[Ffmpeg-devel-irc] ffmpeg-devel.log.20120217

[burek]

[00:29] <CIA-48> ffmpeg: 03Stefano Sabatini 07master * r7bdefc0f12 10ffmpeg/libavfilter/ (version.h vf_overlay.c): 
[00:29] <CIA-48> ffmpeg: lavfi/overlay: add logic for avoiding overlaying frames with PTS > main frame PTS
[00:29] <CIA-48> ffmpeg: Also add debug logging messages for helping tracking down similar
[00:29] <CIA-48> ffmpeg: issues.
[00:29] <CIA-48> ffmpeg: Fix trac ticket #467.
[00:30] <ubitux> saste: hi :)
[00:31] <saste> hi
[00:34] <pyoor> Hi all.  I'm new to the mpeg format and was hoping someone here might be able to answer some questions for me.  Viewing an m4a in atomboxstudio I see a number of fields in the mvhd atom similar to what's outlined here (https://developer.apple.com/library/mac/#documentation/QuickTime/QTFF/QTFFChap2/qtff2.html).  Why are only half of these listed in the ISO (jaadec.sourceforge.net/specs/ISO_14496-12_ISOBMFF.pdf)?
[00:45] <CIA-48> ffmpeg: 03Justin Ruggles 07master * rf320fb894c 10ffmpeg/ (4 files in 4 dirs): 
[00:45] <CIA-48> ffmpeg: bethsoftvid: pass palette in side data instead of in a separate packet.
[00:45] <CIA-48> ffmpeg: Update FATE reference to account for now non-existent palette packet.
[00:45] <CIA-48> ffmpeg: This also fixes the FATE test if frame data is not initialized in
[00:45] <CIA-48> ffmpeg: get_buffer(), so update comment in avconv accordingly.
[00:45] <CIA-48> ffmpeg: 03Martin Storsjö 07master * rf3a094f2da 10ffmpeg/libavformat/rtsp.c: 
[00:45] <CIA-48> ffmpeg: sdp: Ignore RTCP packets when autodetecting RTP streams
[00:45] <CIA-48> ffmpeg: The rtp demuxer which listens for RTP packets and detects the
[00:45] <CIA-48> ffmpeg: RTP payload type will currently get confused if the first packet
[00:45] <CIA-48> ffmpeg: received is an RTCP packet. Thus ignore such packets.
[00:45] <CIA-48> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[00:45] <CIA-48> ffmpeg: 03Christophe GISQUET 07master * rf3e084909b 10ffmpeg/libavcodec/x86/mpegaudiodec_mmx.c: (log message trimmed)
[00:45] <CIA-48> ffmpeg: mpegaudio: replace memcpy by SIMD code
[00:45] <CIA-48> ffmpeg: By replacing memcpy with an unrolled loop using the alignment knowledge
[00:45] <CIA-48> ffmpeg: it has, some speedup can be obtained.
[00:45] <CIA-48> ffmpeg: Before (gcc 4.6.1): ~400 cycles
[00:45] <CIA-48> ffmpeg: After: ~370 cycles
[00:46] <CIA-48> ffmpeg: Overall, around 2% speed increase when decoding a 2400s mp3 to f32le.
[00:46] <CIA-48> ffmpeg: 03Ronald S. Bultje 07master * rae591aeea5 10ffmpeg/libavcodec/vc1dec.c: 
[00:46] <CIA-48> ffmpeg: 03Justin Ruggles 07master * rc9fdf3241a 10ffmpeg/ (libavformat/bethsoftvid.c tests/ref/fate/bethsoft-vid): 
[00:46] <CIA-48> ffmpeg: bethsoftvid: synchronize video timestamps with audio sample rate
[00:46] <CIA-48> ffmpeg: According to unofficial documentation, the video rate is locked to the audio
[00:46] <CIA-48> ffmpeg: sample rate. This results in proper synchronization of audio and video
[00:46] <CIA-48> ffmpeg: timestamps from the demuxer. This only works if the first audio packet occurs
[00:46] <CIA-48> ffmpeg: before the first video packet or the audio sample rate is the default rate of
[00:46] <CIA-48> ffmpeg: 11111 Hz, both of which are true for all samples in our archive.
[00:46] <CIA-48> ffmpeg: 03Justin Ruggles 07master * r773ff823da 10ffmpeg/libavformat/bethsoftvid.c: 
[00:46] <CIA-48> ffmpeg: bethsoftvid: add audio stream only after getting the first audio packet
[00:46] <CIA-48> ffmpeg: This avoids initializing a stream with dummy values or when the file does not
[00:46] <CIA-48> ffmpeg: contain audio.
[00:46] <CIA-48> ffmpeg: Also set duration for audio packets, using the sample rate as the time base.
[00:46] <CIA-48> ffmpeg: 03Ronald S. Bultje 07master * rc6643fddba 10ffmpeg/libavcodec/golomb.h: 
[00:46] <CIA-48> ffmpeg: golomb: avoid infinite loop on all-zero input (or end of buffer).
[00:47] <saste> ubitux: do you have comments on the -count* options for ffprobe?
[00:48] <ubitux> about the latest merged patch by mateo`?
[00:48] <saste> yes
[00:48] <Compn> pyoor : good question, did you also see the mp4 atom list on http://wiki.multimedia.cx ?
[00:49] <pyoor> Compn: no I haven't
[00:49] <Compn> pyoor : not sure why there is a difference, is it possible that apple only uses part of the iso spec? 
[00:49] <Compn> http://wiki.multimedia.cx/index.php?title=MP4
[00:49] <pyoor> actually you know I have seen that
[00:50] <pyoor> but it doesn't list the structure inside the atoms
[00:50] <ubitux> saste: won't that silently infinite loop with network streams for instance?
[00:50] <ubitux> i guess only in case of specifying the option so not really important
[00:50] <saste> yes it will...
[00:51] <saste> but that's also the expected behavior
[00:51] <ubitux> anyway, i guess it is ok if tested
[00:51] <Compn> pyoor : what are you looking for that info for? going to create your own mp4 muxer ?
[00:52] <pyoor> Compn: Actually writing a file fuzzer for mp4 and quicktime based files
[00:52] <Compn> ah
[00:52] <pyoor> as for the apple question, according to wiki "MPEG-4 Part 14 is essentially identical to the QuickTime file format, but formally specifies support for Initial Object Descriptors (IOD) and other MPEG features."
[00:52] <pyoor> heh, but that's all I know
[00:52] <ubitux> saste: i'll push http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2012-February/120999.html soon, no objection?
[00:52] <pyoor> today's really the first day I've dug into the container so finding information so far has been difficult
[00:53] <Compn> i bet running zzuf on some files will cause some crashes in our mp4 demuxer :D
[00:53] <pyoor> heh, haven't seen that in a long time
[00:53] <pyoor> i develop nearly everything in peach
[00:53] <Compn> ah
[00:53] <Compn> zzuf seems kinda popular around here 
[00:54] <pyoor> probably because it requires little intervention
[00:54] <pyoor> which is great in prep time
[00:54] <pyoor> peach requires you to practically define the entire file format
[00:54] <Compn> ah
[00:54] <Compn> are you a 'security researcher' or doing it just for funs ?
[00:55] <pyoor> i.e. http://www.flinkd.org/projects/peach-pits/peach-pit-for-the-zip-file-format-specification/
[00:55] <pyoor> researcher
[00:55] <Compn> oh no xml!
[00:55] Action: Compn shields his eyes
[00:55] <pyoor> heh
[00:56] <saste> ubitux: ok if correct, that's a consequence of the av_get_channel_layout() changes I suppose
[00:56] <Compn> it might be useful to read the mp4 demuxer in ffmpeg, mplayer and vlc (they all have different mp4 demuxers)
[00:56] <Compn> plus mp4box and other mp4 tools
[00:56] <Compn> read the source, i mean
[00:56] <pyoor> right
[00:56] <ubitux> saste: yes, i had some surprises with -af; it was detecting stereo as 2.1 and similar :p
[00:56] <pyoor> that was the next step.  i thought i got lucky with the iso
[00:57] <ubitux> saste: btw, did you receive my mails?
[00:57] <Compn> pyoor : you might also want to ask in #libav-devel 
[00:57] <saste> ubitux: which one?
[00:57] <pyoor> Compn: thanks I'll try that
[00:57] <ubitux> saste: "libavfilter audio AVFrame"
[00:57] <saste> ah yes
[00:58] <ubitux> ok :)
[01:09] <CIA-48> ffmpeg: 03Matthieu Bouron 07master * r29b9aee49f 10ffmpeg/ (Changelog doc/ffprobe.texi doc/ffprobe.xsd ffprobe.c): 
[01:09] <CIA-48> ffmpeg: ffprobe: add count_frames and count_packets options
[01:09] <CIA-48> ffmpeg: Signed-off-by: Stefano Sabatini <stefasab at gmail.com>
[01:20] <kierank> you know there's nothing at http://www.ffmpeg.org/schema/ffprobe, right?
[01:34] <saste> kierank: yes, iirc the xml schema spec doesn't force the schema url to be "linkable", but I may be wrong
[01:34] <saste> anyway we may publish it on ffmpeg.org
[01:38] <ubitux> saste: what kind of rework is needed for AVFilterBufferRef?
[01:40] <ubitux> :(
[06:13] <CIA-48> ffmpeg: 03Zongyao Qu 07master * r1b84cf41e0 10ffmpeg/libavcodec/Makefile: 
[06:13] <CIA-48> ffmpeg: Makefile: fix proresdec dependancies
[06:13] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[08:39] <zamb> Hi everybody.  I tried compiling ffmpeg from git about an hour ago, and got this error:
[08:39] <zamb> libavcodec/libavcodec.so: undefined reference to `ff_jpeg_fdct_islow_10'
[08:39] <zamb> I fixed it by editing "libavcodec/Makefile" and adding "jfdctint.o"
[08:39] <zamb> to the "OBJS-$(CONFIG_PRORES_LGPL_DECODER)" line.  That is all.  Thank you.
[12:41] <ubitux> mmh& AVFrame has no channel layout& just channels :(
[12:45] <ubitux> wait no channels either? beh.
[13:51] <Daemon404> isnt that uses set in codeccontext?
[13:52] <Daemon404> usually*
[13:59] <ubitux> lavfi returns bufferref based on the avframe props
[13:59] <ubitux> so if we want lavfi to do the same as for video, we need all the props in the avframe
[13:59] <ubitux> or maybe i'm missing sth
[14:01] <ubitux> (the same for audio i meant)
[15:57] Action: Daemon404 pokes michaelni
[16:30] <michaelni> Daemon404, yes, how can i help ?
[16:31] <Daemon404> michaelni, i was going to ask you something about git cherry-pick with regards to keeping both dates, but i decided to base standalone postproc on ffmpeg
[16:31] <Daemon404> git filter-brnach is urnning now. takes 1-2 hrs.
[16:33] <j-b> hello
[16:34] <Daemon404> yes i decided to get off my ass and do it after flying to UK and job hunting for a while.
[16:34] <av500> j-b: world!
[16:45] <Daemon404> $ echo "scale=4; `grep Rewrite p.log | wc -l` / 37952 * 100" | bc | sed -e 's/\(.\{5\}\).*/\1%/'
[16:45] <Daemon404> 27.49%
[16:45] <Daemon404> god i hate filter-branch.
[16:49] <michaelni> Daemon404, you could just git clone, git rm all but libpostproc and be done with it
[16:49] <Daemon404> thats a lot of extra history and cruft left in the repo
[16:50] <Daemon404> id rather spend some cpu time now
[16:50] <Daemon404> btw michaelni, what's your take on the whole +100 postproc api version
[16:51] <Daemon404> should standalone postproc keep this?
[16:53] <michaelni> i guess it can be droped with the next minor bump
[16:53] <michaelni> but it could as well be kept
[16:53] Action: Daemon404 is impartial
[16:53] <michaelni> doesnt really do any harm
[17:01] <CIA-48> ffmpeg: 03Kostya Shishkov 07master * r857d14863d 10ffmpeg/libavcodec/proresdsp.c: prores: initialise encoder and decoder parts only when needed
[17:01] <CIA-48> ffmpeg: 03Andrew Wason 07master * ra0e44414df 10ffmpeg/libavformat/movenc.c: (log message trimmed)
[17:01] <CIA-48> ffmpeg: movenc: leave st->codec->frame_size as is instead of forcing it to 1.
[17:01] <CIA-48> ffmpeg: Muxing pcm audio in MOV using avcodec_encode_audio() was failing
[17:01] <CIA-48> ffmpeg: because avcodec_encode_audio() returns an incorrect packet size of 4
[17:01] <CIA-48> ffmpeg: bytes. This can be reproduced by modifying the sample
[17:01] <CIA-48> ffmpeg: ffmpeg/doc/examples/muxing.c to encode PCM, see ML patch
[17:01] <CIA-48> ffmpeg: muxing-test.diff
[17:31] <CIA-48> ffmpeg: 03Reimar Döffinger 07master * r3ce7148dff 10ffmpeg/libavformat/movenc.c: 
[17:31] <CIA-48> ffmpeg: movenc: fix crashes if a stream is empty.
[17:31] <CIA-48> ffmpeg: For some reason this always happens with -f ismv.
[17:31] <CIA-48> ffmpeg: See trac issue #996.
[17:31] <CIA-48> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[17:31] <CIA-48> ffmpeg: 03Reimar Döffinger 07master * r9e0f897fe5 10ffmpeg/ (configure tests/lavf-regression.sh tests/ref/lavf/ismv): 
[17:31] <CIA-48> ffmpeg: Add ismv regression test.
[17:31] <CIA-48> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[17:31] <CIA-48> ffmpeg: 03Reimar Döffinger 07master * rd493170ec9 10ffmpeg/libavformat/matroskadec.c: 
[17:31] <CIA-48> ffmpeg: matroskadec: introduce resync function.
[17:31] <CIA-48> ffmpeg: This allows handling matroska files with errors.
[17:31] <CIA-48> ffmpeg: Fixes test4.mkv and test7.mkv from the official Matroska test suite.
[17:31] <CIA-48> ffmpeg: These are also trac issues #544 and #545.
[17:31] <CIA-48> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[18:09] <lionaneesh> Hey
[18:09] <lionaneesh> michaelni, There?
[18:10] <durandal_1707> michaelni is always busy
[18:12] <Compn> hes here often tho
[18:13] <Compn> just wait around lionaneesh
[18:13] <lionaneesh> Compn, Yeah i am.
[18:18] <michaelni> lionaneesh, how can i help ?
[18:58] <Daemon404> michaelni, FINALLY i got a nice-looking repo
[18:58] <Daemon404> after some intense git-fu
[19:00] <Daemon404> all the commit dates are in-tact and
[19:00] <Daemon404> $ diff -urN ffmpeg/libpostproc/ postproc/libpostproc/ | wc -l
[19:00] <Daemon404> 0
[19:00] <j-b> nice
[19:01] <Daemon404> and git log --pretty=one | wc -l -> 185
[19:01] <Daemon404> nice and sane...
[19:02] <Daemon404> http://pastebin.com/raw.php?i=gr9SUP3M <-- nothing sane here tho.
[19:53] <Daemon404> j-b, just testing cross-compiule for powerpc (i.e. with psotproc's altivec file), and ill push to a new rpeo.
[20:05] <Daemon404> yup works :o
[20:10] <Daemon404> j-b / michaelni : https://github.com/dwbuiten/postproc
[20:44] <j-b> Daemon404: cool
[21:25] <pasteeater> moron question: i know nothing of fate/emulation. how useful would a i7 machine be for fate? it's always on and mostly idle until it has to encode several large batches DV on occasion.
[21:34] <Compn> fate is just building ffmpeg using scripts
[21:34] <CIA-48> ffmpeg: 03Carl Eugen Hoyos 07master * r44f5b26212 10ffmpeg/libavcodec/cdxl.c: Make CDXL palette opaque.
[21:34] <Compn> and cross-compiling, if you have those libs installed
[21:34] <Compn> or you could also run virtual machines and compile in that enviroment
[21:34] <Compn> itd be very useful
[21:35] <Compn> not sure what systems need coverage tho, best ask someone who knows
[21:35] <beastd> + exercising the build binaries (with samples)
[21:35] <Compn> yeah, testing (md5sum output) of all samples in the fate test
[21:36] <pasteeater> that might be a good start.
[21:40] <michaelni> what we need for fate ? ... everything thats not there yet and "build only" means just compile / that is we have no hardware to run it on
[21:41] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * rf9caec0cf9 10ffmpeg/libavcodec/x86/h264_deblock.asm: 
[21:41] <CIA-48> ffmpeg: h264: change deblock_h_chroma_8_mmxext() to prevent valgrind confusion.
[21:41] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:44] <beastd> pasteeater: just some quick thoughts. some  bsd variants are missing, some microsoft windows version and cygwin and os/2 is missing completely
[22:27] <durandal_1707> getting this: [avi @ 0x29202570]Failed to find timestamp -1456 in index 0 .. 49
[22:48] <durandal_1707> michaelni: ffmpeg summer of code 2012 wiki page is gone
[22:52] <michaelni> durandal_1707, if you have a avi that has a seeking bug, give me the file and ill fix it
[22:54] <durandal_1707> michaelni: this is created file with ffmpeg
[22:54] <durandal_1707> michaelni: sample generated from vsynth
[22:54] <michaelni> which ?
[22:56] <durandal_1707> michaelni: y41p, r210
[22:57] <durandal_1707> cljr
[22:59] <durandal_1707> v308
[22:59] <michaelni> it happens when you seek to before or after the file
[23:00] <durandal_1707> it happens if you use left/right with ffplay to seek
[23:00] <michaelni> yes
[23:01] <durandal_1707> it should not give these error, because there is no such timestamp anyway
[23:02] <ohsix> it's told to seek back that far, what's it going to do
[23:02] <ohsix> the person seeking should limit the range if it's going to seek to one that doesn't exist :D
[23:04] <CIA-48> ffmpeg: 03Paul B Mahol 07master * r545e887902 10ffmpeg/libavcodec/y41penc.c: 
[23:04] <CIA-48> ffmpeg: y41penc: switch to encode2()
[23:04] <CIA-48> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[23:04] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:04] <CIA-48> ffmpeg: 03Paul B Mahol 07master * r48ad7f822d 10ffmpeg/libavcodec/r210enc.c: 
[23:04] <CIA-48> ffmpeg: r210enc: switch to encode2()
[23:04] <CIA-48> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[23:04] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:55] <CIA-48> ffmpeg: 03Carl Eugen Hoyos 07master * rf6492476a6 10ffmpeg/cmdutils.c: Fix ffmpeg -codecs output.
[00:00] --- Sat Feb 18 2012