August 2008 - ffmpeg-devel - lists.ffmpeg.org

Memory leak using bitstream filters with shared libs
by mans＠mansr.com 19 Feb '09

19 Feb '09

Baptiste discovered a rather nasty memory leak when using a bitstream filter with shared libraries. In ffmpeg.c:write_frame() the output of bitstream filters is stored in an AVPacket, and its destruct function is set like this (line 417): new_pkt.destruct= av_destruct_packet; Later on, in av_interleave_packet_per_dts(), AVPacket.destruct is compared against av_destruct_packet, like this (utils.c:2439): if(pkt->destruct == av_destruct_packet) The trouble here is that with shared libraries, the address of av_destruct_packet when used in the main executable or libs other than the one defining it resolves to a trampoline that calls the real function using a GOT entry. As a result, this comparison is always false when the destruct pointer was assigned outside libavformat. Does anyone have an idea for fixing this? -- M?ns Rullg?rd mans at mansr.com

7 47

[PATCH] DVB LATM parser
by paul＠kcbbs.gen.nz 16 Feb '09

16 Feb '09

Hi attached is a patch that provides a AAC LATM parser CODEC_ID_AAC_LATM I have been working on. The libfaad decoder is also patched to provide another codec the decoding for this codec id. There are minor patches to mpeg.c & mpegts.c to send stream id 0x11 to this codec id. With this patch I can get audio working perfectly for NZ DVB-T using the stock unpatched libfaad2 library. I would welcome comments and suggestions so this can be included in the ffmpeg codebase. Cheers, Paul Kendall

6 21

[PATCH] support FCP IMX/D-10 30/40/50 mov muxing
by baptiste.coudurier＠smartjog.com 12 Feb '09

12 Feb '09

Hi, This patch will correctly mux IMX/D-10 essences into a FCP compatible mov. This patch is not really meant to be included in svn (if people are against it), but since too many people are contacting me to get it, Im publishing it here. Feedback very welcome especially when muxing NTSC material. -- Baptiste COUDURIER GnuPG Key Id: 0x5C1ABAAA SMARTJOG S.A. http://www.smartjog.com Key fingerprint 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA Phone: +33 1 49966312

4 3

[PATCH] Add versioning information to dlls
by jkolb＠wsi.com 12 Feb '09

12 Feb '09

This patch adds support for generating .rc files and then linking them with the dlls created by ffmpeg. This information can be used to identify the product version by the user and is also available to installer programs for ease of updating etc. Jeremy

10 102

[PATCH] change the order of params for av_init_random()
by justinruggles＠bellsouth.net 05 Feb '09

05 Feb '09

$subject includes lavu minor version bump. -Justin

6 26

[PATCH] DV timecode
by baptiste.coudurier＠smartjog.com 02 Feb '09

02 Feb '09

Hi 3 patches: - add timecode.h to avoid code duplication for drop frame timecode adjustment - use new lavc timecode api to set timecode in DV. Regression tests change because now default behaviour is non drop frame timecode, you can still activate it by using -flags2 +drop_frame_timecode - timecode helper function for ffmpeg to set timecode. Maybe moving brktimegm to lavu and using it is better ? -- Baptiste COUDURIER GnuPG Key Id: 0x5C1ABAAA SMARTJOG S.A. http://www.smartjog.com Key fingerprint 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA Phone: +33 1 49966312

5 14

[Patch] Enable gpl if enabled nonfree
by cehoyos＠rainbow.studorg.tuwien.ac.at 19 Jan '09

19 Jan '09

Hi! IMO, it does not make much sense to be required to enable nonfree and gpl at the same time if one wants to use a nonfree and a gpl'ed library. Patch attached. Carl Eugen

9 17

[PATCH] Enhance ffmpeg.c:opt_default()
by stefano.sabatini-lala＠poste.it 19 Dec '08

19 Dec '08

Hi all, actually opt_default() will keep trying to set an option even if it fails to set it in some context. The correct (or a more correct behaviour) seems to simply give up when the option is found the first time but its value is invalid for that context, in this case the program immediatly exits. Also the patch make opt_default() distinguishes if an option wasn't found in all the checked contexts or if the option was found in some context but its value was invalid. Best regards. -- FFmpeg = Frenzy and Fantastic MultiPurpose EnGine

3 46

[PATCH] Force users to specify memalign_hack in configure
by ramiro＠lisha.ufsc.br 19 Dec '08

19 Dec '08

Hello, Currently on MinGW users are required to explicitly add --enable-memalign-hack to configure. Support for FreeDOS was recently added, and memalign-hack is automatically enabled. I think we should force all users to request memalign-hack when needed. Patch attached. Ramiro Polla

7 25

[PATCH] support for ordered chapters/segment linking in Matroska
by jbfaad＠gmail.com 10 Dec '08

10 Dec '08

Sorry for thread-breaking. Some opinions in this patch: it seems to differ a bit from the "de-facto" behavior of Haali Media Splitter (the only existing implementation). I think it'd be desirable to play along the same way, specially considering this makes things much simpler implementation-wise and I think mostly addresses Michael's comments. In Haali, to accept a segment, TrackType and CodecID must match exactly, for all tracks, in the same order. Everything else is fair game. The values from the "master" segment are used and the ones from the linked segments discarded. This includes CodecPrivate, TrackVideo's childern, TrackAudio's children, etc... Needless to say, the demuxer doesn't handle the case of changing stream parameters: it just passes the frames to the decoder, and doesn't care for anything else. IMHO this is sufficient and even desirable.

6 38