[Ffmpeg-devel-irc] ffmpeg-devel.log.20120428

[burek]

[10:07] <CIA-17> ffmpeg: 03Nicolas George 07master * r7ad0d9864e 10ffmpeg/libavfilter/ (Makefile asrc_abuffer.c vsrc_buffer.c): lavfi: merge asrc_abuffer into vsrc_buffer.
[10:07] <CIA-17> ffmpeg: 03Nicolas George 07master * r2ae5af8094 10ffmpeg/libavfilter/ (Makefile vsrc_buffer.c): lavfi: rename vsrc_buffer.c into src_buffer.c.
[10:47] <CIA-17> ffmpeg: 03Nicolas George 07master * r8f19483d06 10ffmpeg/doc/examples/filtering_audio.c: examples/filtering_audio: fix a memory leak.
[10:48] <CIA-17> ffmpeg: 03Nicolas George 07master * ra0a0199da3 10ffmpeg/doc/examples/filtering_audio.c: 
[10:48] <CIA-17> ffmpeg: examples/filtering_audio: do not stop on decode error.
[10:48] <CIA-17> ffmpeg: Decode errors can happen with concatenated MP3s
[10:48] <CIA-17> ffmpeg: with different formats.
[12:47] <aquarat> hi, does anyone know if ffmpeg supports MPEG multichannel audio ?
[12:47] <aquarat> http://en.wikipedia.org/wiki/MPEG_Multichannel
[12:47] <aquarat> it's a bit of a weird format
[12:48] <j-b> did you tried it?
[12:48] <Compn> did you have sample ?
[12:48] <aquarat> if I hand ffmpeg an mpeg multichannel file it reads it as having 2 chans
[12:48] <aquarat> yes I do
[12:48] <aquarat> they're fairly large though
[12:48] <aquarat> erm
[12:48] <aquarat> I have a sample of code
[12:48] <aquarat> that can decode it
[12:48] <aquarat> a project called mctoolamed
[12:49] <aquarat> can decode it with some mods
[12:49] <Compn> ok
[12:49] <aquarat> btw
[12:49] <Compn> i dont remember if ffmpeg has support for it
[12:49] <Compn> sounds rare :)
[12:49] <aquarat> thatnk you so much for prores support
[12:49] <Compn> have to check mpegaudiodec.c or whatnot
[12:49] <aquarat> it is rare
[12:49] <aquarat> it was originally supposed to be part of the DVD spec
[12:50] <aquarat> but was dropped in 1997
[12:50] <Compn> they went with ac3 multichannel :P
[12:50] <aquarat> Canon XLH1 cameras generate those files in 4 channel mode
[12:50] <aquarat> it's part of the HDV spec
[12:50] <aquarat> but very few cams support it
[12:50] <aquarat> lol
[12:50] <aquarat> yeah :)
[12:50] <aquarat> mctoolame.sourceforge.net
[12:50] <Compn> but uh, if you upload a sample , and a link to the code that works, someone may add it :)
[12:50] <aquarat> kew
[12:51] <j-b> does libmad decode it?
[12:51] <aquarat> my C++ skills are useless :/
[12:51] <aquarat> haven't tried
[12:51] <Compn> j-b works on videolan , so he is also interested in a sample and such :)
[12:51] <aquarat> can I cut up the sample using dd?
[12:51] <j-b> http://mctoolame.sourceforge.net/test.aiff-mpeg1-192.mp2 ?
[12:51] <Compn> yep
[12:51] <aquarat> each sample is about an hour long
[12:51] <aquarat> kew
[12:51] <aquarat> I used ffmpeg to demux the mp2 from the hdv files
[12:52] <Compn> ftp://upload.ffmpeg.org/incoming
[12:52] <aquarat> that'll probably work @ j-b
[12:52] <j-b> so, this is easy
[12:52] <j-b> a sample is alredy there
[12:52] <j-b> and ffplay, VLC and Mplayer says it is stereo
[12:52] <aquarat> from 1996
[12:52] <j-b> roflillol
[12:52] <aquarat> yes
[12:52] <aquarat> but it's not :P
[12:52] <aquarat> hdv cams generate 4 channels
[12:52] <aquarat> mctoolamed dies if you hand it a four channel file
[12:52] <Compn> oh so one of you tested if that sample is multichannel  ? :)
[12:53] <Compn> making sure its not just a 2ch sample
[12:53] <j-b> Compn: lol
[12:53] <aquarat> I haven't recently tested
[12:55] Action: aquarat compiles mctoolamed
[12:57] <j-b> mediainfo says stereo
[12:57] <j-b> but it could be worong too
[12:57] Action: aquarat changes arch to "core2"
[12:59] <Compn> how can you guys just pick a sample and think its correct? :P
[13:00] <aquarat> that mp2 file has 5 channels in it
[13:00] <Compn> yay
[13:00] <aquarat> I just tested
[13:00] <Compn> ok
[13:00] <Compn> if you want to make a report, you can file it here > http://ffmpeg.org/trac/ffmpeg/
[13:00] <aquarat> the code only make provision for 2,6 or 8 channels
[13:00] <Compn> or i can do it if you dont feel like registering
[13:00] <aquarat> I had to mod it for 4 chan
[13:01] <Compn> thanks for testing sample
[13:01] <aquarat> you're probably better equipped to submit a report?
[13:01] <aquarat> I can try
[13:02] <CIA-17> ffmpeg: 03Reimar Döffinger 07master * r9d4a1aca88 10ffmpeg/libavresample/x86/audio_mix.asm: 
[13:02] <CIA-17> ffmpeg: Fix compilation with YASM/NASM without AVX support.
[13:02] <CIA-17> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[13:03] Action: aquarat creates new ticket
[13:04] <Compn> dont forget to link or attach the code you hacked for 4 channel.
[13:10] <aquarat> wow
[13:10] <aquarat> it's really really nasty
[13:11] <aquarat> but okay
[13:11] <aquarat> mctoolame doesn't handle sample rates properly either
[13:11] <aquarat> it assume 44.1kHz
[13:11] <aquarat> HDV cameras shoot 48kHz
[13:12] <aquarat> https://ffmpeg.org/trac/ffmpeg/ticket/1258
[13:12] <aquarat> ^^^ I hope I did that correctly
[13:15] <Compn> yeha, what about your hacks for the 4channel ? 
[13:15] <aquarat> I'm looking for it ;)
[13:15] <aquarat> I last used this code 2 years ago
[13:16] <Compn> oh :)
[13:17] <aquarat> I thought I'd be clever... someone gave me an HDV camera to shoot part of a documentary on
[13:17] <aquarat> and I switched it to 4 channel mode
[13:18] <j-b> but who used those mpeg multichannel codec?
[13:18] <aquarat> it's part of the HDV specification
[13:18] <aquarat> in four channel mode HDv cameras switch the audio stream from MPEG2 layer 1 to layer 2
[13:18] <Compn> j-b : <aquarat> Canon XLH1 cameras generate those files in 4 channel mode
[13:18] <aquarat> and encode the audio using this method
[13:18] <Compn> canon cameras i guess :P
[13:18] <aquarat> yes
[13:18] <j-b> morons
[13:18] <aquarat> the xl h1 in particular
[13:19] <aquarat> welll
[13:19] <aquarat> it's quite cool
[13:19] <j-b> riiight...
[13:19] <aquarat> if the decoding ability is available :P
[13:19] <aquarat> it is
[13:19] <Compn> is your camera from 1996 too ?
[13:19] <aquarat> you basically have a surround set of channels
[13:19] <j-b> there are not enough 5.1 codecs?
[13:19] <aquarat> er
[13:19] <Compn> and what documentary are you working on ? :)
[13:19] <aquarat> the xl h1 is a 2005 camera
[13:19] <aquarat> still being produced
[13:19] <Compn> ah
[13:19] <aquarat> newer cameras like the EX and EX3 use XDCAM
[13:19] <aquarat> which stores channels are pcm
[13:20] <aquarat> in seperate tracks
[13:20] <aquarat> which is, of course, lovely :)
[13:20] <aquarat> lol
[13:20] <aquarat> *as
[13:20] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * redbde52226 10ffmpeg/libswresample/ (6 files): 
[13:20] <CIA-17> ffmpeg: swr: change sample format representation so as to maintain the planer/packed distinction.
[13:20] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[13:20] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r06b62cb8f2 10ffmpeg/libswresample/ (audioconvert.c audioconvert.h): 
[13:20] <CIA-17> ffmpeg: swr: move AudioConvert to header
[13:20] <CIA-17> ffmpeg: the next commits will need it
[13:20] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[13:20] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * rbcc66ff0e4 10ffmpeg/libswresample/ (6 files in 2 dirs): 
[13:20] <CIA-17> ffmpeg: swr: add int16_to_int32_mmx/sse
[13:20] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[13:20] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * rfa5daaca0d 10ffmpeg/libswresample/x86/ (audio_convert.asm swresample_x86.c): 
[13:20] <CIA-17> ffmpeg: swr: seperate functions for aligned & unaligned
[13:20] <CIA-17> ffmpeg: If someone has an idea on how to do this cleaner, its welcome
[13:20] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[13:21] <aquarat> it's so weird that HDV wasn't designed with proper multichannel in mind
[13:21] <aquarat> they had the bitrate for it
[13:22] <Compn> thanks for reporting bug :)
[13:22] Action: Compn goes afk
[13:22] <aquarat> lol
[13:22] <aquarat> thanks for looking at it
[13:22] <aquarat> :)
[16:24] <Compn> aquarat : carl requested 4channel sample if you can make one :)
[16:46] <aquarat> I'll get it on it
[16:55] <CIA-17> ffmpeg: 03Nicolas George 07master * r12eeced8b4 10ffmpeg/libavformat/avio.c: 
[16:55] <CIA-17> ffmpeg: avio: change ffurl_alloc return code.
[16:55] <CIA-17> ffmpeg: If the designated protocol is not found, return
[16:55] <CIA-17> ffmpeg: AVERROR_PROTOCOL_NOT_FOUND instead of AVERROR(ENOENT).
[16:58] <burek> if anyone has a spare minute/two please take a look at this compile error: http://pastebin.com/MkraQvRL
[16:58] <burek> In file included from libavfilter/vf_deshake.c:56:0: ./libavcodec/dsputil.h:448:47: error: expected ‘;’, ‘,’ or ‘)’ before ‘v1’
[16:59] <burek> the line in question is this (dsputil.h): void (*butterflies_float)(float *restrict v1, float *restrict v2, int len);
[16:59] <burek> is it some kind of typo, or some missing #include or what?
[16:59] <burek> it's the latest ffmpeg git in question
[17:03] <aquarat> done
[17:19] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r113738d6c2 10ffmpeg/libswresample/x86/audio_convert.asm: 
[17:19] <CIA-17> ffmpeg: swr: more correct cglobal parameters to int16->int32
[17:19] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:19] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r832c3b10d2 10ffmpeg/libswresample/x86/ (audio_convert.asm swresample_x86.c): 
[17:19] <CIA-17> ffmpeg: swr: add int32_to_float_sse2
[17:19] <CIA-17> ffmpeg: could be done for sse/3dnow too if someone wants
[17:19] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:19] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r95057b1972 10ffmpeg/libswresample/x86/audio_convert.asm: 
[17:19] <CIA-17> ffmpeg: swr: int16->int32: use the old index negate trick to avoid 2 adds
[17:19] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:58] <CIA-17> ffmpeg: 03Nicolas George 07master * r75d5624cb2 10ffmpeg/libavfilter/avfilter.c: 
[18:58] <CIA-17> ffmpeg: avfilter: filter_samples: read pts before filtering.
[18:58] <CIA-17> ffmpeg: The call to the next filter_sample will likely unref the
[18:58] <CIA-17> ffmpeg: current buffer, so it is not possible to read it afterwards.
[19:12] <CIA-17> ffmpeg: 03Jean First 07master * r64943b47e7 10ffmpeg/libavformat/utils.c: 
[19:12] <CIA-17> ffmpeg: lavf remove duplicated check in has_duration
[19:12] <CIA-17> ffmpeg: this hunk was merged in 8b97ae64 and cbf767a8 although the check was there a
[19:12] <CIA-17> ffmpeg: few lines above since cdced09e. I removed the first check to reduce the differences
[19:12] <CIA-17> ffmpeg: to libav.
[19:12] <CIA-17> ffmpeg: Signed-off-by: Jean First <jeanfirst at gmail.com>
[19:12] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:12] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * rb72a0f9c23 10ffmpeg/libswresample/x86/ (audio_convert.asm swresample_x86.c): 
[19:12] <CIA-17> ffmpeg: swr: add int16_to_float_sse2()
[19:12] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:47] <CIA-17> ffmpeg: 03Nicolas George 07master * rf1b82099cd 10ffmpeg/libavfilter/src_buffer.c: src_buffer: add forgotten avfilter_unref_buffer.
[20:07] <CIA-17> ffmpeg: 03Roland Scheidegger 07master * r5520df6a8f 10ffmpeg/libavcodec/x86/cabac.h: 
[20:07] <CIA-17> ffmpeg: h264: (trivial) remove unneeded macro argument in x86/cabac.h
[20:07] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:07] <CIA-17> ffmpeg: 03Roland Scheidegger 07master * re52b96187b 10ffmpeg/libavcodec/cabac.c: 
[20:07] <CIA-17> ffmpeg: h264: (trivial) make ff_h264_lps_state static
[20:07] <CIA-17> ffmpeg: not used outside the cabac test functions (which probably means it's
[20:07] <CIA-17> ffmpeg: a bad test if it doesn't use the same tables as the real functions?)
[20:07] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:07] <CIA-17> ffmpeg: 03Roland Scheidegger 07master * r7f668cd2b5 10ffmpeg/libavcodec/ (6 files in 2 dirs): (log message trimmed)
[20:07] <CIA-17> ffmpeg: h264: use one table instead of several for cabac functions
[20:07] <CIA-17> ffmpeg: The reason is this is easier for PIC code (in particular on darwin...).
[20:07] <CIA-17> ffmpeg: Keep the old names as pointers (static in cabac_functions.h so gcc
[20:07] <CIA-17> ffmpeg: knows these are just immediate offsets) so the c code can nicely stay the same
[20:07] <CIA-17> ffmpeg: (alternatively could use offsets directly in the functions needing the
[20:30] <CIA-17> ffmpeg: 03Nicolas George 07master * r4b0521eca9 10ffmpeg/libavcodec/ (avcodec.h options.c utils.c): 
[20:30] <CIA-17> ffmpeg: lavc: add a channel_layout field to AVFrame.
[20:30] <CIA-17> ffmpeg: The field is filled with the codec context information.
[20:30] <CIA-17> ffmpeg: FIXME need a minor version bump.
[20:30] <CIA-17> ffmpeg: 03Nicolas George 07master * r128dda7016 10ffmpeg/libavcodec/ (avcodec.h options.c utils.c): 
[20:30] <CIA-17> ffmpeg: lavc: add a sample_rate field to AVFrame.
[20:30] <CIA-17> ffmpeg: The field is filled with the codec context information.
[20:43] <CIA-17> ffmpeg: 03Nicolas George 07master * rc31be45e14 10ffmpeg/libavcodec/version.h: lavc: minor bump for the new fields in AVFrame.
[21:10] <CIA-17> ffmpeg: 03Yusuke Nakamura 07master * r546adc1fee 10ffmpeg/libavformat/mov.c: 
[21:10] <CIA-17> ffmpeg: mov: Parse EC3SpecificBox (dec3 atom).
[21:10] <CIA-17> ffmpeg: Skip to parse fields for additional independent substreams and its
[21:10] <CIA-17> ffmpeg: associated dependent substreams since libavcodec's E-AC-3 decoder does not
[21:10] <CIA-17> ffmpeg: support them yet.
[21:10] <CIA-17> ffmpeg: Signed-off-by: Justin Ruggles <justin.ruggles at gmail.com>
[21:10] <CIA-17> ffmpeg: 03Luca Barbato 07master * re1e146a2d1 10ffmpeg/libavformat/aviobuf.c: 
[21:10] <CIA-17> ffmpeg: avio: make avio_close(NULL) a no-op
[21:10] <CIA-17> ffmpeg: Its behaviour in line with ffurl_close(NULL).
[21:10] <CIA-17> ffmpeg: 03Roland Scheidegger 07master * r14e9ffc1e4 10ffmpeg/libavcodec/ (6 files in 2 dirs): (log message trimmed)
[21:10] <CIA-17> ffmpeg: h264: use one table instead of several for cabac functions
[21:10] <CIA-17> ffmpeg: The reason is this is easier for PIC code (in particular on darwin...).
[21:10] <CIA-17> ffmpeg: Keep the old names as pointers (static in cabac_functions.h so gcc
[21:10] <CIA-17> ffmpeg: knows these are just immediate offsets) so the c code can nicely stay the same
[21:10] <CIA-17> ffmpeg: (alternatively could use offsets directly in the functions needing the
[21:11] <CIA-17> ffmpeg: h264: new assembly version of get_cabac for x86_64 with PIC
[21:11] <CIA-17> ffmpeg: This adds a hand-optimized assembly version for get_cabac much like the
[21:11] <CIA-17> ffmpeg: existing one, but it works if the table offsets are RIP-relative.
[21:11] <CIA-17> ffmpeg: Compared to the non-RIP-relative version this adds 2 lea instructions
[21:11] <CIA-17> ffmpeg: and it needs one extra register. get_cabac() gets about 40% faster, for
[21:11] <CIA-17> ffmpeg: an overall speedup of about 5%.
[21:11] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r9d7c1b4cf3 10ffmpeg/: (log message trimmed)
[21:11] <CIA-17> ffmpeg: Merge remote-tracking branch 'qatar/master'
[21:11] <CIA-17> ffmpeg: * qatar/master:
[21:11] <CIA-17> ffmpeg:  h264: new assembly version of get_cabac for x86_64 with PIC
[21:11] <CIA-17> ffmpeg:  h264: use one table instead of several for cabac functions
[21:12] <CIA-17> ffmpeg:  h264: (trivial) remove unneeded macro argument in x86/cabac.h
[21:12] <CIA-17> (8 lines omitted)
[00:00] --- Sun Apr 29 2012