]> git.sesse.net Git - ffmpeg/log
ffmpeg
11 years agofate: ea, h264: prettyprinting and ordering cosmetics
Diego Biurrun [Wed, 28 Nov 2012 16:54:07 +0000 (17:54 +0100)]
fate: ea, h264: prettyprinting and ordering cosmetics

11 years agoconfigure: Strip ordinals from mingw generated def files
Martin Storsjö [Sat, 17 Nov 2012 14:34:54 +0000 (16:34 +0200)]
configure: Strip ordinals from mingw generated def files

The def files are used for generating import libraries for
other toolchains (in particular, for generating import libraries
for MSVC for DLLs built with mingw).

The def files produced by mingw/gcc contains ordinals for each
exported function. When MSVC tools generate import libraries
from such a def file, MSVC links to the DLL by the ordinals
instead of linking by name.

Since the def files aren't maintained by hand, the ordinal
numbers are assigned (more or less) randomly and any caller
linking to the libs by ordinals will break as soon as the libraries
export more/fewer functions.

Therefore, strip out the ordinals from the generated def files,
to make users link to the libraries by name.

Callers linking to the DLLs using the gcc provided import library
link by name as they should.

Signed-off-by: Martin Storsjö <martin@martin.st>
11 years agoconfigure: arm: detect default thumb state of compiler
Mans Rullgard [Fri, 30 Nov 2012 21:58:42 +0000 (21:58 +0000)]
configure: arm: detect default thumb state of compiler

This makes configure honour the compiler default if thumb is not
explicitly enabled or disabled.

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agofate: fft: Fix libavcodec dependency
Diego Biurrun [Wed, 28 Nov 2012 23:22:25 +0000 (00:22 +0100)]
fate: fft: Fix libavcodec dependency

11 years agobuild: Make the ISMV muxer select the MOV muxer
Diego Biurrun [Sat, 1 Dec 2012 16:56:00 +0000 (17:56 +0100)]
build: Make the ISMV muxer select the MOV muxer

They share the same code, so building one w/o the other makes no sense.

11 years agoconfigure: move arm arch extensions to a separate variable
Mans Rullgard [Sat, 1 Dec 2012 12:26:12 +0000 (12:26 +0000)]
configure: move arm arch extensions to a separate variable

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agofate-seek: remove use of gnu make 3.82 only private modifier
Janne Grunau [Sun, 2 Dec 2012 23:37:07 +0000 (00:37 +0100)]
fate-seek: remove use of gnu make 3.82 only private modifier

11 years agofate: move vsynth reference files to their own directory
Janne Grunau [Sun, 2 Dec 2012 23:36:10 +0000 (00:36 +0100)]
fate: move vsynth reference files to their own directory

11 years agofate: move fate-acodec reference files to their own dir
Janne Grunau [Sun, 2 Dec 2012 23:29:35 +0000 (00:29 +0100)]
fate: move fate-acodec reference files to their own dir

11 years agoconfigure: avplay now depends on avresample
Diego Biurrun [Wed, 28 Nov 2012 23:14:23 +0000 (00:14 +0100)]
configure: avplay now depends on avresample

11 years agofate: split dependencies for fate-seek tests
Janne Grunau [Sun, 2 Dec 2012 19:31:20 +0000 (20:31 +0100)]
fate: split dependencies for fate-seek tests

Each fate-seek test depends now only on the corresponding fate-acodec,
fate-vsynth2 or fate-lavf test which creates the file seek-tests
operates on. The tests and references are renamed to match the test they
depend on.

11 years agoh264: error out on unset current_picture_ptr for h->current_slice > 0
Janne Grunau [Wed, 21 Nov 2012 18:41:59 +0000 (19:41 +0100)]
h264: error out on unset current_picture_ptr for h->current_slice > 0

Fixes a segfault with fuzzed sample sample_varPAR_s11622_r001-02.avi.

CC: libav-stable@libav.org
11 years agoavprobe: report per stream bit rate if set by the decoder
Janne Grunau [Fri, 30 Nov 2012 14:44:19 +0000 (15:44 +0100)]
avprobe: report per stream bit rate if set by the decoder

11 years agoaac: avoid a memcpy in sbr_qmf_analysis
Christophe Gisquet [Wed, 28 Nov 2012 21:47:25 +0000 (22:47 +0100)]
aac: avoid a memcpy in sbr_qmf_analysis

Swapping buffer indices allows saving one memcpy that accounts for 1% of the
runtime, according to oprofile.

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
11 years agoconfigure: fix indentation in option parsing loop
Mans Rullgard [Thu, 29 Nov 2012 18:03:50 +0000 (18:03 +0000)]
configure: fix indentation in option parsing loop

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agoconfigure: Use headers in the check for _beginthreadex for w32threads
Martin Storsjö [Fri, 30 Nov 2012 14:57:02 +0000 (16:57 +0200)]
configure: Use headers in the check for _beginthreadex for w32threads

When targeting the metro API subset, this function still exists in
the link libraries, but is excluded from the headers. This makes
sure w32threads is automatically disabled when targeting this API
subset (since not all the necessary functions for it are available).

Signed-off-by: Martin Storsjö <martin@martin.st>
11 years agoavutil: Use a configure check to enable windows console functions
Martin Storsjö [Fri, 30 Nov 2012 14:56:57 +0000 (16:56 +0200)]
avutil: Use a configure check to enable windows console functions

Not all versions or API subsets of windows have these functions.

Signed-off-by: Martin Storsjö <martin@martin.st>
11 years agoavutil: Include io.h with a separate condition from windows console functions
Martin Storsjö [Fri, 30 Nov 2012 14:56:54 +0000 (16:56 +0200)]
avutil: Include io.h with a separate condition from windows console functions

Not all versions of windows have the console color functions,
while io.h might be needed for isatty (which can be found in
unistd.h or io.h).

Signed-off-by: Martin Storsjö <martin@martin.st>
11 years agoavutil: Include io.h with a separate condition from MapViewOfFile
Martin Storsjö [Thu, 29 Nov 2012 15:02:42 +0000 (17:02 +0200)]
avutil: Include io.h with a separate condition from MapViewOfFile

The existence of MapViewOfFile isn't linked to the existence of
io.h.

Not all versions of windows have MapViewOfFile (in particular,
Windows Phone 8 and the "metro" windows 8 API subset don't),
while they still have io.h (and need it for open/read/close).

Signed-off-by: Martin Storsjö <martin@martin.st>
11 years agocmdutils: Use a configure check for enabling CommandLineToArgvW
Martin Storsjö [Thu, 29 Nov 2012 15:02:24 +0000 (17:02 +0200)]
cmdutils: Use a configure check for enabling CommandLineToArgvW

This simplifies the condition to avoid hardcoding the systems
where the function exists. This also simplifies support for
newer Windows API subsets where this function doesn't exist,
such as Windows Phone 8 and the "metro" API subset of Windows 8.

Signed-off-by: Martin Storsjö <martin@martin.st>
11 years agobuild: The FLAC encoder also depends on the flacdsp code
Diego Biurrun [Thu, 29 Nov 2012 09:34:05 +0000 (10:34 +0100)]
build: The FLAC encoder also depends on the flacdsp code

Fixes linking with only the FLAC encoder enabled.

11 years agoimg2: K&R formatting cosmetics
Diego Biurrun [Thu, 29 Nov 2012 03:56:40 +0000 (04:56 +0100)]
img2: K&R formatting cosmetics

Also introduce local img_ namespace to simplify debugging.

11 years agoh264: check context state before decoding slice data partitions
Janne Grunau [Wed, 28 Nov 2012 21:17:14 +0000 (22:17 +0100)]
h264: check context state before decoding slice data partitions

Fixes mov_h264_aac__Demo_FlagOfOurFathers.mov.SIGSEGV.4e9.656.

Found-by: Mateusz "j00ru" Jurczyk
CC: libav-stable@libav.org
11 years agoflashsv: make sure data for zlib priming is available
Janne Grunau [Wed, 28 Nov 2012 16:31:35 +0000 (17:31 +0100)]
flashsv: make sure data for zlib priming is available

Fixes a segfault in the fuzzed sample resolutionchange.flv_s314809.

CC: libav-stable@libav.org
11 years agoavplay: Do not use removed av_get_int()
Anton Khirnov [Wed, 28 Nov 2012 20:23:57 +0000 (21:23 +0100)]
avplay: Do not use removed av_get_int()

Fixes build with lavfi disabled.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
11 years agoavconv: fix variable shadowing in configure_input_audio_filter()
Justin Ruggles [Tue, 2 Oct 2012 18:12:27 +0000 (14:12 -0400)]
avconv: fix variable shadowing in configure_input_audio_filter()

11 years agofate: h264: Add dependencies
Diego Biurrun [Wed, 28 Nov 2012 16:53:47 +0000 (17:53 +0100)]
fate: h264: Add dependencies

11 years agofate: ea: Add dependencies
Diego Biurrun [Wed, 28 Nov 2012 16:16:34 +0000 (17:16 +0100)]
fate: ea: Add dependencies

11 years agofate: Do not unconditionally run libavutil tests
Diego Biurrun [Wed, 28 Nov 2012 15:59:13 +0000 (16:59 +0100)]
fate: Do not unconditionally run libavutil tests

Now that libavutil can be disabled, this should be conditional.

11 years agortpenc_chain: Remove unused variable
Diego Biurrun [Wed, 28 Nov 2012 15:23:23 +0000 (16:23 +0100)]
rtpenc_chain: Remove unused variable

11 years agonuv: check for malloc failure when allocating extradata
Justin Ruggles [Tue, 27 Nov 2012 22:47:19 +0000 (17:47 -0500)]
nuv: check for malloc failure when allocating extradata

Also make sure extradata is freed in the case where multiple
NUV_EXTRADATA frame types are found. This may not happen in practice,
but it could happen in a malformed stream, which would lead to a memleak
if not handled.

11 years agonuv: use the stream indices generated by avformat_new_stream()
Justin Ruggles [Tue, 27 Nov 2012 21:41:11 +0000 (16:41 -0500)]
nuv: use the stream indices generated by avformat_new_stream()

11 years agonuv: cosmetics: pretty-printing
Justin Ruggles [Tue, 27 Nov 2012 20:19:00 +0000 (15:19 -0500)]
nuv: cosmetics: pretty-printing

11 years agolavf: move nuv fourcc audio tags from riff to nuv
Justin Ruggles [Tue, 27 Nov 2012 04:36:13 +0000 (23:36 -0500)]
lavf: move nuv fourcc audio tags from riff to nuv

11 years agolavf: add a common function for selecting a pcm codec from parameters
Justin Ruggles [Tue, 27 Nov 2012 19:52:38 +0000 (14:52 -0500)]
lavf: add a common function for selecting a pcm codec from parameters

11 years agolavf: move ff_codec_get_tag() and ff_codec_get_id() definitions to internal.h
Justin Ruggles [Tue, 27 Nov 2012 16:52:28 +0000 (11:52 -0500)]
lavf: move ff_codec_get_tag() and ff_codec_get_id() definitions to internal.h

11 years agolavf: move "MP3 " fourcc from riff to nut
Justin Ruggles [Tue, 27 Nov 2012 04:23:01 +0000 (23:23 -0500)]
lavf: move "MP3 " fourcc from riff to nut

Original commit, 7b24f9b, says it was added because it is used in libnut.

11 years agofate: vpx: Add dependencies
Diego Biurrun [Wed, 28 Nov 2012 10:27:59 +0000 (11:27 +0100)]
fate: vpx: Add dependencies

11 years agofate: Fix wavpack-matroskamode test dependencies
Diego Biurrun [Wed, 28 Nov 2012 10:14:22 +0000 (11:14 +0100)]
fate: Fix wavpack-matroskamode test dependencies

11 years agox86: dsputilenc: port to cpuflags
Diego Biurrun [Mon, 16 Jul 2012 00:05:16 +0000 (02:05 +0200)]
x86: dsputilenc: port to cpuflags

11 years agofate: Add dependencies for Vorbis, ProRes, QTRLE, utvideo tests
Diego Biurrun [Wed, 28 Nov 2012 10:45:17 +0000 (11:45 +0100)]
fate: Add dependencies for Vorbis, ProRes, QTRLE, utvideo tests

11 years agofate: real: Add dependencies
Diego Biurrun [Wed, 28 Nov 2012 10:44:56 +0000 (11:44 +0100)]
fate: real: Add dependencies

11 years agofate: lossless-audio: Add dependencies
Diego Biurrun [Wed, 28 Nov 2012 10:44:24 +0000 (11:44 +0100)]
fate: lossless-audio: Add dependencies

11 years agox86: h264dsp: Fix linking with yasm and optimizations disabled
Diego Biurrun [Tue, 27 Nov 2012 19:35:27 +0000 (20:35 +0100)]
x86: h264dsp: Fix linking with yasm and optimizations disabled

Some optimized functions reference optimized symbols, so the functions
must be explicitly disabled when those symbols are unavailable.

11 years agovf_fieldorder: require write permissions
Anton Khirnov [Tue, 27 Nov 2012 07:59:55 +0000 (08:59 +0100)]
vf_fieldorder: require write permissions

This filter modifies the input frame.

11 years agovf_fieldorder: reindent
Anton Khirnov [Tue, 27 Nov 2012 07:59:22 +0000 (08:59 +0100)]
vf_fieldorder: reindent

11 years agolavfi: merge start_frame/draw_slice/end_frame
Anton Khirnov [Tue, 27 Nov 2012 06:49:45 +0000 (07:49 +0100)]
lavfi: merge start_frame/draw_slice/end_frame

Any alleged performance benefits gained from the split are purely
mythological and do not justify added code complexity.

11 years agolavfi: remove vf_slicify
Anton Khirnov [Tue, 27 Nov 2012 09:41:22 +0000 (10:41 +0100)]
lavfi: remove vf_slicify

The following commit will make it useless.

The crop_scale_vflip FATE test changes because of off-by-one differences
in output when vflipped slices are passed to sws.

11 years agoasink_nullsink: plug a memory leak.
Anton Khirnov [Tue, 27 Nov 2012 10:35:55 +0000 (11:35 +0100)]
asink_nullsink: plug a memory leak.

11 years agox86: h264_idct: port to cpuflags
Diego Biurrun [Sun, 15 Jul 2012 13:43:47 +0000 (15:43 +0200)]
x86: h264_idct: port to cpuflags

11 years agox86: cpu: Drop unused HAVE_RWEFLAGS condition
Diego Biurrun [Tue, 27 Nov 2012 18:47:34 +0000 (19:47 +0100)]
x86: cpu: Drop unused HAVE_RWEFLAGS condition

The test for rweflags was dropped in a previous commit.

11 years agovble: Do not abort decoding when version is not 1
Piotr Bandurski [Tue, 27 Nov 2012 14:05:46 +0000 (14:05 +0000)]
vble: Do not abort decoding when version is not 1

Some combinations of OS, VirtualDub, and VBLE can accidentally
set the version to a value other than 1. Since no other version
of VBLE was ever released, simply warn about it.

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
11 years agolavr: do not pass consumed samples as a parameter to ff_audio_resample()
Justin Ruggles [Thu, 1 Nov 2012 15:08:08 +0000 (11:08 -0400)]
lavr: do not pass consumed samples as a parameter to ff_audio_resample()

Since the resampler handles buffering of unconsumed samples internally, the
caller does not need this information.

11 years agolavr: correct the documentation for the ff_audio_resample() return value
Justin Ruggles [Thu, 1 Nov 2012 05:10:37 +0000 (01:10 -0400)]
lavr: correct the documentation for the ff_audio_resample() return value

11 years agolavr: do not pass sample count as a parameter to ff_audio_convert()
Justin Ruggles [Thu, 1 Nov 2012 04:44:11 +0000 (00:44 -0400)]
lavr: do not pass sample count as a parameter to ff_audio_convert()

It will always be the number of samples in the input buffer, so just use that
directly instead of passing it as a separate parameter.

11 years agox86: h264_weight: port to cpuflags
Diego Biurrun [Sun, 29 Jul 2012 13:54:55 +0000 (15:54 +0200)]
x86: h264_weight: port to cpuflags

11 years agoconfigure: Enable avconv filter dependencies automatically
Martin Storsjö [Fri, 9 Nov 2012 12:32:38 +0000 (14:32 +0200)]
configure: Enable avconv filter dependencies automatically

This makes sure minimal configurations such as
"--disable-everything --enable-avconv" will enable the filters
necessary for running avconv, instead of just keeping avconv
disabled (even if the user specified "--enable-avconv").

Signed-off-by: Martin Storsjö <martin@martin.st>
11 years agoh264: set Picture.owner2 to the current thread
Janne Grunau [Sun, 25 Nov 2012 18:42:16 +0000 (19:42 +0100)]
h264: set Picture.owner2 to the current thread

This does not seem to have an effect currently. Fate-h264 passes with
THREADS=1..16 and both threading types as before. It fixes however a
segfault during error resilience with my adaptive-frame-mt patchset.
A picture in use during error resilience gets realloced in another
thread in the fuzzed sample sample_varPAR.avi_s226019.

11 years agoh264: check ref_count validity for num_ref_idx_active_override_flag
Janne Grunau [Sun, 25 Nov 2012 15:02:02 +0000 (16:02 +0100)]
h264: check ref_count validity for num_ref_idx_active_override_flag

Fixes segfault in the fuzzed sample bipbop234.ts_s226407.

CC: libav-stable@libav.org
11 years agoh264: add missing new line to log message
Janne Grunau [Wed, 21 Nov 2012 19:13:15 +0000 (20:13 +0100)]
h264: add missing new line to log message

11 years agodcadec: skip QMF on unused channels
Michael Niedermayer [Mon, 29 Oct 2012 04:21:47 +0000 (05:21 +0100)]
dcadec: skip QMF on unused channels

When the extra rear channel is present but unused, the
s->channel_order_tab[] value for that channel is -1. The QMF can be
skipped for the extra channel, and doing so avoids an out-of-array read
on s->samples_chanptr[].

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Justin Ruggles <justin.ruggles@gmail.com>
11 years agowavenc: write fact chunk sample count at the correct file position
Michael Niedermayer [Tue, 30 Oct 2012 20:48:39 +0000 (21:48 +0100)]
wavenc: write fact chunk sample count at the correct file position

Fixes curruption of metadata in the INFO chunk.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Justin Ruggles <justin.ruggles@gmail.com>
11 years agoriff: do not add empty metadata tags in INFO chunk
Justin Ruggles [Sun, 25 Nov 2012 17:40:54 +0000 (12:40 -0500)]
riff: do not add empty metadata tags in INFO chunk

11 years agoriff: only warn on a bad INFO chunk code size instead of failing
Justin Ruggles [Sun, 25 Nov 2012 17:38:58 +0000 (12:38 -0500)]
riff: only warn on a bad INFO chunk code size instead of failing

fixes Bug 392

11 years agoconfigure: Add separate list for libraries and use where appropriate
Diego Biurrun [Sun, 11 Nov 2012 12:37:40 +0000 (13:37 +0100)]
configure: Add separate list for libraries and use where appropriate

11 years agox86: float_dsp: add SSE version of vector_fmul_scalar()
Justin Ruggles [Sat, 22 Sep 2012 22:41:25 +0000 (18:41 -0400)]
x86: float_dsp: add SSE version of vector_fmul_scalar()

11 years agodsputil: move vector_fmul_scalar() to AVFloatDSPContext in libavutil
Justin Ruggles [Sat, 22 Sep 2012 22:13:57 +0000 (18:13 -0400)]
dsputil: move vector_fmul_scalar() to AVFloatDSPContext in libavutil

11 years agoaacenc: use the correct output buffer
Michael Niedermayer [Mon, 26 Nov 2012 15:30:51 +0000 (10:30 -0500)]
aacenc: use the correct output buffer

This fixes segfault caused by 3d3cf6745e2a5dc9c377244454c3186d75b177fa
when SingleChannelElement.ret was renamed to SingleChannelElement.ret_buf.

Signed-off-by: Justin Ruggles <justin.ruggles@gmail.com>
11 years agoaacdec: fix signed overflows in lcg_random()
Mans Rullgard [Sat, 24 Nov 2012 13:12:47 +0000 (13:12 +0000)]
aacdec: fix signed overflows in lcg_random()

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agobase64: fix signed overflow in shift
Mans Rullgard [Sat, 24 Nov 2012 13:03:53 +0000 (13:03 +0000)]
base64: fix signed overflow in shift

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agolavf: avoid integer overflow in ff_compute_frame_duration()
Janne Grunau [Fri, 23 Nov 2012 13:05:36 +0000 (14:05 +0100)]
lavf: avoid integer overflow in ff_compute_frame_duration()

Scaling the denominator instead of the numerator if it is too large
loses precision. Fixes an assert caused by a negative frame duration in
the fuzzed sample nasa-8s2.ts_s202310.

CC: libav-stable@libav.org
11 years agoflashsv: check for keyframe before using differential coding
Janne Grunau [Sat, 24 Nov 2012 14:50:03 +0000 (15:50 +0100)]
flashsv: check for keyframe before using differential coding

Fixes a segfault in te fuzzed sample resolutionchange.flv_s211713.

CC: libav-stable@libav.org
11 years agoh264: enable low delay only if no delayed frames were seen
Janne Grunau [Fri, 16 Nov 2012 13:31:09 +0000 (14:31 +0100)]
h264: enable low delay only if no delayed frames were seen

Dropping frames is undesirable but that is the only way by which the
decoder could return to low delay mode. Instead emit a warning and
continue with delayed frames.
Fixes a crash in fuzzed sample nasa-8s2.ts_s20033 caused by a larger
than expected has_b_frames value. Low delay keeps getting re-enabled
from a presumely broken SPS.

CC: libav-stable@libav.org
11 years agox86: fix build without inline asm
Diego Biurrun [Mon, 26 Nov 2012 00:44:29 +0000 (01:44 +0100)]
x86: fix build without inline asm

The qpel functions referenced here are not related to h264 and should
thus never have been under CONFIG_H264QPEL.

Signed-off-by: Mans Rullgard <mans@mansr.com>
Signed-off-by: Diego Biurrun <diego@biurrun.de>
11 years agoaacdec: use float planar sample format for output
Justin Ruggles [Sun, 14 Oct 2012 04:12:55 +0000 (00:12 -0400)]
aacdec: use float planar sample format for output

11 years agolavc: clarify get_buffer() documentation
Justin Ruggles [Sun, 25 Nov 2012 16:37:14 +0000 (11:37 -0500)]
lavc: clarify get_buffer() documentation

This is needed for the AAC decoder, which may need to call get_buffer()
more than once if the channel configuration changes.

11 years agompegaudiodec: use planar sample format for output unless packed is requested
Justin Ruggles [Mon, 27 Aug 2012 22:17:33 +0000 (18:17 -0400)]
mpegaudiodec: use planar sample format for output unless packed is requested

11 years agox86: h264 qpel: use the correct number of utilized xmm regs in cglobal
Justin Ruggles [Sun, 25 Nov 2012 21:11:57 +0000 (16:11 -0500)]
x86: h264 qpel: use the correct number of utilized xmm regs in cglobal

Fixes xmm register clobbering on win64.

11 years agoremove #defines to prevent use of discouraged external functions
Janne Grunau [Fri, 26 Oct 2012 18:26:29 +0000 (20:26 +0200)]
remove #defines to prevent use of discouraged external functions

Preventing the use of discouraged or 'insecure' external functions
through defines in an internal header is not a good solution. The
header is not guaranteed to be included universally which makes
overlooking bad use of said functions during review more likely.

There are cases were those functions either are the most straight
forward solution or even have to be used. Using malloc or free is
required if the allocation or release is done by other libraries.

11 years agox86: h264: Convert 8-bit QPEL inline assembly to YASM
Daniel Kang [Sat, 13 Oct 2012 15:04:50 +0000 (10:04 -0500)]
x86: h264: Convert 8-bit QPEL inline assembly to YASM

Signed-off-by: Diego Biurrun <diego@biurrun.de>
11 years agox86: h264: Remove 3dnow QPEL code
Daniel Kang [Sat, 13 Oct 2012 15:04:49 +0000 (10:04 -0500)]
x86: h264: Remove 3dnow QPEL code

The only CPUs that have 3dnow and don't have mmxext are 12 years old.
Moreover, AMD has dropped 3dnow extensions from newer CPUs.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
11 years agox86: h264_chromamc: port to cpuflags
Diego Biurrun [Fri, 27 Jul 2012 12:25:46 +0000 (14:25 +0200)]
x86: h264_chromamc: port to cpuflags

11 years agoyop: fix typo
Paul B Mahol [Wed, 21 Nov 2012 20:15:02 +0000 (20:15 +0000)]
yop: fix typo

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Justin Ruggles <justin.ruggles@gmail.com>
11 years agoavconv: fix copying per-stream metadata.
Anton Khirnov [Sat, 24 Nov 2012 06:55:42 +0000 (07:55 +0100)]
avconv: fix copying per-stream metadata.

It is handled separately from other types because it uses stream
specifiers and currently that triggers an assert in SET_DICT.

CC:libav-stable@libav.org

11 years agodoc: avtools-common-opts: Fix terminology concerning metric prefixes
Marcus Stollsteimer [Mon, 19 Nov 2012 20:39:20 +0000 (21:39 +0100)]
doc: avtools-common-opts: Fix terminology concerning metric prefixes

'k', 'M', and 'G' are SI (unit) prefixes or metric prefixes,
not 'number postfixes'. Also, the statement regarding binary
prefixes ("powers of 2 are used instead of powers of 10")
might be misinterpreted (1 kB = 10^3 B, but 1 KiB != 2^3 B).

Signed-off-by: Diego Biurrun <diego@biurrun.de>
11 years agoconfigure: suncc: Add compiler arch support for Nehalem & Sandy Bridge
Sean McGovern [Sun, 25 Nov 2012 06:32:57 +0000 (01:32 -0500)]
configure: suncc: Add compiler arch support for Nehalem & Sandy Bridge

GCC does not appear to have a -march= string for Westmere, which is a
bit surprising as it has a few more instructions than a Nehalem, but
a few less than a Sandy Bridge.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
11 years agoriff: Make ff_riff_tags static and move under appropriate #ifdef
Diego Biurrun [Wed, 21 Nov 2012 18:28:02 +0000 (19:28 +0100)]
riff: Make ff_riff_tags static and move under appropriate #ifdef

The table is not used outside the file.

11 years agoconfigure: sort cpuflags section by architecture
Mans Rullgard [Sun, 18 Nov 2012 15:22:04 +0000 (15:22 +0000)]
configure: sort cpuflags section by architecture

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agoconfigure: properly support DEC/Compaq compiler
Mans Rullgard [Mon, 22 Oct 2012 23:05:44 +0000 (00:05 +0100)]
configure: properly support DEC/Compaq compiler

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agoav_memcpy_backptr: optimise some special cases
Mans Rullgard [Fri, 26 Oct 2012 13:42:23 +0000 (14:42 +0100)]
av_memcpy_backptr: optimise some special cases

- Add special cases for offsets of 2, 3, or 4 bytes.  This means the
  offset is always >4 in the generic case, allowing 32-bit copies to
  be used there.
- Don't use memcpy() for sizes less than 16 bytes.

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agompegvideo: simplify dxy calculation in hpel_motion()
Mans Rullgard [Sat, 4 Aug 2012 01:57:53 +0000 (02:57 +0100)]
mpegvideo: simplify dxy calculation in hpel_motion()

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agobuild: add rules to generate preprocessed source files
Mans Rullgard [Wed, 3 Oct 2012 12:48:27 +0000 (13:48 +0100)]
build: add rules to generate preprocessed source files

This is useful for debugging.  Dependencies for these files are not
generated due to limitations in many compilers.

Signed-off-by: Mans Rullgard <mans@mansr.com>
11 years agoid3v2: fix reading unsynchronized frames.
Anton Khirnov [Wed, 21 Nov 2012 07:48:47 +0000 (08:48 +0100)]
id3v2: fix reading unsynchronized frames.

Current code would incorrectly process e.g. 'ff 00 ff 00 ff' to
'ff ff ff', while it should be 'ff ff 00 ff'.

Fixes Bug 395.

CC: libav-stable@libav.org
11 years agocdgraphics: fix incorrect vertical offset mask in cdg_scroll()
Xi Wang [Wed, 14 Nov 2012 08:42:37 +0000 (03:42 -0500)]
cdgraphics: fix incorrect vertical offset mask in cdg_scroll()

The vertical offset mask 0x07 is suspicious.

    v_off = FFMIN(data[2] & 0x07, CDG_BORDER_HEIGHT - 1);

Note that v_off is up to 11 (CDG_BORDER_HEIGHT - 1), the correct mask
should be 0x0F.

Signed-off-by: Xi Wang <xi.wang@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
11 years agoapetag: fix error handling in ff_ape_parse_tag()
Xi Wang [Wed, 14 Nov 2012 08:25:41 +0000 (03:25 -0500)]
apetag: fix error handling in ff_ape_parse_tag()

The following error handling is broken due to signedness.

    int file_size;
    uint32_t tag_bytes;
    int64_t tag_start;
    ...
    tag_start = file_size - tag_bytes - APE_TAG_FOOTER_BYTES;
    if (tag_start < 0) { ... }

Note that tag_bytes is unsigned, which makes the right-hand side of
`tag_start = ...' unsigned, too.  The 32-bit unsigned value is then
zero-extended to 64 bits.  Therefore, tag_start must be non-negative,
and the check (tag_start < 0) is always false, which breaks the error
handling.  This patch fixes the check.

Signed-off-by: Xi Wang <xi.wang@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
11 years agoflashsv: Drop unused function and struct parameters
Diego Biurrun [Wed, 24 Oct 2012 17:14:24 +0000 (19:14 +0200)]
flashsv: Drop unused function and struct parameters

11 years agopcm: fix decoding of pcm_s16le_planar on big-endian
Justin Ruggles [Tue, 20 Nov 2012 05:57:41 +0000 (00:57 -0500)]
pcm: fix decoding of pcm_s16le_planar on big-endian

The sample count is decremented by the DECODE() macro and needs to be reset
in each loop iteration. Also, DECODE() increments the src pointer so that does
not need to be done separately.

11 years agoimgconvert: remove PixFmtInfo
Luca Barbato [Sat, 17 Nov 2012 19:34:41 +0000 (20:34 +0100)]
imgconvert: remove PixFmtInfo

It is pleonastic and was used in stale functions pending replacement.

11 years agoavcodec: split avpicture from imgconvert
Luca Barbato [Sun, 11 Nov 2012 23:47:57 +0000 (00:47 +0100)]
avcodec: split avpicture from imgconvert

All the non deprecated functions are in avpicture.c now.

11 years agopixdesc: add PIX_FMT_ALPHA flag
Luca Barbato [Mon, 12 Nov 2012 00:17:20 +0000 (01:17 +0100)]
pixdesc: add PIX_FMT_ALPHA flag