[Ffmpeg-devel-irc] ffmpeg-devel.log.20120101

burek burek021 at gmail.com
Mon Jan 2 02:05:02 CET 2012


[00:09] <michaelni> happy new year everyone ... :)
[00:10] <microchip_> michaelni: for you too!
[00:10] <microchip_> \o/
[00:32] <ubitux> :)
[00:32] <ubitux> time to update: cmdutils.c:static const int this_year = 2011;
[02:49] <CIA-101> ffmpeg: 03Shitiz Garg 07master * rc3513477bb 10ffmpeg/libavcodec/cook.c: 
[02:49] <CIA-101> ffmpeg: cook: K&R formatting cosmetics
[02:49] <CIA-101> ffmpeg: Signed-off-by: Diego Biurrun <diego at biurrun.de>
[02:49] <CIA-101> ffmpeg: 03Janne Grunau 07master * rc11b92b304 10ffmpeg/Changelog: 
[02:49] <CIA-101> ffmpeg: Changelog: restore version <next> header
[02:49] <CIA-101> ffmpeg: Indeo 4 decoder and SMJPEG demuxer are not included in 0.8_beta1. Move
[02:49] <CIA-101> ffmpeg: them to version <next>.
[02:49] <CIA-101> ffmpeg: 03Janne Grunau 07master * rbcc7396065 10ffmpeg/ (configure libavcodec/pthread.c): 
[02:49] <CIA-101> ffmpeg: threads: add sysconf based number of CPUs detection
[02:49] <CIA-101> ffmpeg: Can act as fallback and should work on a couple of Unix systems.
[02:49] <CIA-101> ffmpeg: 03Justin Ruggles 07master * r481a7ff5bd 10ffmpeg/libavcodec/mpegaudiodec.c: mpegaudiodec: fix compilation when testing the unchecked bitstream reader
[02:49] <CIA-101> ffmpeg: 03Ronald S. Bultje 07master * r464f26889c 10ffmpeg/libavcodec/mpegaudiodec.c: 
[02:49] <CIA-101> ffmpeg: mpegaudiodec: fix seeking.
[02:49] <CIA-101> ffmpeg: The safe bitstream reader does not allow using skip_bits_long() to seek to a
[02:49] <CIA-101> ffmpeg: point before the start of the buffer, which was needed by the mp3 decoder.
[02:49] <CIA-101> ffmpeg: This change instead calculates the start point of the first valid granule and
[02:49] <CIA-101> ffmpeg: skips to that position.
[02:49] <CIA-101> ffmpeg: Signed-off-by: Justin Ruggles <justin.ruggles at gmail.com>
[02:49] <CIA-101> ffmpeg: 03Oana Stratulat 07master * rfebd022228 10ffmpeg/tests/ (13 files in 3 dirs): 
[03:52] <CIA-101> ffmpeg: 03Michael Niedermayer 07master * r57581e4110 10ffmpeg/libavcodec/pthread.c: 
[03:52] <CIA-101> ffmpeg: pthread: dont autoenable frame threads with visualization
[03:52] <CIA-101> ffmpeg: This combination doesnt work
[03:52] <CIA-101> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:52] <CIA-101> ffmpeg: 03Michael Niedermayer 07master * r490c97bdf5 10ffmpeg/libavformat/mpegenc.c: 
[03:52] <CIA-101> ffmpeg: mpegpsenc: Fix integer overflow of the muxrate calculation.
[03:52] <CIA-101> ffmpeg: this should fix the failure of h264-bsf-mp4toannexb on freebsd/openbsd
[03:52] <CIA-101> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:44] <CIA-101> ffmpeg: 03Carl Eugen Hoyos 07master * r03d7d8faef 10ffmpeg/libavcodec/pamenc.c: Support gray8a pam encoding.
[05:04] <CIA-101> ffmpeg: 03Carl Eugen Hoyos 07master * r8dc973e6d1 10ffmpeg/cmdutils.c: Update copyright year.
[10:38] <Granjow> Why does this http://codepad.org/HruatjSg segfault in valgrind with an invalid write in avpicture_alloc on line 70?
[10:40] <Granjow> Backtrace looks like this http://codepad.org/6PmmKl1h
[10:44] <nevcairiel> you allocate a rgb24 buffer, but for the rgb stride you use width*4 .. shouldnt that be *3? 24-bit and all
[10:46] <nevcairiel> (also, i would make sure that the linesize is at least 16-byte aligned)
[10:46] <nevcairiel> but that may cause issues with writing the file, but huh
[10:46] <nevcairiel> the first one is a clear mistake, imho
[10:48] <Granjow> nevcairiel: Wow! Indeed, this fixed the invalid write.  How can I ensure alignment?
[10:49] <nevcairiel> there is the FFALIGN macro, however i don't know if PNG can deal with that. :)
[10:50] <nevcairiel> probably should
[10:50] <nevcairiel> just use FFALIGN(3*pCodecContext->width, 16) and see if everything still works
[10:51] <nevcairiel> oh
[10:51] <nevcairiel> need to do that when allocating the picture, too, of course
[10:53] <nevcairiel> i would recommend to use av_image_alloc instead of avpicture_alloc, it lets you specify an alignment, and you don't need to do anything manually
[10:56] <Granjow> Still works. At least the part that worked before :) http://codepad.org/UTRIChQa line 92; did I use it correctly?
[10:56] <Granjow> I have an invalid read on line 135 now: http://codepad.org/9mBvQbYk but in my opinion I passed all parameters correctly. 
[10:57] <nevcairiel> yes, however your buffer might be too small now, you only allocate it for "width" but with the alignment it may use more
[10:59] <Granjow> So I should use int rgbLines[4+2] approximately for the additional 16 bits?
[11:04] <Granjow> Hm. Changed to av_picture_alloc now and the invalid read is gone; however I get:    [IMGUTILS @ 0x7fefffdc0] Picture size 0x0 is invalid   [NULL @ 0xce45f80] buffer smaller than minimum size 
[11:06] <nevcairiel> that would only happen if you give it width/height of 0
[11:10] <Granjow> Yes; I thought if I already had to give width/height to av_image_alloc it would also fill in those fields, but it did not so they were still 0. Again the same invalid read as I posted above.
[11:16] <Granjow> sws_scale seems to break a lot :S When I watch pRgbFrame->data in the debugger, it disappears after this call.
[11:24] <Granjow> If I got this correctly then I can use the linesizes created by av_image_alloc, right? Do I have any chance now to find out why sws_scale does not do what it should? http://codepad.org/tIjOD0GT line 100
[11:26] <Granjow> pFrame->data is filled in by avcodec_decode_video2, pFrame->linesize too, so nothing I could change there. pRgbFrame fields are filled by av_image_alloc. pCodecContext->height is the correct height too according to my debugger.
[11:32] <nevcairiel> looks fine so far
[11:33] <nevcairiel> one thing i noticed is that you seem to allocate a buffer further down but never free it (the "buf" buffer), but unless you let it run for a while and it runs out of memory, thats not the cause i guess :p
[11:34] <Granjow> I currently don't even run it once because of the segfault :D
[11:37] <Granjow> But you're right, I'm freeing it now. What else could be wrong? Is the swsContext bad?
[11:37] <Granjow> Or my call to avcodec_decode_video2?
[11:39] <Granjow> Do I need to av_image_alloc pFrame too?
[11:53] <Granjow> Tjoppen: If you happen to be online -- Do you maybe have an idea how I can find out where the problem lies? I get a segfault on line 138 (avcodec_encode_video) here http://codepad.org/UmbptCkV, looks like this: http://codepad.org/A8sVNz1j 
[11:54] <Granjow> In my debugger some local variables already disappear after the sws_scale call.
[12:11] <Granjow> :(
[12:13] <nevcairiel> hm
[12:14] <nevcairiel> dont you need to call avcodec_open on the encoder, too
[12:14] <nevcairiel> you dont seem to do that
[16:14] <cbsrobot> michaelni: are you around ?
[16:19] <CIA-101> ffmpeg: 03Stefano Sabatini 07master * r42a8ac94d9 10ffmpeg/ (3 files in 2 dirs): 
[16:19] <CIA-101> ffmpeg: vf_tinterlace: implement interlace mode 5
[16:19] <CIA-101> ffmpeg: Allow creating interlaced bottom field first video.
[16:40] <michaelni> cbsrobot, yes, how can i help ?
[16:45] <CIA-101> ffmpeg: 03Jean First 07master * r2e0efa332b 10ffmpeg/libavfilter/af_astreamsync.c: 
[16:45] <CIA-101> ffmpeg: af_astreamsync: remove unused variable
[16:45] <CIA-101> ffmpeg: Signed-off-by: Jean First <jeanfirst at gmail.com>
[16:49] <cbsrobot> michaelni: the flvdec patch i sent
[16:49] <cbsrobot> do you mean to initialise like this :
[16:49] <cbsrobot> int av_unused sample_rate = 0;
[16:49] <cbsrobot> or
[16:49] <cbsrobot> int av_unused sample_rat;
[16:50] <cbsrobot> *rate;
[16:50] <michaelni> av_uninit(sample_rate)
[16:50] <michaelni> the unused was a typo, sorry
[16:52] <cbsrobot> need to run ...
[17:42] <durandal_1707> can we get more y41p samples?
[17:44] <durandal_1707> i think that this codec just always encodes image bottom to up
[17:57] <durandal_1707> is #322 so hard to fix?
[21:48] <durandal_1707> michaelni: so what I should do with this flipping for Y41P? Apparently other software/hardware can output this format too (v4l2)...
[21:52] <CIA-101> ffmpeg: 03Michael Niedermayer 07master * recb14b8af7 10ffmpeg/libavcodec/h264.c: 
[21:52] <CIA-101> ffmpeg: h264: add forgotten \n to "Invalid mix of idr and non-idr slices"
[21:52] <CIA-101> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:52] <CIA-101> ffmpeg: 03Michael Niedermayer 07master * r393253b4b3 10ffmpeg/libavcodec/h264.c: 
[21:52] <CIA-101> ffmpeg: h264: Print pass in NAL debug output
[21:52] <CIA-101> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:52] <CIA-101> ffmpeg: 03Jean First 07master * r6bd8d61378 10ffmpeg/libavformat/mov.c: 
[21:52] <CIA-101> ffmpeg: mov: av_dlog use PRId64 instead of %ld
[21:52] <CIA-101> ffmpeg: Signed-off-by: Jean First <jeanfirst at gmail.com>
[21:52] <CIA-101> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:52] <CIA-101> ffmpeg: 03Michael Niedermayer 07master * r36fa781bd6 10ffmpeg/libavcodec/h264.c: 
[21:52] <CIA-101> ffmpeg: h264: reset picture->sync on allocation.
[21:52] <CIA-101> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:52] <CIA-101> ffmpeg: 03Michael Niedermayer 07master * rfba6777363 10ffmpeg/libavcodec/h264.c: 
[21:52] <CIA-101> ffmpeg: h264: Fix recovery_frame initialization when recovery_frame is -1
[21:52] <CIA-101> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:52] <CIA-101> ffmpeg: 03Michael Niedermayer 07master * r9ba43e50ef 10ffmpeg/libavcodec/ (h264.c h264_refs.c): 
[21:57] <michaelni> durandal_1707, iam not sure i understood the fliping issue?
[21:57] <michaelni> does some software output it top down and some bottom up ?
[21:59] <michaelni> i thoight your latest patch fixed the flip issue ?
[22:00] <durandal_1707> i dunno, i have only one sample, both xanim and bt20 do flipping
[22:01] <durandal_1707> s/bt20/bt411 via mplayer
[22:04] <michaelni> hmm, t
[22:04] <michaelni> hmm, try setting codec_tag in the encoder
[22:09] <durandal_1707> how that should help? avi muxer already write Y41P twice
[22:10] <michaelni> does it write positive height ? 
[22:10] <durandal_1707> yes
[22:10] <michaelni> hmm, the linked bug was speaking of negative ....
[22:11] <durandal_1707> yeah, I just thought that was causing crash but apperently it is something else
[22:11] <michaelni> could be all the docs are wrong and Y41P with positive height is actually fliped
[22:11] <michaelni> someone with a windows box that supports it could test that 
[22:12] <durandal_1707> but sample which works have not flipped height
[22:12] <michaelni> can i have that sample ?
[22:13] <durandal_1707> http://ffmpeg.org/trac/ffmpeg/ticket/102
[23:13] <rahulr92> Hi
[23:17] <rahulr92> Hi I am new to open source and would like to contribute to FFmpeg. I picked the following task from the Small FFmpeg Tasks page: Implement a Vivo demuxer. It corresponds to ticket #132 https://avcodec.org/trac/ffmpeg/ticket/132 .I have went through mplayer demux_viv.c code, but could use additional assistance and pointers. Any guidance will be greatly appreciated. Thanks.
[23:19] <durandal_1707> rahulr92: there is already patch for that demuxer and it is 2 years old (i have synced it with master locally but it segfaults though)
[23:21] <rahulr92> durandal_1707: Alright. So I think I will scout for another beginner level task. 
[23:21] <durandal_1707> you can still work on it
[23:22] <durandal_1707> but it may be hard for begginer...
[23:23] <rahulr92> Okay. So can you please direct me towards something suitable for a beginner.
[23:24] <CIA-101> ffmpeg: 03Paul B Mahol 07master * rc0f7c574e9 10ffmpeg/libavcodec/ (j2kdec.c mjpegenc.c): Cosmetics: Replace -1 with PIX_FMT_NONE.
[23:27] <rahulr92> How about Port missing demuxers from MPlayer? Are they similarly hard?
[23:27] <rahulr92> TiVo?
[23:28] <durandal_1707> well if you can get samples...
[23:29] <durandal_1707> ah they are on usual location
[23:29] <durandal_1707> whan is almost 1GB though
[23:31] <rahulr92> Sorry, but I am totally new to this. Slight hand holding or reference would be of great help.
[23:31] <rahulr92> samples?
[23:31] <durandal_1707> ty file
[23:33] <durandal_1707> http://wiki.multimedia.cx/index.php?title=Tivo
[23:34] <rahulr92> On a differeent note, how do I create an account in the wiki page.
[23:35] <rahulr92> Is it same for all wikimedia sites?
[23:37] <durandal_1707> no you need to contact mike at multimedia.cx
[23:38] <rahulr92> Alright. I will go through the documentation first and be back. Thanks a lot.
[23:40] <rahulr92> Just I doubt, should I send a mail to mike at multimedia.cx requesting an account?
[23:41] <michaelni> durandal_1707, xanim seems to unconditionally flip y41p
[23:41] <michaelni> see AVI_Decode_Y41P()
[23:42] <durandal_1707> michaelni: yes I'm aware of that, i'm more interested in encoders side of story :)
[23:43] <durandal_1707> rahulr92: if you want to edit wiki than yes
[23:43] <rahulr92> Ok. Thanks
[23:45] <michaelni> its very likely the encoder will flip too
[23:45] <michaelni> otherwise the existing sample is hard to explain
[23:50] <durandal_1707> there is flipped y41p colorspace thread on doom9 forum
[23:55] <michaelni> maybe theres a fliped and non fliped one
[23:56] <michaelni> id say add support for the sample we have and let users find a sample that is the other way around
[23:56] <michaelni> then we can figure out what the difference is and support it too
[00:00] --- Mon Jan  2 2012


More information about the Ffmpeg-devel-irc mailing list