Kamil Nowosad [Fri, 30 Mar 2007 09:26:13 +0000 (09:26 +0000)]
Fix a bug in av_find_opt(). Because some of the AVOption structures have field
unit = NULL, the function could pass NULL to strcmp and cause a segfault.
Patch by Kamil Nowosad, k nowosad % students mimuw edu pl.
Original thread:
Subject: [PATCH] small bugfix in av_find_opt()
Date: 03/23/2007 12:20 PM
Originally committed as revision 8553 to svn://svn.ffmpeg.org/ffmpeg/trunk
Limin Wang [Fri, 30 Mar 2007 09:05:52 +0000 (09:05 +0000)]
add 'all' and 'none' options for partitions config
Patch by Limin Wang % lance P lmwang A gmail P com %
Original thread:
date: Mar 24, 2007 3:53 PM
subject: [Ffmpeg-devel] [PATCH] add all and none options for partitions config
Originally committed as revision 8552 to svn://svn.ffmpeg.org/ffmpeg/trunk
Bobby Bingham [Thu, 29 Mar 2007 05:24:35 +0000 (05:24 +0000)]
Change the vhook code to send real timestamps to the filters instead of the
current time of day, which is useless, and which the filters could just as
easily query for themselves.
patch by Bobby Bingham, uhmmmm gmail com
Originally committed as revision 8541 to svn://svn.ffmpeg.org/ffmpeg/trunk
fix decoding of big audio packets (48k 16bit 2 channels), needed size is related to samples which is short * while len passed to decode_audio2 is related to pkt->data which is uint8_t *
Originally committed as revision 8537 to svn://svn.ffmpeg.org/ffmpeg/trunk
Ivan Kalvachev [Mon, 26 Mar 2007 15:21:19 +0000 (15:21 +0000)]
get rid of unused entries in the middle of xvmc_render struct,
extensions are added to the end of the struct (avoids build issues on 64 bits platform)
add osd extension to be used by mplayer (required by the resent changes in the build system)
breaking the compability, thus bumping minor lavc version number
Originally committed as revision 8527 to svn://svn.ffmpeg.org/ffmpeg/trunk
Aurelien Jacobs [Mon, 26 Mar 2007 00:00:43 +0000 (00:00 +0000)]
properly set *data_size when returning >= 0 values in shorten_decode_frame()
Original thread:
Date: Sat, 24 Mar 2007 04:04:53 +0100
Subject: [Ffmpeg-devel] Shorten file playback broken in SVN trunk
Originally committed as revision 8523 to svn://svn.ffmpeg.org/ffmpeg/trunk
check input validity, this prevents a few variables from reachin odd values which might have lead to out of array writes and thus might have been exploitable
Originally committed as revision 8522 to svn://svn.ffmpeg.org/ffmpeg/trunk
move eob_reached logic into ff_lzw_decode_tail() which simplifies the code, avoids some checks in the innermost loop and also gets rid of the controversal break while hopefully retaining the last byte in a valid bytestream, invalid bytestreams still can have very significant overread
Originally committed as revision 8517 to svn://svn.ffmpeg.org/ffmpeg/trunk
Aurelien Jacobs [Sun, 25 Mar 2007 01:11:00 +0000 (01:11 +0000)]
add support for V_THEORA into Matroska
Original thread:
Date: Thu, 22 Mar 2007 20:23:08 -0400
Subject: [Ffmpeg-devel] [PATCH] Theora in MKV (GSoC '07 Qualification)
Originally committed as revision 8507 to svn://svn.ffmpeg.org/ffmpeg/trunk
Aurelien Jacobs [Sun, 25 Mar 2007 01:06:45 +0000 (01:06 +0000)]
extract vorbis header spliting code into a reusable function
Original thread:
Date: Thu, 22 Mar 2007 20:23:08 -0400
Subject: [Ffmpeg-devel] [PATCH] Theora in MKV (GSoC '07 Qualification)
Originally committed as revision 8505 to svn://svn.ffmpeg.org/ffmpeg/trunk
Limin Wang [Sat, 24 Mar 2007 12:07:07 +0000 (12:07 +0000)]
Fix H.264 8x8 transform selection
It seems that it's opt parse bug, it can't process the flags which start by
digit. After change 8x8dct to dct8x8, I can set it without problem. I guess
nobody will use it since it can't work as expected, so the quick fix is change
the option name.
Patch by Limin Wang lance dot lmwang at gmail com
Originally committed as revision 8492 to svn://svn.ffmpeg.org/ffmpeg/trunk
1000000l for me ive misstakely reverted the regression checksum change from r8444 in r8486
how did i succeed doing such a ridiculously silly thing? well i think it happened like:
1. verifying that the regression tests pass with old resample2.c
2. updating the regressions to the new resample2.c ... failed svn complained
3. svn up
4. updating the regressions to the new resample2.c success (r8485)
at that point everything was still ok
5. some more resample2.c work update regressions, read diff, commit (r8486)
my misstake was that the svn up at point 3 was run in tests/ -> iam an idiot
Originally committed as revision 8489 to svn://svn.ffmpeg.org/ffmpeg/trunk
Tim Chick [Fri, 23 Mar 2007 00:05:45 +0000 (00:05 +0000)]
iwmmxt ffmpeg code does not implement dct_unquantize_h263_inter_iwmmxt, but
armv5t does - use the armv5t function rather than default c code.
patch by Tim Chick chick at computergeek freeserve co uk
Originally committed as revision 8487 to svn://svn.ffmpeg.org/ffmpeg/trunk
replace long double in the audiophile kiddy mode with doubles, its faster (and more portable) and the audiophile placebo 120db stopband attenuation isnt affected
Originally committed as revision 8482 to svn://svn.ffmpeg.org/ffmpeg/trunk
remove dithering of filter coefficients, improves precision by 1-2 bits and
improves subjective sound quality on artificial sample (udial.wav resampling
to 32khz)
Originally committed as revision 8467 to svn://svn.ffmpeg.org/ffmpeg/trunk
David Conrad [Wed, 21 Mar 2007 11:05:35 +0000 (11:05 +0000)]
The NSV demuxer assumes that a video frame's timestamp increases by one on each
frame, but some low-bitrate NSV files omit video frames for some NSV frames,
and expect the timestamp to increase by one every NSV frame. This is noticeable
in 64vp3.nsv where the video runs several times faster than the audio. Fix this
by unconditionally incrementing the video's timestamp with each NSV frame.
patch by David Conrad, umovimus gmail com
Originally committed as revision 8466 to svn://svn.ffmpeg.org/ffmpeg/trunk
Wolfram Gloger [Wed, 21 Mar 2007 10:48:10 +0000 (10:48 +0000)]
av_estimate_timings_from_pts() flushes the packet queue but doesn't
reset the streams' cur_dts values. This can lead to a fatal "error,
non monotone timestamps ..." message later, because the out-of-date
cur_dts values are used to compute some packet's dts.
Fix this by calling av_read_frame_flush() and eliminate code
duplication in the process.
The additional hunk gives more detailed error messages.
patch by Wolfram Gloger, wmglo dent.med.uni-muenchen de
Originally committed as revision 8465 to svn://svn.ffmpeg.org/ffmpeg/trunk