[Ffmpeg-devel-irc] ffmpeg-devel.log.20131010

[burek]

[00:17] <cone-239> ffmpeg.git 03Michael Niedermayer 07master:5ac6b6028f17: avcodec/h264_refs: modify key frame detection heuristic to detect more cases
[02:43] <cone-239> ffmpeg.git 03Michael Niedermayer 07master:65c2fe7159f7: avcodec/h264: call flush_change() on init
[03:39] <BBB> michaelni: yes fixed
[05:26] <mathstuf> hi, im looking at adding support for mid-stream ogg comment changes to libavcodec to get support for printing changes in an mpd stream in mpv
[05:26] <mathstuf> i see "av_log(avccontext, AV_LOG_DEBUG, "Ignoring comment header\n");"
[05:26] <mathstuf> in vorbisdec.c
[05:27] <mathstuf> the comment parsing code is in avformat/oggparsevorbis.c currently; would it make sense to move the comment parsing into vorbisdec.c and call it from avformat or is there a better way?
[05:28] <mathstuf> hrm, seems flac's picture parsing would go as well
[05:40] <mathstuf> nope...looking at the call pattern from mpv, it looks like decode() needs some way to signal that a metadata packet is incoming and that the header parsing function on the avformat for the codec needs to be called
[06:18] <Compn> mathstuf : just want to warn you, many developers have tried to work on ogg, but gave up :)
[06:18] <Compn> good luck man!
[06:19] <mathstuf> well, i just spent the last week in matroska edition land, so who knows
[06:19] <mathstuf> and the code is all there...it "just" needs to be hooked together
[10:38] <cone-976> ffmpeg.git 03Luca Barbato 07master:1018a92219a3: jpeg2000: Check block length
[10:38] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:e96c80154f7c: Merge commit '1018a92219a38a812cf97761c6b3a5e66a400f4b'
[10:40] <durandal_1707> saste: that command you mentioned for interleave does not work
[10:41] <durandal_1707> try with different inputs, one testsrc, other smptebars or similar
[10:41] <durandal_1707> i can't get one input in left and another in right
[10:45] <cone-976> ffmpeg.git 03Luca Barbato 07master:1cac9accbd1f: avi: properly fail if the dv demuxer is missing
[10:45] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:c589c4c52d13: Merge commit '1cac9accbd1f9b8596122d0735e37b97a844c514'
[11:29] <cone-976> ffmpeg.git 03Luca Barbato 07master:756547ce7f4d: avi: DV in AVI must be considered single stream
[11:29] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:dd0429e1de54: Merge commit '756547ce7f4d67a18663503e2157aebed3531703'
[11:37] <cone-976> ffmpeg.git 03Luca Barbato 07master:ceec6e792e4b: avi: directly resync on DV in AVI read failure
[11:37] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:5fcd0a80a25d: Merge commit 'ceec6e792e4b5baaa23b220f4fd33417631f5288'
[11:43] <cone-976> ffmpeg.git 03Martin Storsjö 07master:41a10f3ba149: vp6: Support cropping to AVCodecContext.width/height
[11:43] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:70025c07a703: Merge commit '41a10f3ba149a2012de499d0b4ad4955d81f28d5'
[11:51] <cone-976> ffmpeg.git 03Martin Storsjö 07master:4d7aeff9a108: flvenc: Don't pretend to support muxing "plain" VP6
[11:51] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:0efbd98e2331: Merge commit '4d7aeff9a1080e64125f3374d993b2c9c4c2751b'
[12:04] <cone-976> ffmpeg.git 03Martin Storsjö 07master:be1e1373d267: flvenc: Support muxing VP6A as well
[12:04] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:c2555bcbd147: Merge commit 'be1e1373d267bae2af8a62d79eef736736f24565'
[12:13] <cone-976> ffmpeg.git 03Martin Storsjö 07master:e6ed8668597c: flvenc: Write proper cropping for VP6 even if there's no extradata
[12:13] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:7acb464f0e8c: Merge commit 'e6ed8668597cfea25dfb350a9b4df7fb2efc1d90'
[12:21] <cone-976> ffmpeg.git 03Luca Barbato 07master:44690dfa683f: prores: Add a codepath for decoding errors
[12:21] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:5c99ca2288f5: Merge commit '44690dfa683f620c77e9f0e8e9bc5682608636b1'
[12:29] <cone-976> ffmpeg.git 03Luca Barbato 07master:2df7f7714a12: prores: Error out only on surely incomplete ac_coeffs
[12:29] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:3fa4c528a92f: Merge remote-tracking branch 'qatar/master'
[13:18] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:0391f4617f1b: avcodec/vmnc/reset_buffers(): reset cur_hx/hy too for saftey
[13:18] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:ef6b85e9b825: avcodec/proresdec2: propagate error codes
[13:36] <cone-976> ffmpeg.git 03Ronald S. Bultje 07master:20d78a86064a: libavcodec/x86: Fix emulated_edge_mc SSE code to not contain SSE2 instructions on x86-32.
[13:41] <durandal_1707> ubitux: that ts_unit approach does not generally work as it resets pts to 0, which is not nice for audio
[13:41] <ubitux> ah?
[13:41] <ubitux> why 0 is wrong?
[13:42] <durandal_1707> video stream pts may start from !=0, while audio starts from 0
[13:42] <durandal_1707> for example in ts
[13:44] <ubitux> ah, i see
[13:45] <ubitux> i guess you could have a ts_start
[13:45] <ubitux> from the first frame
[13:45] <ubitux> to use as a basis
[13:45] <ubitux> if you have ts jumps you're doomed though
[13:46] <ubitux> but well, not worse than currently
[13:47] <durandal_1707> i just stole code from yadif
[13:49] <cone-976> ffmpeg.git 03Timothy Gu 07master:7ab8bceabc33: doc/encoder/aac: Remove -cutoff trick
[13:51] <ubitux> durandal_1707: did you submit/commit anything?
[13:51] <ubitux> ah, just now in w3fdid ok
[13:52] <ubitux> dif*
[14:36] <lhs> hi
[16:04] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:cdc9657a0c7b: doc/codecs: Document field_order
[17:12] <cone-976> ffmpeg.git 03Paul B Mahol 07master:d1d875d30138: avfilter/vf_w3fdif: do not reset pts to 0
[18:08] <cone-976> ffmpeg.git 03Michael Niedermayer 07master:a830915b78e2: swscale: Use full resolution chroma for yuv2rgb when the input has full chroma
[18:41] <cone-976> ffmpeg.git 03Maxim Poliakovski 07master:4fb8de08e110: atrac3: Remove unused gain compensation tables
[18:41] <cone-976> ffmpeg.git 03Maxim Poliakovski 07master:2d52834910a4: atrac3: Better name for imdct window initialization
[19:53] <llogan> BBB: i suppose pre SSE is too old for your testing (I am ignorant of the topic and I didn't really follow thread)
[20:30] <cone-976> ffmpeg.git 03Paul B Mahol 07master:996c73490c29: avcodec/exr: remove pointless initialization
[21:44] <durandal11707> i get following spam: deprecated pixel format used, make sure you did set range correctly
[22:13] <durandal_1707> michaelni: pts resets to 0 when input size changes when filtering
[22:14] <durandal_1707> this is suboptimal
[22:25] <rcombs> 0x1010101 feels like a very black-magic-y constant
[22:25] <rcombs> anyone know what it means? (it's in the ff_draw functions)
[22:26] <Compn> nsa backdoor
[22:26] <rcombs> nah, it's something with color blending
[22:26] <rcombs> but fuck if I know why
[22:27] Action: rcombs is noticing some funny artifacting around subtitles drawn with ff_draw, and is trying to determine if that's coming from the draw functions or from x264 settings, and can't understand what any of this code's actually doing
[22:29] <rcombs> like, blend_pixel appears to loop across the entire draw area
[22:32] <durandal_1707> rcombs: its just for rounding so no floats are used....
[22:33] <iive> rcombs: can you point to the code in question?
[22:33] <durandal_1707> blend_pixel/line
[22:50] <rcombs> I'm looking at http://ffmpeg.org/doxygen/trunk/drawutils_8c_source.html#l00398
[22:56] <durandal_1707> just rewrite it
[22:57] <rcombs> I think it's just H.264 artifacts anyway, but that doesn't stop me from being concerned about how confusing and undocumented that code is
[22:57] <durandal_1707> with this unmodified code?
[22:58] <rcombs> huh?
[22:58] <durandal_1707> ff_draw you use is unmodified or?
[22:58] <rcombs> oh, yeah
[22:59] <durandal_1707> well you can always use lossless codecs to make sure its not that code fault
[23:01] <rcombs> yeah, I'll be screwing around with this a bit more
[00:00] --- Fri Oct 11 2013