[Ffmpeg-devel-irc] ffmpeg-devel.log.20120531

[burek]

[01:02] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r76853a3e0c 10ffmpeg/libavformat/ (avidec.c riff.c riff.h wtvdec.c): 
[01:02] <CIA-119> ffmpeg: libavformat: ff_get_bmp_header: return esize too
[01:02] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:02] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r57778f61d0 10ffmpeg/libavformat/avidec.c: 
[01:02] <CIA-119> ffmpeg: avidec: fix odd extradata size case.
[01:02] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:02] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r21fdf1ccf0 10ffmpeg/libavcodec/ffv1.c: 
[01:02] <CIA-119> ffmpeg: ffv1: fix reading global header with CRC
[01:02] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:02] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r0dfcd7e7e3 10ffmpeg/libavcodec/rangecoder.c: 
[01:02] <CIA-119> ffmpeg: range coder test code: decoding errors are av_log_error not debug.
[01:02] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:34] <saste> michaelni: ffplay -f lavfi "testsrc,format=yuva420p,lutrgb=g=0:b=0"
[02:34] <saste> michaelni: hints?
[02:34] <saste> i'd expect a red output, but something funny is going on with the alpha channel
[02:38] <saste> wait maybe it's a bug in lutrgb
[02:40] <michaelni> saste, seems r<->b and a<->g are swaped or something
[02:40] <michaelni> i mean b=0 kills red, r=0 kills blue
[02:44] <saste> i suspect it is the conversion yuva420p -> argb
[02:45] <saste> ffplay -f lavfi -i "color=red,format=yuva420p,lutrgb=r=0"
[02:50] <saste> bah forget seems a lut problem...
[03:14] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r3b275ce374 10ffmpeg/tests/ (rotozoom.c videogen.c): 
[03:14] <CIA-119> ffmpeg: fate: dont freopen() stdout
[03:14] <CIA-119> ffmpeg: it doesnt work and breaks most BSDs
[03:14] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:08] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r8aaa00c301 10ffmpeg/libavcodec/indeo5.c: 
[05:08] <CIA-119> ffmpeg: indeo5: check quant_mat
[05:08] <CIA-119> ffmpeg: prevents out of array read
[05:08] <CIA-119> ffmpeg: Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
[05:08] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[07:19] <ubitux> michaelni: do you have a public repo where i can checkout the colors branch (the patchset doesn't apply anymore and i feel lazy)
[07:19] <ubitux> ?
[11:18] <michaelni> ubitux, https://github.com/michaelni/FFmpeg/tree/coloring
[11:18] <ubitux> ah, thx :)
[11:22] <CIA-119> ffmpeg: 03Paul B Mahol 07master * r824a6975ee 10ffmpeg/libavcodec/binkaudio.c: 
[11:22] <CIA-119> ffmpeg: binkaudio: check number of channels
[11:22] <CIA-119> ffmpeg: Fixes #1380.
[11:22] <CIA-119> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[12:13] <durandal_1707> hmm, why swscale cant do RGBA64le to RGBA64be conversion?
[12:13] <durandal_1707> that is pretty trivial
[12:15] <av500> add it
[12:15] <av500>  that is pretty trivial
[12:17] <durandal_1707> ENOTIME
[12:17] <av500> ENORANT
[12:21] <cbsrobot> durandal_1707: I cant produce single channel exr ... sorry
[12:21] <cbsrobot> I tried yesterday but I always get rgb files
[12:22] <durandal_1707> cbsrobot: and YUV(A) one?
[12:22] <cbsrobot> well as I said - it's always rgb - so no - sorry
[12:22] <durandal_1707> what utilitility you used? ImageMagick?
[12:22] <cbsrobot> nuke
[12:23] <durandal_1707> I looked at ImageMagick code and it always outputs RGB, it uses openexr lib
[12:24] <cbsrobot> and the spec says it's possible ?
[12:24] <durandal_1707> i just wanted to know what YUV colorspace is supported by exr
[12:25] <durandal_1707> i already know that is supports gray16, gray16 with alpha
[12:26] <durandal_1707> it uses strings to specify color channels (yet another "smart" idea)
[12:26] <cbsrobot> well in my understanding you can put multiple pipes with multiple channels in one exr
[12:27] <cbsrobot> and you can name them the way you like
[12:27] <durandal_1707> defined ones in spec are "R", "G", "B", "A", "Y", "RY" and "BY"
[12:27] <cbsrobot> it's useful for 3d programs putting rgb, alpha, z, shadow, diffuse etc. channels in one file
[12:28] <durandal_1707> and Y + RY + BY are perhaps YUV420p16
[12:29] <cbsrobot> my guess it's rather yuv444p16 (in float) or yuv444p32 (in float)
[12:40] <michaelni> Tjoppen: poke poke, its morning -> michaelni: kinda late, poke me in the morning <-
[12:45] <durandal_1707> michaelni: is exr decoding lossless?
[12:46] <durandal_1707> our decoder and ImageMagick one outputs different data
[12:55] <Tjoppen> right
[12:55] <Tjoppen> I'll just cross-reference the specs
[12:55] <Tjoppen> looks fine afaict
[12:57] <michaelni> durandal_1707, exr does float->int so it doesnt look lossless 
[13:00] <durandal_1707> ffmpeg tiff  and ImageMagick tiff are different
[13:00] <ubitux> TimNich: heh i'm working on it today ;)
[13:00] <ubitux> (mov timecode)
[13:00] <ubitux> btw, i've added it to the AVI too
[13:00] <ubitux> (pretty trivial after the move to the metadata)
[13:01] <durandal_1707> michaelni: so it is more like bug in code or decoder is just hack
[13:03] <michaelni> we would have to output exr floats instead of 8/16bit ints for samples for it to be lossless
[13:03] <michaelni> that would need float based pixel formats
[13:04] <durandal_1707> but conversion to int should be bit exact as done in openexr (depending on spec)
[13:04] <michaelni> yes, if its not its a bug
[13:06] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r60de761657 10ffmpeg/ffmpeg.c: 
[13:06] <CIA-119> ffmpeg: ffmpeg: allow overriding input top field first
[13:06] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[13:58] <TimNich> ubitux: Yes, I thought the move might make life easier overall...
[13:59] Action: TimNich currently trying to get my head around the move from codec_regressions.sh to makefile....
[14:00] Action: TimNich but deciding that a beer in the sun is a better idea...
[14:01] <kierank> the sun has gone here :(
[14:01] <ubitux> yes, at last.
[14:01] Action: ubitux hates the sun
[14:01] <ubitux> btw i think it's better to have all the tests in one place, the fate/*.mak are clear enough
[14:02] <ubitux> all the tests/*.sh called implictely are not easy to get
[15:42] <TimNich> ubitux:  I am sure your right, is just a case of familiarising myself with the new layout....
[15:42] <ubitux> :)
[15:42] <ubitux> i found the issue with the duration
[15:43] <ubitux> should be fixed soon©®"
[15:50] <TimNich> I can give it a good thrashing to try and break it when you are ready )
[15:50] <ubitux> i'll share the patchset asap
[15:51] <burek> does anyone know what does this err msg mean: "[buffer @ 0x1c3839c0] Changing frame properties on the fly is not supported." -> http://pastebin.com/8FUHLcY5
[15:51] <burek> I've never seen it before
[15:57] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r76170f5373 10ffmpeg/libavformat/yop.c: 
[15:57] <CIA-119> ffmpeg: yop_probe: check more fields to avoid detecting random data as yop
[15:57] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:58] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r6cfaccabc4 10ffmpeg/tools/probetest.c: 
[15:58] <CIA-119> ffmpeg: probetest: allow specifying parameters on the command line
[15:58] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:28] <michaelni> ubitux, jacosub_probe() is calling sscanf() in a loop which is pretty slow ...
[16:30] <michaelni> is there some easy way to avoid or reduce this for probing ?
[16:32] <michaelni> and jacosub probe fails: tools/probetest 256 4096
[16:33] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rebfe0c6eb8 10ffmpeg/libavformat/ac3dec.c: 
[16:33] <CIA-119> ffmpeg: ac3_probe: speedup by checking for header earlier
[16:33] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:33] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rcc4d80c99f 10ffmpeg/libavformat/ipmovie.c: 
[16:33] <CIA-119> ffmpeg: ipmovie_probe: speedup by avoiding memcmp() call
[16:33] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:33] <ubitux> michaelni: yes it should be possible
[16:33] <ubitux> skipping the blank and '#' lines
[16:34] <ubitux> and then just try 3-4 timed_line()
[16:35] <ubitux> michaelni: http://unicorn.us.com/jacosub/demo.txt
[16:35] <ubitux> actually, just one timed_line() call should be enough
[16:35] <ubitux> as long as you skip '#' and blank lines
[16:35] <ubitux> (assuming trimming for every line)
[16:36] <saste_> mateo`: you there?
[16:37] <michaelni> ubitux, want to work on it or should i try ?
[16:37] <ubitux> michaelni: if i do, not until a few hours
[16:37] <michaelni> no hurry
[16:39] <ubitux> saste_: this loafer is taking a break, he will be back soon ;)
[16:40] <saste_> mateo`: http://gitorious.org/~saste/ffmpeg/sastes-ffmpeg/commits/add-movie-sink2-20120531
[16:40] <saste_> ubitux: it would also be possible to merge movie+amovie, and specify dynamically the number of supported inputs/outputs
[16:41] <saste_> so we have a libavfilter replacement for ffmpeg
[16:43] <ubitux> awesome!
[16:43] <Daemon404> TimNich, where can i find that lgpl patch?
[16:47] <Daemon404> -rw-rw-r-- 1 derekb derekb 252K Jun 18  2010 ffmpeg-DVCPROHD-0.5.patch
[16:47] <Daemon404> -rw-rw-r-- 1 derekb derekb 361K Jun  2  2010 ffmpeg-DVCPROHD-r15593.patch
[16:47] <Daemon404> holy crap
[16:47] <Daemon404> a bit hefty
[16:48] <Tjoppen> I tried to port it a while back. let's just say I gave up
[16:49] <mateo`> saste_: this is awesome :)
[16:49] <Daemon404> Tjoppen, well ffmpeg could always merge a gpl encoder
[16:50] <TimNich> Well the ingex guys were looking at updating their ffmpeg to something newer, so if they do that would break the back of the port..
[16:51] <Daemon404> TimNich, i thought bbc just use ffmbc 
[16:51] <Daemon404> why would they update it?
[16:51] <TimNich> Daemon404:  Where did you find that one?
[16:51] <Daemon404> ingex's cvs
[16:51] <Daemon404> (yes! CVS!)\
[16:51] <TimNich> Ahhh. Last time I looked they had omitted the source...
[16:52] <Daemon404> there are about 18 patches
[16:52] <Daemon404> ...to 0.5 and and otehr misc ancient vers
[16:52] <TimNich> well they having been talking about moving to git for centuries. They only thing that has moved is the bmxlib stuff
[16:52] <Daemon404> i think it would be far easier to port a patch series than one massive blob like this
[16:53] <TimNich> I did say it was prehistoric.. but then  they did update their mxf patches which (mostly) got into upstream ffmpeg
[16:54] <Daemon404> i hate to ask the obvious, but has anyone actually just asked baptiste about backport to ffmpeg?
[16:55] <TimNich> Well his stuff is GPL (although the original was LGPL) so it would be fair game I suppose....
[16:55] <Daemon404> i dont actually see him responding to that original threads
[16:55] <Daemon404> -s
[16:58] <mateo`> here is some stuff about DVCPRO HD: f8007a14734ff9fc8a4dd0d0194ab2c645f72578
[16:59] <ubitux> decoder only :p
[16:59] <mateo`> sorry this is just the decoder :(
[16:59] <burek> is there an option in ffmpeg to just "extract" some data stream into an outfile (without muxing) ?
[17:00] <TimNich> just run the code backwards then ;)
[17:00] <ubitux> |rot13
[17:00] <av500> burek: mux into raw format
[17:01] <burek> -f rawvideo?
[17:02] <av500> probaj
[17:03] <burek> rawvideo doesn't work :( http://tonsofpcs.com/paste/ffmpeg_demux_attempt004.txt
[17:06] <Compn> burek : just use mplayer -dumpvideo :P
[17:06] <burek> it's a data stream
[17:06] <burek> unknown to ffmpeg
[17:06] <Compn> oh
[17:06] <Daemon404> image2
[17:07] <Compn> Daemon404 : it depends , ffmpeg-mt was merged as one big blob, iirc
[17:07] <Daemon404> Compn, with help from its author, no?
[17:07] <Compn> a series of patches kind of makes no sense, especially if its huge and requires more code
[17:07] <Compn> from astrange ? 
[17:07] <Compn> iirc michael merged it during the ... bad times
[17:08] <Daemon404> HEH
[17:08] <Daemon404> heh*
[17:08] <Compn> astrange got it closer to being intergrated anyways
[17:09] <Compn> Daemon404 : http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2011-March/109620.html
[17:09] <Compn> :)
[17:10] <Compn> but i'm not git admin, so i dont really have opinion on doing a full history merge vs a blob commit...
[17:11] <Compn> i just make observations based on previous actions
[17:25] <CIA-119> ffmpeg: 03Matthieu Bouron 07master * r320e537baf 10ffmpeg/libavformat/mxfenc.c: 
[17:25] <CIA-119> ffmpeg: mxfenc: support 50 and 60 frame rates
[17:25] <CIA-119> ffmpeg: Reviewed-by: Tomas Härdin <tomas.hardin at codemill.se>
[17:25] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:25] <CIA-119> ffmpeg: 03Piotr Bandurski 07master * re03ddbcd91 10ffmpeg/libavcodec/gif.c: 
[17:25] <CIA-119> ffmpeg: gifenc: support resolutions up to 65535x65535
[17:25] <CIA-119> ffmpeg: Maybe someone can add a check in the second gif encoder (rgb24), because I'm not sure where it should be added.
[17:25] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:25] <CIA-119> ffmpeg: 03Piotr Bandurski 07master * rd2581dcca6 10ffmpeg/libavcodec/mjpegenc.c: 
[17:25] <CIA-119> ffmpeg: mjpegenc: add a limit for maximum supported resolution
[17:25] <CIA-119> ffmpeg: jpeg does not allow more than 65000 pixels for width and height
[17:25] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:39] <CIA-119> ffmpeg: 03Nicolas George 07master * r6d679c4359 10ffmpeg/libavfilter/bufferqueue.h: lavfi: add a generic API for buffer queues.
[17:40] <CIA-119> ffmpeg: 03Nicolas George 07master * r72f3786d0f 10ffmpeg/libavfilter/sink_buffer.c: 
[17:40] <CIA-119> ffmpeg: sink_buffer: guard against NULL frames.
[17:40] <CIA-119> ffmpeg: This only happens if some filter somewhere is bogus,
[17:40] <CIA-119> ffmpeg: but it helps debugging.
[17:40] <CIA-119> ffmpeg: 03Nicolas George 07master * r06bf6d3bc0 10ffmpeg/libavfilter/vf_overlay.c: (log message trimmed)
[17:40] <CIA-119> ffmpeg: vf_overlay: rewrite request/push logic.
[17:40] <CIA-119> ffmpeg: The old implementation, upon receiving a frame on the main
[17:40] <CIA-119> ffmpeg: input, would request an overlay frame if necessary. This could
[17:40] <CIA-119> ffmpeg: generate an infinite recursion, off-by-one frame mismatch and
[17:40] <CIA-119> ffmpeg: other harmful effects, especially it the split filter is present
[17:40] <CIA-119> ffmpeg: upgraph.
[17:44] <ubitux> TimNich: would it make sense to have one timecode track per video stream?
[19:59] <Daemon404> ubitux, is there any way for -show_streams to output if a stream is interlaced or not
[19:59] <Daemon404> or do i always need to use -show-frames
[20:01] Action: Daemon404 supposes hell fallback on mediainfo for now
[20:09] <ubitux> no idea :)
[20:26] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rb02cc2ddc6 10ffmpeg/libavcodec/mpeg4videodec.c: 
[20:26] <CIA-119> ffmpeg: mpeg4videodec: Check that cplx_estimation_* fits in the available space
[20:26] <CIA-119> ffmpeg: Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
[20:26] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:26] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r18b46a494e 10ffmpeg/libavformat/oggparsevorbis.c: 
[20:26] <CIA-119> ffmpeg: oggparsevorbis: fix null ptr dereference
[20:26] <CIA-119> ffmpeg: Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
[20:26] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:26] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rb7d1488393 10ffmpeg/libavcodec/ (h264.c h264.h h264data.h): (log message trimmed)
[20:26] <CIA-119> ffmpeg: h264: move q0 scan tables into context
[20:27] <CIA-119> ffmpeg: This fixes out of global array reads.
[20:27] <CIA-119> ffmpeg: The alternative solutions of checking the index or modifying the VLC tables
[20:27] <CIA-119> ffmpeg: to prevent the index going outside are each about 1-2 cpu cyclces slower
[20:27] <CIA-119> ffmpeg: per coded 4x4 block.
[20:27] <CIA-119> ffmpeg: The alternative of padding the global tables directly is more ugly and
[21:42] <CIA-119> ffmpeg: 03Ronald S. Bultje 07master * r64bde80563 10ffmpeg/libavformat/ (ac3dec.c mp3dec.c): mp3/ac3 probe: search for PES headers to prevent probing MPEG-PS as MP3.
[21:42] <CIA-119> ffmpeg: 03Ronald S. Bultje 07master * r6163d880c0 10ffmpeg/libavcodec/vp8.c: 
[21:42] <CIA-119> ffmpeg: vp8: move block coeff arithcoder on stack.
[21:42] <CIA-119> ffmpeg: This prevents gcc from assuming that contents of it may have changed
[21:42] <CIA-119> ffmpeg: between calls to vp56_range_get_prob(), thus preventing countless (and
[21:42] <CIA-119> ffmpeg: unnecessary) movs. Decoding of sintel trailer goes from (avg+SG) 9.796
[21:42] <CIA-119> ffmpeg: +/- 0.003 to 9.635 +/- 0.010.
[21:42] <CIA-119> ffmpeg: 03Justin Ruggles 07master * rc7bd556d48 10ffmpeg/libavfilter/af_amix.c: 
[21:42] <CIA-119> ffmpeg: lavfi: amix: check active input count before calling request_samples
[21:42] <CIA-119> ffmpeg: fixes use of the amix filter with only 1 input
[21:42] <CIA-119> ffmpeg: 03Samuel Pitoiset 07master * r1876e7c0c2 10ffmpeg/libavformat/http.c: 
[21:42] <CIA-119> ffmpeg: http: Add 'post_data', a new option which sets custom HTTP post data
[21:42] <CIA-119> ffmpeg: This allows doing http posts with a content-length header sent
[21:42] <CIA-119> ffmpeg: in advance, avoiding chunked encoding.
[21:42] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r5afdb3e6b1 10ffmpeg/: (log message trimmed)
[21:42] <CIA-119> ffmpeg: Merge remote-tracking branch 'qatar/master'
[21:42] <CIA-119> ffmpeg: * qatar/master:
[21:42] <CIA-119> ffmpeg:  fate: Fix fate-ac3-fixed-encode for pre-ssse3 x86 machines
[21:42] <CIA-119> ffmpeg:  http: Pass the proper return code of net IO operations
[21:42] <CIA-119> ffmpeg:  http: Add 'post_data', a new option which sets custom HTTP post data
[21:43] <CIA-119> ffmpeg:  lavfi: amix: check active input count before calling request_samples
[21:46] <ubitux> michaelni: https://github.com/ubitux/FFmpeg/commit/0b008d6ba61bbe34805b277443e5fa474410a831
[21:46] <ubitux> i didn't test it much
[21:46] <ubitux> but it should be better
[21:49] <michaelni> ubitux, i suspect it will need a check for ' ' and tab too otherwise LGTM & thanks
[21:49] Action: michaelni kicks his ISP once again ...
[21:49] <ubitux> jss_whitespace() should get rid of the spaces
[21:50] <Daemon404> this reminds me
[21:50] <michaelni> ahh right, i missed that
[21:50] Action: Daemon404 was contemplating rewriting patcheck
[21:50] <Daemon404> too many false positives and missed thigns
[21:54] <michaelni> Daemon404, you can fix it if its buggy
[21:58] <Daemon404> michaelni, im actually trying to locate the one i used to use
[21:58] <Daemon404> for kernel patches
[21:58] <Daemon404> it was a lot more hardcore
[22:00] <pasteeater> he kicked too hard
[22:09] <ubitux> michaelni: ok to return AVPROBE_SCORE_MAX/2 + 1 btw?
[22:09] <ubitux> otherwise it triggers an equal score with ansi or sth like this
[22:09] <ubitux> (currently AVPROBE_SCORE_MAX/2)
[22:10] <ubitux> mmh sth looks wrong.
[22:11] <ubitux> oh get it.
[22:11] <ubitux> it seems the ffprobe tests are lacking a dependency to the samples dir
[22:14] <ubitux> anyway, i'll go for the +1
[22:16] <CIA-119> ffmpeg: 03Clément BSsch 07master * r303619d3ca 10ffmpeg/libavformat/jacosubdec.c: jacosub_probe: slightly increase the score to limit misdetection.
[22:16] <CIA-119> ffmpeg: 03Clément BSsch 07master * r908293d1bc 10ffmpeg/libavformat/jacosubdec.c: jacosub_probe: speedup by making only one call to sscanf
[00:00] --- Fri Jun  1 2012