11 years agoARM: add -marm flag since we do not support Thumb
Måns Rullgård [Fri, 2 Oct 2009 08:41:52 +0000 (08:41 +0000)]
ARM: add -marm flag since we do not support Thumb

Originally committed as revision 20144 to svn://

11 years agowav demuxer: return AVERROR_EOF instead of AVERROR(EIO) when the end of
Reimar Döffinger [Fri, 2 Oct 2009 07:27:08 +0000 (07:27 +0000)]
wav demuxer: return AVERROR_EOF instead of AVERROR(EIO) when the end of
the file is reached normally, without any error.

Originally committed as revision 20143 to svn://

11 years agowav: return av_get_packet errors unchanged.
Reimar Döffinger [Fri, 2 Oct 2009 07:22:44 +0000 (07:22 +0000)]
wav: return av_get_packet errors unchanged.

Originally committed as revision 20142 to svn://

11 years agoFix seek regression test: several AVERROR(EIO) correctly changed to AVERROR_EOF.
Reimar Döffinger [Fri, 2 Oct 2009 07:19:21 +0000 (07:19 +0000)]
Fix seek regression test: several AVERROR(EIO) correctly changed to AVERROR_EOF.

Originally committed as revision 20141 to svn://

11 years agoRemove url_feof check that would be triggered only after incorrectly producing
Reimar Döffinger [Fri, 2 Oct 2009 07:05:52 +0000 (07:05 +0000)]
Remove url_feof check that would be triggered only after incorrectly producing
an empty packet and also av_get_packet already handles EOF now.

Originally committed as revision 20140 to svn://

11 years agoRemove useless setting of pkt->size, av_get_packet already handles that.
Reimar Döffinger [Fri, 2 Oct 2009 07:03:04 +0000 (07:03 +0000)]
Remove useless setting of pkt->size, av_get_packet already handles that.

Originally committed as revision 20139 to svn://

11 years agoReturn special EOF checking for aiff and au demuxers, the check was at
Reimar Döffinger [Fri, 2 Oct 2009 06:42:20 +0000 (06:42 +0000)]
Return special EOF checking for aiff and au demuxers, the check was at
the wrong place, causing an empty packet to be read before EOF was
detected and the eof detection is already now handled by av_get_packet anyway.

Originally committed as revision 20138 to svn://

11 years agoReturn any error return values from av_get_packet, get_buffer etc. unchanged
Reimar Döffinger [Fri, 2 Oct 2009 06:40:50 +0000 (06:40 +0000)]
Return any error return values from av_get_packet, get_buffer etc. unchanged
in the raw demuxers.
Also remove special handling of 0-size reads, if they are due to an error/eof,
these are already converted to the appropriate error by get_buffer.

Originally committed as revision 20137 to svn://

11 years agoMake seek regression tests fit after previous get_buffer change.
Reimar Döffinger [Fri, 2 Oct 2009 06:38:56 +0000 (06:38 +0000)]
Make seek regression tests fit after previous get_buffer change.
The au and aiff demuxers would always return an empty packet before
detecting EOF, this is now fixed with av_get_packet etc. returning EOF.

Originally committed as revision 20136 to svn://

11 years agoMake get_buffer and get_partial_buffer return url_ferror or AVERROR_EOF as
Reimar Döffinger [Fri, 2 Oct 2009 06:36:39 +0000 (06:36 +0000)]
Make get_buffer and get_partial_buffer return url_ferror or AVERROR_EOF as
appropriate if it couldn't read any data at all.
This should make handling of EOF and error simpler or make it work right without
extra code in a few place (e.g. raw demuxer).

Originally committed as revision 20135 to svn://

11 years agoCheck whether 32-bit x86 is really 64-bit
David Conrad [Thu, 1 Oct 2009 21:38:25 +0000 (21:38 +0000)]
Check whether 32-bit x86 is really 64-bit
Fixes configure on OS X 10.6

Originally committed as revision 20134 to svn://

11 years agomatroskadec: cosmetics: indentation
Aurelien Jacobs [Thu, 1 Oct 2009 21:15:36 +0000 (21:15 +0000)]
matroskadec: cosmetics: indentation

Originally committed as revision 20133 to svn://

11 years agomatroskadec: don't overwrite extradata already read by ff_get_wav_header()
Aurelien Jacobs [Thu, 1 Oct 2009 21:14:46 +0000 (21:14 +0000)]
matroskadec: don't overwrite extradata already read by ff_get_wav_header()

Originally committed as revision 20132 to svn://

11 years agomatroskadec: fix handling of A_MS/ACM track with no extradata
Aurelien Jacobs [Thu, 1 Oct 2009 21:14:05 +0000 (21:14 +0000)]
matroskadec: fix handling of A_MS/ACM track with no extradata

Originally committed as revision 20131 to svn://

11 years agoMake rawvideo_read_packet return partial frames.
Reimar Döffinger [Thu, 1 Oct 2009 17:32:22 +0000 (17:32 +0000)]
Make rawvideo_read_packet return partial frames.
This is consistent with other demuxers and also fixes a memleak (memory
allocated for partial frame data was leaked).

Originally committed as revision 20130 to svn://

11 years agoSimplify: remove pointless {} and else
Reimar Döffinger [Thu, 1 Oct 2009 17:26:20 +0000 (17:26 +0000)]
Simplify: remove pointless {} and else

Originally committed as revision 20129 to svn://

11 years agoRemove pointless setting of pkt->size, av_get_packet already handles this.
Reimar Döffinger [Thu, 1 Oct 2009 17:22:01 +0000 (17:22 +0000)]
Remove pointless setting of pkt->size, av_get_packet already handles this.

Originally committed as revision 20128 to svn://

11 years agoau demuxer: pass av_get_packet error on unchanged instead of converting to
Reimar Döffinger [Thu, 1 Oct 2009 17:08:33 +0000 (17:08 +0000)]
au demuxer: pass av_get_packet error on unchanged instead of converting to

Originally committed as revision 20127 to svn://

11 years agoav_new_packet failing should return ENOMEM, not EIO.
Reimar Döffinger [Thu, 1 Oct 2009 16:10:09 +0000 (16:10 +0000)]
av_new_packet failing should return ENOMEM, not EIO.

Originally committed as revision 20126 to svn://

11 years agoMake avcodec_open set codec_id and codec_type if they haven't been set.
Reimar Döffinger [Thu, 1 Oct 2009 15:54:55 +0000 (15:54 +0000)]
Make avcodec_open set codec_id and codec_type if they haven't been set.
This fixes the API breakage introduced by the check that avctx codec id and type
match the opened codec and should make (almost?) all applications work again.

Originally committed as revision 20125 to svn://

11 years agoput_bits can only reliably write up to 31 bit bits, above it relies on
Reimar Döffinger [Thu, 1 Oct 2009 15:40:29 +0000 (15:40 +0000)]
put_bits can only reliably write up to 31 bit bits, above it relies on
undefined shift behaviour.
Document this, fix the assert and add a put_bits32 to handle writing 32
bits and use that where necessary.

Originally committed as revision 20124 to svn://

11 years agoUse MANGLE in cavsdsp, the current version using "m" constraints will not
Reimar Döffinger [Thu, 1 Oct 2009 15:30:27 +0000 (15:30 +0000)]
Use MANGLE in cavsdsp, the current version using "m" constraints will not
compile on e.g. OpenBSD due to running out of registers.

Originally committed as revision 20123 to svn://

11 years agoDocument -sn option.
Diego Biurrun [Thu, 1 Oct 2009 13:48:23 +0000 (13:48 +0000)]
Document -sn option.

Originally committed as revision 20122 to svn://

11 years agoFix indentation after last commit.
Benoit Fouet [Thu, 1 Oct 2009 12:21:54 +0000 (12:21 +0000)]
Fix indentation after last commit.

Originally committed as revision 20121 to svn://

11 years agoAdd support for TARGA images without RLE compression.
Anne-Laure de Smit [Thu, 1 Oct 2009 12:20:08 +0000 (12:20 +0000)]
Add support for TARGA images without RLE compression.
Patch by Anne-Laure de Smit: gmail_address(annelaure, desmit)

Originally committed as revision 20120 to svn://

11 years agoExplain the process of patch review and commit a bit.
Diego Biurrun [Thu, 1 Oct 2009 11:58:34 +0000 (11:58 +0000)]
Explain the process of patch review and commit a bit.

Originally committed as revision 20119 to svn://

11 years agoFix playback of strange AVI files with stray LIST.
Aurelien Jacobs [Thu, 1 Oct 2009 10:24:02 +0000 (10:24 +0000)]
Fix playback of strange AVI files with stray LIST.
fix suggested by Michael

Originally committed as revision 20118 to svn://

11 years agoFix wrongly indented block.
Diego Biurrun [Thu, 1 Oct 2009 09:30:08 +0000 (09:30 +0000)]
Fix wrongly indented block.

Originally committed as revision 20117 to svn://

11 years agoUse "!exp" instead of "exp == NULL" in if condition.
Diego Biurrun [Thu, 1 Oct 2009 09:13:21 +0000 (09:13 +0000)]
Use "!exp" instead of "exp == NULL" in if condition.

Originally committed as revision 20116 to svn://

11 years agocosmetics: K&R coding style, prettyprinting
Diego Biurrun [Thu, 1 Oct 2009 09:11:07 +0000 (09:11 +0000)]
cosmetics: K&R coding style, prettyprinting

Originally committed as revision 20115 to svn://

11 years agoDo not attempt to decode TIFF files containing fax data with uncompressed
Kostya Shishkov [Thu, 1 Oct 2009 06:31:49 +0000 (06:31 +0000)]
Do not attempt to decode TIFF files containing fax data with uncompressed
mode allowed for there is no code to decode it (yet).

Originally committed as revision 20114 to svn://

11 years agolavc MJPEG decoder is capable of decoding some extended sequential
Kostya Shishkov [Thu, 1 Oct 2009 05:52:38 +0000 (05:52 +0000)]
lavc MJPEG decoder is capable of decoding some extended sequential
Huffman-compressed JPEG files, so make it decode those files
(samples are welcome).
This fixes issue 1420

Originally committed as revision 20113 to svn://

11 years agoMake BMP decoder produce flipped picture with RLE compression.
Kostya Shishkov [Thu, 1 Oct 2009 05:46:17 +0000 (05:46 +0000)]
Make BMP decoder produce flipped picture with RLE compression.
This fixes issue 1415

Originally committed as revision 20112 to svn://

11 years agoMake MS RLE decoder produce both bottom-up and top-down pictures
Kostya Shishkov [Thu, 1 Oct 2009 05:42:55 +0000 (05:42 +0000)]
Make MS RLE decoder produce both bottom-up and top-down pictures

Originally committed as revision 20111 to svn://

11 years agoChange an error to a warning to support broken AC-3 files known to exist.
Justin Ruggles [Thu, 1 Oct 2009 00:06:54 +0000 (00:06 +0000)]
Change an error to a warning to support broken AC-3 files known to exist.
Fixes Issue 1426.

Originally committed as revision 20110 to svn://

11 years agoSupport constant-quant encoding for libtheora
David Conrad [Wed, 30 Sep 2009 22:25:49 +0000 (22:25 +0000)]
Support constant-quant encoding for libtheora
Based on a patch by j f (gonzaz-z # operamail (d) com)

Originally committed as revision 20109 to svn://

11 years agoconfigure: break a long line
Måns Rullgård [Wed, 30 Sep 2009 22:08:16 +0000 (22:08 +0000)]
configure: break a long line

Originally committed as revision 20108 to svn://

11 years agoRe-add check for compilation failure with ebx clobber, and only set
Reimar Döffinger [Wed, 30 Sep 2009 21:57:13 +0000 (21:57 +0000)]
Re-add check for compilation failure with ebx clobber, and only set
ebx_available if both "b" constraint and exb clobber code compile.
This is to fix compilation with Mac OSX 10.5.8's gcc 4.0.1 in 64
bit mode (-m64) which only fails for ebx/rbx clobbers but not
"b" constraints - this is probably a compiler bug.

Originally committed as revision 20107 to svn://

11 years ago10l: wrong operation in stereo rematrixing
Justin Ruggles [Wed, 30 Sep 2009 21:51:02 +0000 (21:51 +0000)]
10l: wrong operation in stereo rematrixing

Originally committed as revision 20106 to svn://

11 years agoconfigure: return result of test from check_asm
Måns Rullgård [Wed, 30 Sep 2009 21:38:11 +0000 (21:38 +0000)]
configure: return result of test from check_asm

This makes it possible to use check_asm in conditionals.

Originally committed as revision 20105 to svn://

11 years agoAdd support for SGI images without RLE compression.
Anne-Laure de Smit [Wed, 30 Sep 2009 21:23:47 +0000 (21:23 +0000)]
Add support for SGI images without RLE compression.
patch by Anne-Laure de Smit, annelaure.desmit gmail com

Originally committed as revision 20104 to svn://

11 years agoDo not needlessly add 7 to put_buts_count before dividing by 8,
Reimar Döffinger [Wed, 30 Sep 2009 20:47:59 +0000 (20:47 +0000)]
Do not needlessly add 7 to put_buts_count before dividing by 8,
flush_put_bits ensures it is divisible 8.

Originally committed as revision 20103 to svn://

11 years ago10l in code reading codec_name, the first byte is the length and should not
Reimar Döffinger [Wed, 30 Sep 2009 18:27:39 +0000 (18:27 +0000)]
10l in code reading codec_name, the first byte is the length and should not
become part of the string.

Originally committed as revision 20102 to svn://

11 years agoUse a "b" constraint instead of ebx clobber to detect if EBX is available.
Reimar Döffinger [Wed, 30 Sep 2009 16:55:32 +0000 (16:55 +0000)]
Use a "b" constraint instead of ebx clobber to detect if EBX is available.
The test using clobber incorrectly indicates ebx to be available with the
default compilers on OpenBSD 4.5 and Haiku.

Originally committed as revision 20101 to svn://

11 years agoUse floating point mathematics when encoding mpeg audio.
Lasse Reinhold [Wed, 30 Sep 2009 15:13:09 +0000 (15:13 +0000)]
Use floating point mathematics when encoding mpeg audio.
Fixes issue 975: high db peak levels when encoding mp2

Original patch by Lasse Reinhold, lassemikkelreinhold hotmail

Originally committed as revision 20100 to svn://

11 years agoReturn an error when the parsed mpc chunk size is negative, otherwise we
Reimar Döffinger [Wed, 30 Sep 2009 13:35:13 +0000 (13:35 +0000)]
Return an error when the parsed mpc chunk size is negative, otherwise we
might end up in an endless loop where the same chunk is parsed over and over.
Fixes a hang near the end for

Originally committed as revision 20099 to svn://

11 years agoFix compilation of beosaudio.cpp, not tested if it actually works though.
Reimar Döffinger [Wed, 30 Sep 2009 13:01:48 +0000 (13:01 +0000)]
Fix compilation of beosaudio.cpp, not tested if it actually works though.

Originally committed as revision 20098 to svn://

11 years agoFix typo in comment.
Reimar Döffinger [Wed, 30 Sep 2009 12:39:18 +0000 (12:39 +0000)]
Fix typo in comment.

Originally committed as revision 20097 to svn://

11 years agoUse skip_put_bytes in MJPEG encoder instead of filling all bytes with 0
Reimar Döffinger [Wed, 30 Sep 2009 10:48:14 +0000 (10:48 +0000)]
Use skip_put_bytes in MJPEG encoder instead of filling all bytes with 0
with put_bits.

Originally committed as revision 20096 to svn://

11 years agoUse ff_put_string in vorbis encoder.
Reimar Döffinger [Wed, 30 Sep 2009 10:37:37 +0000 (10:37 +0000)]
Use ff_put_string in vorbis encoder.

Originally committed as revision 20095 to svn://

11 years agoReplace several #ifdef PIC with the more obvious and correct
Reimar Döffinger [Wed, 30 Sep 2009 09:49:12 +0000 (09:49 +0000)]
Replace several #ifdef PIC with the more obvious and correct
#if !HAVE_EBX_AVAILABLE, since all it does is avoid using ebx.

Originally committed as revision 20094 to svn://

11 years agoDisable parsing for ogg streams where no ogg header was found,
Reimar Döffinger [Wed, 30 Sep 2009 09:46:48 +0000 (09:46 +0000)]
Disable parsing for ogg streams where no ogg header was found,
if no header was found the parser was not initialized and thus will
crash when trying to use it.

Originally committed as revision 20093 to svn://

11 years agoConvert latin1 codec_name in mov to UTF-8, since all strings in FFmpeg
Reimar Döffinger [Wed, 30 Sep 2009 09:43:56 +0000 (09:43 +0000)]
Convert latin1 codec_name in mov to UTF-8, since all strings in FFmpeg
must be valid UTF-8.

Originally committed as revision 20092 to svn://

11 years agoWhen BitsPerSample tag is not present in TIFF, that means file is
Kostya Shishkov [Wed, 30 Sep 2009 05:49:18 +0000 (05:49 +0000)]
When BitsPerSample tag is not present in TIFF, that means file is
monochrome, so initialize picture before decoding.
This fixes decoding monochrome files produced by lavc TIFF encoder.

Originally committed as revision 20091 to svn://

11 years agoSimplify stereo rematrixing by only using one temporary variable. It is also
Justin Ruggles [Wed, 30 Sep 2009 01:25:04 +0000 (01:25 +0000)]
Simplify stereo rematrixing by only using one temporary variable. It is also
about 1.8% faster on my system.

Originally committed as revision 20090 to svn://

11 years agoremove unneeded assignment in inner loop. rematrixing bands are contiguous.
Justin Ruggles [Wed, 30 Sep 2009 01:09:57 +0000 (01:09 +0000)]
remove unneeded assignment in inner loop. rematrixing bands are contiguous.

Originally committed as revision 20089 to svn://

11 years agoPrefer -vcodec rawvideo over -vcodec huffyuv for lavfi regression
Stefano Sabatini [Tue, 29 Sep 2009 21:53:36 +0000 (21:53 +0000)]
Prefer -vcodec rawvideo over -vcodec huffyuv for lavfi regression
tests, as the use of huffyuv may cause the auto-insertion of a scale
filter, thus masking the mis-behavior of some filter.

Originally committed as revision 20088 to svn://

11 years agoRemove disabled code cruft.
Diego Biurrun [Tue, 29 Sep 2009 21:08:33 +0000 (21:08 +0000)]
Remove disabled code cruft.

Originally committed as revision 20087 to svn://

11 years agoFix md5 program invocation for BSD.
Michael Kostylev [Tue, 29 Sep 2009 19:49:04 +0000 (19:49 +0000)]
Fix md5 program invocation for BSD.
patch by Michael Kostylev, michael.kostylev gmail com

Originally committed as revision 20086 to svn://

11 years agoUse context instead of NULL for logging.
Reimar Döffinger [Tue, 29 Sep 2009 19:18:51 +0000 (19:18 +0000)]
Use context instead of NULL for logging.

Originally committed as revision 20085 to svn://

11 years agoMake sure that dv encoder initializes all encoded packet data.
Reimar Döffinger [Tue, 29 Sep 2009 19:17:18 +0000 (19:17 +0000)]
Make sure that dv encoder initializes all encoded packet data.
The specification does not say which value to use for unused
parts, so fill all unused bytes with 0xff, which is consistent
with what DV usually uses for reserved or unused parts.

Originally committed as revision 20084 to svn://

11 years agocosmetics: K&R coding style, prettyprinting
Diego Biurrun [Tue, 29 Sep 2009 18:03:30 +0000 (18:03 +0000)]
cosmetics: K&R coding style, prettyprinting

Originally committed as revision 20083 to svn://

11 years agoFix indentation in avcodec.h
Jason Garrett-Glaser [Tue, 29 Sep 2009 17:34:36 +0000 (17:34 +0000)]
Fix indentation in avcodec.h

Originally committed as revision 20082 to svn://

11 years agoWMA: fix loop unrolling in decode_exp_vlc()
Måns Rullgård [Tue, 29 Sep 2009 12:48:24 +0000 (12:48 +0000)]
WMA: fix loop unrolling in decode_exp_vlc()

The count can be a non-multiple of 4 after all.

Originally committed as revision 20081 to svn://

11 years agowhitespace cosmetics, prettyprinting, K&R coding style
Diego Biurrun [Tue, 29 Sep 2009 11:43:58 +0000 (11:43 +0000)]
whitespace cosmetics, prettyprinting, K&R coding style

Originally committed as revision 20080 to svn://

11 years agoSimplify r20025: use align_get_bits instead of reimplementing it.
Reimar Döffinger [Tue, 29 Sep 2009 11:06:04 +0000 (11:06 +0000)]
Simplify r20025: use align_get_bits instead of reimplementing it.

Originally committed as revision 20079 to svn://

11 years agoWMA: store level_table as floats, use type punning for sign flip in decode
Måns Rullgård [Tue, 29 Sep 2009 10:38:34 +0000 (10:38 +0000)]
WMA: store level_table as floats, use type punning for sign flip in decode

Originally committed as revision 20078 to svn://

11 years agoWMA: use type punning and unroll loops in decode_exp_vlc()
Måns Rullgård [Tue, 29 Sep 2009 10:38:30 +0000 (10:38 +0000)]
WMA: use type punning and unroll loops in decode_exp_vlc()

GCC does stupid things if these assignments are done using floats
directly, so fill the runs using integer operations instead.  Also
unroll the loops since the length is always a multiple of 4.

Originally committed as revision 20077 to svn://

11 years agoWMA: use a table instead of pow() in decode_exp_vlc
Måns Rullgård [Tue, 29 Sep 2009 10:38:27 +0000 (10:38 +0000)]
WMA: use a table instead of pow() in decode_exp_vlc

Originally committed as revision 20076 to svn://

11 years agoSync AC3 probe values with MP3 probe values, they have to avoid similar issues.
Reimar Döffinger [Tue, 29 Sep 2009 10:23:47 +0000 (10:23 +0000)]
Sync AC3 probe values with MP3 probe values, they have to avoid similar issues.
This fixes ffmpeg-generated files with -acodec ac3 being detected as raw ac3
instead of MPEG.

Originally committed as revision 20075 to svn://

11 years agoExtend DV autodetection to also reliably detect single-frame DVs with
Reimar Döffinger [Tue, 29 Sep 2009 10:12:18 +0000 (10:12 +0000)]
Extend DV autodetection to also reliably detect single-frame DVs with
a higher score that MAX/4.
It checks that there are at least 10 DIF headers and at least one per
24000 bytes, and if so considers the file reliably detected as DV.
Passes probetest, too.

Originally committed as revision 20074 to svn://

11 years agoAdd id3v2 metadata conversion table and use it in mp3 muxer.
Anton Khirnov [Tue, 29 Sep 2009 07:05:03 +0000 (07:05 +0000)]
Add id3v2 metadata conversion table and use it in mp3 muxer.

Patch by Anton Khirnov, wyskas gmail

Originally committed as revision 20073 to svn://

11 years agoMake sure all the bits are written to output in fax data decoder.
Kostya Shishkov [Tue, 29 Sep 2009 05:55:14 +0000 (05:55 +0000)]
Make sure all the bits are written to output in fax data decoder.
This fixes decoding TIFF images with fax compression and width being not
multiple of eight (and issue 1429).

Originally committed as revision 20072 to svn://

11 years agosome more linebreak and brace placement cosmetics
Diego Biurrun [Mon, 28 Sep 2009 21:46:22 +0000 (21:46 +0000)]
some more linebreak and brace placement cosmetics

Originally committed as revision 20071 to svn://

11 years agobrace placement and linebreak cosmetics
Diego Biurrun [Mon, 28 Sep 2009 21:13:47 +0000 (21:13 +0000)]
brace placement and linebreak cosmetics

Originally committed as revision 20070 to svn://

11 years agoRefactor channel element configuration and mapping code into its own function
Robert Swain [Mon, 28 Sep 2009 15:37:18 +0000 (15:37 +0000)]
Refactor channel element configuration and mapping code into its own function
to allow reuse

Originally committed as revision 20069 to svn://

11 years agoReindent after last commit
Robert Swain [Mon, 28 Sep 2009 15:35:51 +0000 (15:35 +0000)]
Reindent after last commit

Originally committed as revision 20068 to svn://

11 years agoAAC: Add channel layout support for channel configuration as defined in the
Robert Swain [Mon, 28 Sep 2009 15:35:13 +0000 (15:35 +0000)]
AAC: Add channel layout support for channel configuration as defined in the

Originally committed as revision 20067 to svn://

11 years agocosmetics: reindent and reformat function declarations
Kostya Shishkov [Mon, 28 Sep 2009 05:19:13 +0000 (05:19 +0000)]
cosmetics: reindent and reformat function declarations

Originally committed as revision 20066 to svn://

11 years agoFinally distinguish TIFF_CCITT_RLE and TIFF_G3 1-D case, so both of them
Kostya Shishkov [Mon, 28 Sep 2009 05:17:20 +0000 (05:17 +0000)]
Finally distinguish TIFF_CCITT_RLE and TIFF_G3 1-D case, so both of them
will be decoded correctly.
This fixes issue 1423.

Originally committed as revision 20065 to svn://

11 years agoMake TIFF decoder load compression options only for corresponding codec
Kostya Shishkov [Mon, 28 Sep 2009 05:15:27 +0000 (05:15 +0000)]
Make TIFF decoder load compression options only for corresponding codec

Originally committed as revision 20064 to svn://

11 years agoARM: NEON optimised vector_fmul_add
Måns Rullgård [Sun, 27 Sep 2009 16:52:05 +0000 (16:52 +0000)]
ARM: NEON optimised vector_fmul_add

Originally committed as revision 20063 to svn://

11 years agocosmetics: fix indentation after previous commit
Måns Rullgård [Sun, 27 Sep 2009 16:52:00 +0000 (16:52 +0000)]
cosmetics: fix indentation after previous commit

Originally committed as revision 20062 to svn://

11 years agoDrop unused args from vector_fmul_add_add, simpify code, and rename
Måns Rullgård [Sun, 27 Sep 2009 16:51:54 +0000 (16:51 +0000)]
Drop unused args from vector_fmul_add_add, simpify code, and rename

The src3 and step arguments to vector_fmul_add_add() are always zero
and one, respectively.  This removes these arguments from the function,
simplifies the code accordingly, and renames the function to better
match the new operation.

Originally committed as revision 20061 to svn://

11 years agoRemove duplicated / stray code which slipped in with r20057.
Stefano Sabatini [Sun, 27 Sep 2009 10:29:24 +0000 (10:29 +0000)]
Remove duplicated / stray code which slipped in with r20057.

Originally committed as revision 20060 to svn://

11 years agoGet rid of "function" "keywords" in lavfilter testing code, hopefully will
Reimar Döffinger [Sun, 27 Sep 2009 10:06:35 +0000 (10:06 +0000)]
Get rid of "function" "keywords" in lavfilter testing code, hopefully will
make regressions tests runnable again with all shells.

Originally committed as revision 20059 to svn://

11 years agoDo not compile ZLib data uncompressing function in TIFF decoder when ZLib is
Martin Storsjö [Sun, 27 Sep 2009 09:43:11 +0000 (09:43 +0000)]
Do not compile ZLib data uncompressing function in TIFF decoder when ZLib is
not present.
Patch by Martin Storsjö
($firstname <at> $firstname <dot> two first letters of $lastname)

Originally committed as revision 20058 to svn://

11 years agoImplement a lavfitest target for testing regressions in libavfilter.
Stefano Sabatini [Sun, 27 Sep 2009 09:33:37 +0000 (09:33 +0000)]
Implement a lavfitest target for testing regressions in libavfilter.
The target is disabled.

See the thread: "[PATCH] libavfilter-soc: regression test for libavfilter".

Originally committed as revision 20057 to svn://

11 years agoWMAPRO: use vector_fmul_scalar from dsputil where possible
Måns Rullgård [Sun, 27 Sep 2009 08:16:50 +0000 (08:16 +0000)]
WMAPRO: use vector_fmul_scalar from dsputil where possible

Originally committed as revision 20056 to svn://

11 years agoRearrange loop structure for approx. 35-50% faster calc_transform_coeffs_cpl()
Justin Ruggles [Sun, 27 Sep 2009 07:16:51 +0000 (07:16 +0000)]
Rearrange loop structure for approx. 35-50% faster calc_transform_coeffs_cpl()
depending on content.

Originally committed as revision 20055 to svn://

11 years agoLooks like ZLib uncompress() cannot deal with some kinds of TIFF deflated data,
Kostya Shishkov [Sun, 27 Sep 2009 07:01:01 +0000 (07:01 +0000)]
Looks like ZLib uncompress() cannot deal with some kinds of TIFF deflated data,
so replace it with custom code.
This fixes issue 1419.

Originally committed as revision 20054 to svn://

11 years agoMove variable declaration to inside of loop.
Justin Ruggles [Sun, 27 Sep 2009 06:38:57 +0000 (06:38 +0000)]
Move variable declaration to inside of loop.

Originally committed as revision 20053 to svn://

11 years agoCosmetics: Rename some variables.
Justin Ruggles [Sun, 27 Sep 2009 06:38:13 +0000 (06:38 +0000)]
Cosmetics: Rename some variables.

Originally committed as revision 20052 to svn://

11 years agoSimplify coupling band loop.
Justin Ruggles [Sun, 27 Sep 2009 06:33:23 +0000 (06:33 +0000)]
Simplify coupling band loop.

Originally committed as revision 20051 to svn://

11 years agoRemove unneeded table lookup.
Justin Ruggles [Sun, 27 Sep 2009 06:16:49 +0000 (06:16 +0000)]
Remove unneeded table lookup.

Originally committed as revision 20050 to svn://

11 years agoCosmetics: Rename some variables to be more descriptive of their use.
Justin Ruggles [Sun, 27 Sep 2009 06:13:01 +0000 (06:13 +0000)]
Cosmetics: Rename some variables to be more descriptive of their use.

Originally committed as revision 20049 to svn://

11 years agoMove some variable declarations to inside of loops.
Justin Ruggles [Sun, 27 Sep 2009 06:10:36 +0000 (06:10 +0000)]
Move some variable declarations to inside of loops.

Originally committed as revision 20048 to svn://

11 years agoRemove unneeded variable.
Justin Ruggles [Sun, 27 Sep 2009 06:01:32 +0000 (06:01 +0000)]
Remove unneeded variable.

Originally committed as revision 20047 to svn://

11 years agoCosmetics: Rename some variables to be more descriptive of their use.
Justin Ruggles [Sun, 27 Sep 2009 05:48:03 +0000 (05:48 +0000)]
Cosmetics: Rename some variables to be more descriptive of their use.

Originally committed as revision 20046 to svn://

11 years agoRemove some unneeded variables.
Justin Ruggles [Sun, 27 Sep 2009 05:39:57 +0000 (05:39 +0000)]
Remove some unneeded variables.

Originally committed as revision 20045 to svn://