get_audio.h, globalflags.h, lame.c:
	  major work down towards libmp3lame.a: split out parse_args and
	  parameter setting routines. might have broken some things...

2000-01-25 17:48  markt

	* rtp.h:
	  bug fix from Felix

2000-01-20 23:21  robert

	* lame.c:
	  sfb21 parameter messages added

2000-01-20 08:03  markt

	* lame.c:
	  tweaked filter and other encoding parameter messages

2000-01-20 07:21  markt

	* gtkanal.c, gtkanal.h, mpglib/layer3.c:
	  fixed gtk scfsi printout - mpglib/layer3.c seems correct now.

2000-01-20 06:09  markt

	* Makefile, TODO, mpglib/layer3.c:
	  scfsi fixes to mpglib from Iwasa Kazmi

2000-01-19 22:11  takehiro

	* mpglib/layer3.c:
	  scalefactor for short block last sfb(12) was wrong.

2000-01-18 18:51  afaber

	* l3bitstream.c:
	  Made my bug fixes against version 1.9 instead of a previous
	  version

2000-01-17 22:31  afaber

	* l3bitstream.c:
	  Guarded BF_FlushBitstream function call if PartHoldersInitialized
	  equals zero.
	  This to avoid GPF when III_Flushbitstream is called when no data
	  is encoder

2000-01-16 06:49  markt

	* lame.c, lame.h, main.c:
	  cleaned up main.c

2000-01-15 20:33  markt

	* loopold.c, takehiro.c, testcase.mp3:
	  added best_scalefac_store to non -h code also

2000-01-15 14:35  takehiro

	* psymodel.c:
	  little bit optimization

2000-01-14 13:05  takehiro

	* takehiro.c:
	  debug message is removed...

2000-01-14 12:57  takehiro

	* l3bitstream.c, loopold.c, psymodel.h, quantize-pvt.c, quantize.c,
	  takehiro.c:
	  to make it use scfsi

2000-01-14 07:47  markt

	* doc/html/history.html:
	  *** empty log message ***

2000-01-14 07:34  markt

	* version.h:
	  version 3.61

2000-01-14 06:35  markt

	* lame.c, quantize.c:
	  filters were incorrectly being used under VBR

2000-01-14 05:54  markt

	* USAGE, lame.c:
	  more on the -k option

2000-01-14 05:46  markt

	* USAGE, lame.c, mpglib/main.c:
	  -k option will disable *all* filtering and frequency cutoffs

2000-01-13 20:41  robert

	* Makefile, quantize.c:
	  more VBR (RH_masking) finetuning
	  now -V3 should give the same result as -V6 before

2000-01-13 17:21  markt

	* lame.c, psymodel.h:
	  fixed noshort option, fixed l3psycho_anal.h

2000-01-13 16:26  takehiro

	* filters.c, get_audio.c, l3bitstream.c, lame.c, loopold.c,
	  newmdct.c, newmdct.h, psymodel.c, quantize-pvt.c, quantize-pvt.h,
	  quantize.c, reservoir.c, reservoir.h, takehiro.c, util.c, util.h,
	  vbrquantize.c:
	  moved info.stereo into gf.stereo

2000-01-13 07:43  markt

	* loopold.c, quantize.c:
	  got rid of old, no longer used preemph code

2000-01-13 02:07  markt

	* filters.c, get_audio.c, globalflags.h, gtkanal.c, lame.c,
	  mpglib/main.c, psymodel.c, quantize-pvt.c, quantize-pvt.h,
	  quantize.c, takehiro.c, vbrquantize.c:
	  moved the rest of the global variables into gf struct

2000-01-12 22:38  afaber

	* ., VbrTag.c, lame.dsp:
	  fixed a few casting warnings
	  Updated .cvsignore to ignore output of testcase

2000-01-12 14:30  takehiro

	* filters.c, globalflags.h, l3bitstream.c, lame.c, newmdct.c,
	  newmdct.h, quantize-pvt.c, quantize-pvt.h, reservoir.c,
	  takehiro.c, util.c, vbrquantize.c:
	  more simple & fast scalefac_scale use
	  and mode_gr is moved into gf structure.

2000-01-11 20:23  cisc

	* amiga_mpega.c:
	  Fixed timestatus bug (forgot nsamp variable).

2000-01-11 05:46  markt

	* gtkanal.c, lame.c:
	  changed axis in gtk flag

2000-01-10 16:37  cisc

	* VbrTag.h, get_audio.c, machine.h, util.c:
	  Removed AMIGA_ASYNCIO, custom modified libsndfile worked better.

2000-01-10 09:34  takehiro

	* mpglib/layer3.c:
	  debug for scalefactor reading with scfsi.

2000-01-10 03:44  cisc

	* get_audio.c:
	  Disabled open/close of mp3 when AMIGA_MPEGA and LIBSNDFILE is
	  enabled.

2000-01-10 02:44  cisc

	* amiga_mpega.c:
	  MPGLIB replacement using mpega.library. (AmigaOS)

2000-01-10 02:40  cisc

	* get_audio.c, mpglib/main.c:
	  Removed AMIGA_ASYNCIO stuff for mpglib now that we got
	  AMIGA_MPEGA.

2000-01-10 02:07  cisc

	* get_audio.c, lame.h:
	  Added mpega.library support for decoding mpeg audio, enable with
	  AMIGA_MPEGA. (AmigaOS)

2000-01-10 00:45  markt

	* loopold.c, quantize.c:
	  fixed bug in loopold: scalesave loop was using SBMAX instead of
	  SBPSY,
	  and exceeding bounds.

2000-01-09 23:11  markt

	* Dll/BladeMP3EncDLL.c, doc/html/examples.html, doc/html/id3.html,
	  mpglib/layer3.c:
	  globalflags update

2000-01-09 23:10  markt

	* get_audio.c, globalflags.h, gtkanal.c, gtkanal.h, lame.c,
	  loopold.c, main.c, mp3x.c, psymodel.c, quantize-pvt.c,
	  quantize.c, reservoir.c, vbrquantize.c:
	  moved more globalflags into the gf. struct

2000-01-09 22:56  cisc

	* VbrTag.h, get_audio.c, machine.h, mpglib/main.c, util.c:
	  Added usage of asyncio.library for improved I/O, enable with
	  AMIGA_ASYNCIO. (AmigaOS)

2000-01-09 22:26  markt

	* doc/html/history.html, quantize.c, testcase.mp3:
	  updated validation testcase.

2000-01-09 20:59  markt

	* Makefile, version.h:
	  release 3.60beta

2000-01-09 20:55  markt

	* testcase.mp3, testcase.wav:
	  new test case: testcase.wav and 'official' output testcase.mp3
	  'make test' will perform a comparison. see makefile for details

2000-01-09 20:52  markt

	* Makefile:
	  new validation procedure

2000-01-09 20:21  markt

	* doc/html/history.html, doc/html/index.html,
	  doc/html/switchs.html:
	  updated web pages from Gabriel

2000-01-09 16:08  cisc

	* mpglib/main.c:
	  VbrTag.h wasn't included on Amiga

2000-01-09 16:07  cisc

	* fft.c:
	  Removed INLINE from fht (will get inlined as a local function
	  anyway)

2000-01-09 10:54  takehiro

	* l3bitstream.c, l3side.h, loopold.c, quantize-pvt.c,
	  quantize-pvt.h, quantize.c, takehiro.c:
	  All Huffman code search algorithm is implemented.
	  (-h option to enable this)
	  more slower, but better quality.

2000-01-09 07:51  markt

	* Makefile, psymodel.c, quantize.c:
	  AAC masking function now the default. This represents less
	  masking then the old masking function.

2000-01-09 07:34  markt

	* gtkanal.c, gtkanal.h, lame.c, psymodel.c, psymodel.h, quantize.c:
	  compute mid/side energy rations in addition to masking ratios.
	  in the future we might use this for side channel bit allocation.
	  it is not used at all right now.

2000-01-09 04:11  markt

	* Makefile, psymodel.c, quantize.c:
	  someone mistakenly uncommented the call to preemphasis().
	  this routine should *not* be used. preemph is now taken care of
	  in scale_bitcount.

2000-01-08 17:57  robert

	* psymodel.c:
	  define AAC_TMN_NMT at compile time to activate AAC tone masking
	  noise
	  and noise masking tone values

2000-01-08 17:56  robert

	* quantize.c:
	  tuned VBR code, inactive until you define RH_masking at compile
	  time

2000-01-08 17:54  robert

	* Makefile:
	  added suggested compiler options for gcc 2.95.2, uncomment them
	  to activate them

2000-01-08 17:52  robert

	* PRESETS.draft:
	  changed frequencies to Lames format

2000-01-07 06:40  markt

	* HACKING, Makefile:
	  updated HACKING notes

2000-01-07 06:13  markt

	* get_audio.c, globalflags.h, lame.c, loopold.c, psymodel.c,
	  quantize-pvt.c, quantize.c, reservoir.c:
	  Robert's cw_lower_limit, upper_limit code back in. Default is
	  compute
	  cw[] up to 8.9Khz. set with --cwlimit <freq>
	  
	  started putting global variables into global_flags struct.

2000-01-06 05:12  markt

	* mp3rtp.c:
	  more rtp updates

2000-01-06 04:58  markt

	* lame.c, mp3rtp.c, rtp.c:
	  more rtp updates

2000-01-05 23:52  robert

	* quantize.c:
	  VBR: masking lower was not set correctly
	  it was only set for sloppy mode, fixed now

2000-01-05 17:40  afaber

	* mpglib/mpglib.h:
	  Fixed BOOL definition conflict for Win32 systems

2000-01-05 16:47  markt

	* Makefile, mp3rtp.c, rtp.c, rtp.h:
	  added mp3rtp code. uses lame library, but not part of normal lame
	  encoder

2000-01-05 10:04  markt

	* VbrTag.c, VbrTag.h, get_audio.c, l3side.h, lame.h, mpglib/main.c:
	  Check for Xing header in mp3 input files

2000-01-05 06:29  markt

	* version.h:
	  updated version

2000-01-05 06:20  markt

	* Makefile, psymodel.c:
	  norm_l, norm_s table data replaced by formulas.

2000-01-03 05:15  markt

	* INSTALL, Makefile, doc/html/contributors.html,
	  doc/html/index.html, doc/html/lame.css, doc/html/node6.html,
	  doc/html/node7.html, doc/html/switchs.html, encoder.h, gtkanal.c,
	  lame.c, main.c:
	  new html from Gabriel. Fixed but that was causing the first frame
	  to
	  always be zero. changed encoder delay to 800 to fix some other
	  problems
	  and remove the buffering of the first frame in lame_encode().

1999-12-29 15:34  takehiro

	* doc/html/contributors.html, doc/html/history.html,
	  doc/html/index.html, doc/html/node6.html, doc/html/node7.html,
	  doc/html/switchs.html:
	  updated html doc written by Gabriel Bouvigne.

1999-12-28 22:02  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.h, Dll/MP3EncDll.dsp,
	  lame.dsp, mpglib/mpglib.dsp:
	  Some fixes to the Encoder DLL:
	  use -h instead of -k flag for high qualtiy
	  The voice option was set in HQ mode, which is obviously not
	  correct

1999-12-26 14:48  takehiro

	* l3bitstream.c, lame.c:
	  some foolish bug is removed :)

1999-12-26 09:51  takehiro

	* globalflags.h, lame.c, quantize-pvt.c:
	  new option "--noath" to disable the ATH masking

1999-12-26 09:50  takehiro

	* .:
	  ignore mp3x binary exe.

1999-12-23 12:27  takehiro

	* quantize-pvt.c:
	  guess wiser initial StepSize for bin_search_stepsize

1999-12-22 17:38  markt

	* Makefile, psymodel.c:
	  AAC NMT/TMN values (commented out)

1999-12-22 08:03  markt

	* quantize-pvt.c:
	  back to oringal (ave_noise < ATH) definition of analog silence
	  for
	  VBR mode.

1999-12-22 07:39  markt

	* psymodel.c:
	  added AAC spreading function (disabled by default)

1999-12-21 08:18  markt

	* Makefile, fft.c, l3side.h, lame.c, psymodel.c, psymodel.h,
	  quantize-pvt.c, quantize.c:
	  Bug fix in Mid/Side masking thresholds

1999-12-19 21:33  afaber

	* Dll/BladeMP3EncDLL.c:
	  Set home page link to Lame home page

1999-12-19 01:47  markt

	* Makefile, globalflags.h, lame.c, psymodel.c:
	  went back to ISO layer III recommended cw formulas

1999-12-19 01:15  robert

	* globalflags.h, lame.c, psymodel.c:
	  patched GPSYCHO, enable with --psy-model 0

1999-12-17 04:24  markt

	* USAGE, globalflags.h, lame.c, reservoir.c:
	  added the --nores option to disable the bitreservoir. only
	  usefull
	  in special circumstances

1999-12-16 05:24  markt

	* quantize-pvt.c:
	  changed analog silence switch to use max_ener rather than
	  ave_ener.
	  only affects VBR mode: when analog silence is detected, the min
	  bitrate
	  is allowed to drop to 32kbs even if set higher with -b

1999-12-14 23:51  cisc

	* fft.c:
	  Changed M_PI to PI. (Not everyone has M_PI, but PI is always
	  defined in util.h)

1999-12-14 22:44  robert

	* PRESETS.draft:
	  first sketch of what could become PRESETS for LAME

1999-12-14 22:40  robert

	* lame.c, quantize.c:
	  added new experimentalX quant_compare -X6 mode

1999-12-14 04:38  markt

	* Makefile, fft.c, fft.h, psymodel.c:
	  Takehiro's fft's back in. fft_short2(), fft_long2() will call
	  original
	  fft's (with one minor change: 0 protection for ax[] and bx[] was
	  not
	  needed and has been removed). Takehiro's routines are fft_short()
	  and
	  fft_long(). They dont give bit-for-bit identical answers, and I
	  still
	  want to track this down before making them the defaults. .

1999-12-14 00:19  robert

	* lame.c:
	  changed filter options

1999-12-13 21:56  markt

	* version.h:
	  updated version to 3.58

1999-12-12 21:14  robert

	* lame.c:
	  grrr, mist some filter comandline combinations :(

1999-12-12 19:22  robert

	* lame.c:
	  fixed filter commandline parameter inconsistencies

1999-12-12 19:02  markt

	* filters.c, newmdct.c, util.h:
	  removed extra #define's related to PI

1999-12-12 03:53  cisc

	* filters.c, loopold.c, newmdct.c, quantize.c, util.h:
	  Added new PI and LOG predefined variants, and stuck them in at
	  respective places.

1999-12-11 23:37  robert

	* lame.c:
	  added low/highpass filter switches

1999-12-11 23:36  robert

	* filters.c, globalflags.h:
	  added highpass filter similar to lowpass filter
	  fixed lowpass filter to avoid endless loops

1999-12-11 20:13  robert

	* mlame:
	  fix for filenames containing spaces

1999-12-11 00:57  takehiro

	* .:
	  added lame binary executable itself

1999-12-10 01:57  takehiro

	* lame.c, loopold.c, quantize.c, util.c, util.h:
	  little bit good subblock_gain calculation.
	  and nint2 is removed.

1999-12-09 15:17  takehiro

	* loopold.c:
	  little bit clean up. some variable restoring is removed.

1999-12-09 07:07  markt

	* quantize-pvt.c, quantize-pvt.h, quantize.c, takehiro.c:
	  PRECALC_SIZE bug?
	  PRECALC_SIZE should be 8191+15+1, since max value of xr/ix is
	  8191+15.

1999-12-09 06:29  markt

	* loopold.c:
	  cod_info-> should be cod_info[ch]->

1999-12-09 02:34  takehiro

	* loopold.c, quantize.c:
	  Oops, stepsize history was not recorded.

1999-12-09 02:28  takehiro

	* loopold.c, quantize-pvt.c, quantize-pvt.h, quantize.c:
	  initial value of binary search is separeted by channels.
	  maybe faster bin_search_stepsize

1999-12-09 00:44  cisc

	* l3bitstream-pvt.h:
	  Removed write_ancillary_data() prototype. (No longer used)

1999-12-09 00:38  cisc

	* takehiro.c:
	  Included the correct quantize header. (quantize-pvt.h instead of
	  quantize.h)

1999-12-08 19:39  cisc

	* util.h:
	  Adjusted PI/SQRT2 precision. (SQRT2 deviated from standard math.h
	  value)

1999-12-08 19:37  cisc

	* mpglib/mpg123.h:
	  Fixed M_PI/M_SQRT2 to work on all platforms. Removed unused
	  INLINE.

1999-12-08 13:45  takehiro

	* mpglib/layer3.c:
	  debug about scfsi treatment

1999-12-08 06:27  markt

	* Makefile, lame.c:
	  fixed display bug for VBR histogram under ncurses. Now termcap
	  is replaced by the more modern ncurses.

1999-12-08 05:46  markt

	* psymodel.c, takehiro.c:
	  avoid bounds check on xr[] for non-hq mode, from Mat.
	  spelling error in psymodel.c fixed :-)

1999-12-08 05:24  markt

	* TODO:
	  *** empty log message ***

1999-12-08 03:49  takehiro

	* psymodel.c:
	  debugged possible buffer overrun.

1999-12-07 22:58  cisc

	* mpglib/layer3.c:
	  No GTK on Amiga. :(

1999-12-07 22:56  cisc

	* main.c:
	  Removed redundant __buffsize (SAS/C (AmigaOS))

1999-12-07 05:11  markt

	* Makefile, TODO, quantize-pvt.c:
	  removed -.5 from quantize_xrpow

1999-12-07 02:04  markt

	* fft.c, fft.h, lame.c, psymodel.c, quantize-pvt.c:
	  backed out takehiro's fft changes for now
	  added latest quantize_xrpow asm from Acy and Mat

1999-12-06 22:45  takehiro

	* fft.c:
	  lookup table integration

1999-12-06 20:38  afaber

	* fft.c, lame.h, machine.h, psymodel.c:
	  Fixed compiler errors

1999-12-03 10:13  takehiro

	* fft.c, fft.h, psymodel.c:
	  maybe faster FFT coding

1999-12-03 10:11  takehiro

	* l3bitstream.c:
	  foolish mistake has removed :)

1999-12-03 09:45  takehiro

	* l3bitstream.c, l3bitstream.h, lame.c:
	  little bit cleanup

1999-12-02 17:01  takehiro

	* loopold.c, quantize-pvt.c, quantize-pvt.h, quantize.c,
	  takehiro.c:
	  correct adj43 overrun(maybe)

1999-11-30 19:30  afaber

	* Dll, mpglib:
	  Added .cvsignore files in sub projects

1999-11-30 19:27  afaber

	* ., gtkanal.c, lame.dsp, mpglib/mpglib.dsp, util.c:
	  Added Win32 specific output files to .cvsignore
	  updated project files to set proper output directory and fixed
	  alignment
	  problem between mpglib project and lame project, now all modules
	  have 8 byte alignment in all configurations

1999-11-29 19:54  takehiro

	* .:
	  configure ignoring *.d file

1999-11-29 19:48  takehiro

	* quantize.c:
	  And more cleanup. needless variable store/restores are removed.

1999-11-29 19:39  takehiro

	* quantize.c:
	  little bit code cleanup(deleted needless call of scale_bitcount)
	  sorry, it is not bugfix of "TOO LARGE quantizerStepSize" ;p

1999-11-29 16:34  markt

	* loopold.c, quantize-pvt.h, quantize.c:
	  removed unused VBRbits from non-VBR quant routines

1999-11-29 03:13  markt

	* loopold.c, quantize-pvt.c, quantize.c:
	  scalefactor preemph *correctly* turned back on

1999-11-29 02:55  markt

	* loopold.c, quantize-pvt.c:
	  scalefac pre-emph turned back on in loopold.c

1999-11-29 02:45  markt

	* encoder.h, gtkanal.h, l3side.h, lame.c, newmdct.c, psymodel.c,
	  psymodel.h, quantize.c, util.c:
	  MS stereo switch slightly improved: old formula was based on the
	  average
	  of ms_ratio of both granules. New formula uses ms_ratio from both
	  granules and the previous and next granule. This will help avoid
	  toggleing
	  MS stereo off for a single frame. Long runs of MS stereo or
	  regular
	  stereo will not be affected.
	  
	  also fixed a bug in frame analyzer - it was accessing l3_xmin in
	  the last
	  scalefactor (l3_xmin and maskings are not computed for last
	  scalefactor)

1999-11-28 23:00  afaber

	* Dll/MP3EncDll.dsp, lame.c, lame.dsp:
	  More fixes in order to compile under Win32

1999-11-28 22:42  afaber

	* lame.dsp, quantize-pvt.c:
	  Fixes to compile under Win32

1999-11-28 22:21  afaber

	* Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.def,
	  Dll/BladeMP3EncDLL.h, Dll/MP3EncDll.dsp, Dll/MP3EncDll.dsw:
	  Update of Blade DLL interface in order to support new lame
	  features

1999-11-28 06:45  markt

	* quantize-pvt.c:
	  updated comments

1999-11-28 06:36  markt

	* USAGE, quantize-pvt.c, quantize-pvt.h, vbrquantize.c:
	  ASM routines (gcc & MSVC) for quantize_xrpow

1999-11-28 01:01  markt

	* Makefile, TODO, lame.c, quantize-pvt.c, quantize-pvt.h,
	  quantize.c:
	  If analog silence is detected, allow VBR code to over ride the
	  VBR_min_bitrate
	  setting and encode at 32kbs.

1999-11-27 23:39  markt

	* Makefile, formatBitstream.c, l3bitstream.c, lame.c,
	  quantize-pvt.c, quantize-pvt.h, quantize.c, util.c,
	  vbrquantize.c:
	  More accurate quantization from Segher Boessenkool

1999-11-25 19:27  markt

	* util.c:
	  fixed allowable bitrate displays

1999-11-24 08:45  markt

	* ., COPYING, Dll, Dll/BladeMP3EncDLL.c, Dll/BladeMP3EncDLL.def,
	  Dll/BladeMP3EncDLL.h, Dll/MP3EncDll.dsp, Dll/MP3EncDll.dsw,
	  HACKING, INSTALL, LICENSE, Makefile, Makefile.MSVC, README,
	  README.WINGTK, TODO, USAGE, VbrTag.c, VbrTag.h, auenc,
	  debugscalefac.c, doc, doc/html, doc/html/contributors.html,
	  doc/html/history.html, doc/html/index.html, doc/html/lame.css,
	  doc/html/node5.html, doc/html/node6.html, doc/html/node7.html,
	  doc/html/switchs.html, doc/man, doc/man/lame.1, encoder.h, fft.c,
	  fft.h, filters.c, filters.h, formatBitstream.c,
	  formatBitstream.h, get_audio.c, get_audio.h, globalflags.h,
	  gpkplotting.c, gpkplotting.h, gtkanal.c, gtkanal.h, id3tag.c,
	  id3tag.h, ieeefloat.c, ieeefloat.h, l3bitstream-pvt.h,
	  l3bitstream.c, l3bitstream.h, l3side.h, lame.c, lame.dsp,
	  lame.dsw, lame.h, loopold.c, machine.h, main.c, mlame, mp3x.c,
	  mpglib, mpglib/Makefile, mpglib/README, mpglib/TODO,
	  mpglib/common.c, mpglib/dct64_i386.c, mpglib/decode_i386.c,
	  mpglib/huffman.h, mpglib/interface.c, mpglib/layer3.c,
	  mpglib/main.c, mpglib/mpg123.h, mpglib/mpglib.dsp,
	  mpglib/mpglib.h, mpglib/tabinit.c, newmdct.c, newmdct.h,
	  portableio.c, portableio.h, psymodel.c, psymodel.h,
	  quantize-pvt.c, quantize-pvt.h, quantize.c, quantize.h,
	  reservoir.c, reservoir.h, tables.c, tables.h, takehiro.c,
	  timestatus.c, timestatus.h, util.c, util.h, vbrquantize.c,
	  version.c, version.h:
	  initial checkin of LAME
	  Starting with LAME 3.57beta with some modifications

