[Ffmpeg-devel-irc] ffmpeg-devel.log.20111217

Sun Dec 18 02:05:03 CET 2011

[00:00] <cbsrobot> it the mt implementation for decoding the same as for encoding ?
[00:00] <cbsrobot> slice de- and encoding use execute and execute2
[00:01] <Cosmin> michaelni: http://www.google-melange.com/gci/task/view/google/gci2011/7209212  use the NEW files now. it should be a sig9 memory exceeded
[00:01] <michaelni> for slices encoding should be pretty much the same as decoding
[00:01] <cbsrobot> and for frame de- and encoding do I need to use the .init_thread_copy in the context ?
[00:02] <cbsrobot> *frame encoding
[00:02] <cbsrobot> there a only a few codecs mt for encoding
[00:03] <cbsrobot> dnxhd, libx264 and 2 others
[00:03] <michaelni> Cosmin, i think you are doing something wrong, make sure you use the correct (latest)  ffmpeg version in zzuf
[00:03] <cbsrobot> but I did not find a nice frame threaded encoder 
[00:04] <michaelni> cbsrobot, i dont think theres frame MT support yet
[00:04] <michaelni> except some unfinished in a branch of ffmpeg-mt
[00:04] <cbsrobot> michaelni: ah ok
[00:05] <cbsrobot> thats why
[00:24] <cbsrobot> michaelni: seems it was one of astranges last commits: http://gitorious.org/~astrange/ffmpeg/ffmpeg-mt/commit/c34e0871c1f287b54be0aec7442316b604b94907
[00:25] <cbsrobot> or maybe there is another repo somewhere
[00:25] <michaelni> if you find a new repo i surely want to know of it too ...
[00:33] <Dragooon> How different ffmpeg is from libav?
[00:34] <bcoudurier> lol
[00:34] <bcoudurier> Dragooon, ffmpeg is the original
[00:36] <durandal_1707> Libav is fork of FFmpeg, for how: you can use diff(1)
[00:36] <Dragooon> Any GCI mentor here?
[00:37] <Compn> Dragooon : michaelni is 
[00:37] <Compn> Dragooon : ffmpeg merges most of libav, so they arent very different
[00:37] <Dragooon> Can I report and solve the same bug?
[00:38] <durandal_1707> isn't that cheating?
[00:39] <Dragooon> How?
[00:39] <durandal_1707> if bug has already been fixed in Libav
[00:40] <durandal_1707> or reported in Libav ml
[00:40] <Dragooon> Ah yeah
[00:40] <Dragooon> I mean fuzzing and finding my own bug, then fixing it
[00:41] <durandal_1707> that is fine
[00:42] <Dragooon> Although I do have a few unclaimed ones from libav, I guess I'll let them go
[00:43] <Compn> to answer your question, you cant get points for reporting the same bug to both projects
[00:43] <Dragooon> I'm not going to do that
[00:43] <Compn> hehe
[00:43] <Compn> if theres a bug in ffmpeg but not libav, i think thats ok to report
[00:43] <Compn> as long as its not a bug that was just fixed in the last week or anything :P
[00:43] <Compn> but i dunno whats best
[00:44] Action: Compn goes afk
[00:44] <Dragooon> I'm not sure if I will be fixing stuff or not
[00:46] <Dragooon> How about a bug reported in libav but not fixed?
[00:47] <Compn> just reporting it to ffmpeg ? or fixing it ?
[00:47] <Dragooon> Fixing it
[00:47] <Dragooon> And not claiming for the report
[00:47] <pasteeater> Dragooon: the task says, "Its not ok to take it from the old ffmpeg or the forks bug tracker"
[00:48] <durandal_1707> Dragooon: you can still fix it for free
[00:49] <Dragooon> Heh, okay. Thanks
[00:49] <Dragooon> I'll look into ffmpeg in a while
[00:54] <michaelni> Dragooon, reporting something that hasnt been reportet is ok, fixing something that hasnt been fixed is too as a different task
[00:54] <michaelni> so if noone fixed it in either project and noone submited a patch anywhere that fixes it
[00:54] <Dragooon> Oh okay
[00:56] Action: Dragooon loves cheese
[00:59] <michaelni> Dragooon, which cheese ? 
[00:59] <Dragooon> All of them, mostly Mozzarella 
[00:59] <Daemon404> feta~
[01:02] <michaelni> i like roquefort, jerome, appenzeller, ...
[01:03] <ubitux> do we have a task for a favicon?
[01:03] <ubitux> oh we already have one
[01:03] <ubitux> just not integrated in the doxy pages etc
[01:07] <durandal_1707> what about making general.html prettier
[01:44] <durandal_1707> Daemon404: it appears FFmpeg does not crash
[01:45] <Daemon404> i see
[01:45] <Daemon404> libav does
[01:45] <Daemon404> can you reproduce teh bug with that?
[01:46] <Daemon404> using my sample
[01:46] <durandal_1707> I would need to compile Libav first
[01:46] <Daemon404> true
[01:46] <Daemon404> let me compile ffmpeg
[01:46] <Daemon404> i have both repos on my hdd
[01:47] <durandal_1707> but anyway if I try to play file with ffplay I just get bunch of red lines
[01:48] <Daemon404> the file doesnt even have video in t
[01:48] <Daemon404> >_>
[01:48] <Daemon404> it's just ac3 audio
[01:48] <durandal_1707> sample.nut is fine but created .ts is not
[01:48] <Daemon404> ic
[01:48] <durandal_1707> so it is muxer bug perhaps
[01:49] <Daemon404> you saw my bt right?
[01:49] <Daemon404> crashes inside avconv.c and ffmpeg.c
[01:49] <durandal_1707> with ffmpeg too?
[01:49] <Daemon404> ffmpeg is building
[01:49] <Daemon404> youll have an answer in a sec
[01:50] <durandal_1707> bt may be misleading ...
[01:50] <Daemon404> i know
[01:50] <Daemon404> i didnt really have much drive to go through those 2 files though and trace it
[01:50] <Daemon404> cause... yeah well their code is something else.
[01:51] <Daemon404> durandal_1707, indeed ffmpeg is fine
[01:52] <Daemon404> only libav crashes
[01:53] <Daemon404> updated by report to reflect that
[01:54] <durandal_1707> muxing into avi seems fine
[01:55] <durandal_1707> so it must be ts muxer issue
[01:55] <Daemon404> yea
[02:00] <durandal_1707> does it make sense to add pcm16 support to mpegts?
[02:00] <Daemon404> so you dont get audio btw?
[02:01] <Daemon404> when you play it
[02:01] <Daemon404> and yes pcm support is sort of really important in ts
[02:01] <Daemon404> for example, blurays
[02:01] <Daemon404> (mind you tehy use 192-byte packets, and they use pcm_bluray)
[02:01] <durandal_1707> ffplay thinks audio is aac and produce crap
[02:02] <durandal_1707> crap = red lines and aborts
[02:02] <Daemon404> lulz
[02:02] <Daemon404> thats kind of funny.
[02:02] <Daemon404> oh
[02:02] <Daemon404> found the fix
[02:02] <Daemon404> michaelni pushed a fix for this exact issue
[02:02] <durandal_1707> [aac @ 0x29bdac00] More than one AAC RDB per ADTS frame is not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented.
[02:03] <Daemon404> lol.
[02:03] <Daemon404> http://git.videolan.org/?p=ffmpeg.git;a=commit;h=e31c5ebe1146d98d17a5121312c5444432c81904
[02:03] <Daemon404> ^ this stops teh crash
[02:05] <Daemon404> durandal_1707, on windows, using haali's media splitter
[02:05] <Daemon404> it is detected as pcm
[02:05] <Daemon404> and demuxes and plays fine
[02:05] <Daemon404> the file created is OK
[02:08] <Daemon404> lets try some other tools
[02:08] <durandal_1707> so pcm16 does not need its own stream id?
[02:08] <Daemon404> hmm
[02:08] <Daemon404> only haali recognizes it
[02:09] <Daemon404> eac3to says it isnt valid
[02:09] <Daemon404> tsmuxer said it isnt valid
[02:09] <Daemon404> mkvmerge sees it as ac3
[02:10] <durandal_1707> well it probably is invalid lets improve muxer
[02:10] <Daemon404> yea
[02:11] <Daemon404> too bad i dont have the specs
[02:11] <Daemon404> so i cant see what pcm's id is
[02:11] <Daemon404> but looking at teh file it creates
[02:11] <durandal_1707> but do you have tool which succesfully mux pcm into ts?
[02:11] <Daemon404> it's muxing it as private data
[02:11] <Daemon404> hmm
[02:11] <Daemon404> lets try~
[02:13] <Daemon404> durandal_1707, it is detected as pcm_bluray when muxed with tsmuxer
[02:13] <Daemon404> pcm_bluray is pcm with a 4-byte header
[02:13] <Daemon404> so tsmuxer may add that
[02:16] <kierank> in blu-ray there is some header for pcm iirc
[02:16] <Daemon404> yes a 4-byte one
[02:17] <kierank> cba to find the spec
[02:21] <Tjoppen> cbsrobot: haven't checked ml or git yet, but did the patches solve whatever problem you had?
[02:21] <durandal_1707> http://forum.videolan.org/viewtopic.php?f=7&t=15928
[02:22] <durandal_1707> Daemon404: found tool which mux pcm into ts?
[02:22] <Daemon404> durandal_1707, they all convert ot to pcm_bluray (add teh 4-byte header)
[02:22] <Daemon404> so im thinking it may be sane to just add pcm_bluray support.
[02:23] <CIA-49> ffmpeg: 03Diego Biurrun 07master * r54e75be420 10ffmpeg/configure: configure: refactor lists of tests and components into variables
[02:23] <CIA-49> ffmpeg: 03Alexandra Khirnova 07master * ra7b5e841ff 10ffmpeg/ (avconv.c doc/avconv.texi doc/avtools-common-opts.texi): 
[02:23] <CIA-49> ffmpeg: avconv: support stream specifiers in -metadata and -map_metadata
[02:23] <CIA-49> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[02:23] <CIA-49> ffmpeg: 03Derek Buitenhuis 07master * rbb9747c8ee 10ffmpeg/libavcodec/wavpack.c: 
[02:23] <CIA-49> ffmpeg: wavpack: Fix 32-bit clipping
[02:23] <CIA-49> ffmpeg: In the case that (frame_flags & 0x03) == 3, hybrid_maxclip
[02:23] <CIA-49> ffmpeg: may have had a signed integer overflow.
[02:23] <CIA-49> ffmpeg: Signed-off-by: Derek Buitenhuis <derek.buitenhuis at gmail.com>
[02:23] <CIA-49> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[02:23] <CIA-49> ffmpeg: 03Mans Rullgard 07master * ra1e98f198e 10ffmpeg/libavcodec/ (dv.c get_bits.h imc.c proresdec.c): (log message trimmed)
[02:23] <CIA-49> ffmpeg: get_bits: remove A32 variant
[02:23] <CIA-49> ffmpeg: The A32 bitstream reader variant is only used on ARMv5 and for
[02:23] <CIA-49> ffmpeg: Prores due to the larger bit cache this decoder requires.
[02:23] <CIA-49> ffmpeg: In benchmarks on ARMv5 (Marvell Sheeva) with gcc 4.6, the only
[02:23] <CIA-49> ffmpeg: statistically significant difference between ALT and A32 is
[02:23] <CIA-49> ffmpeg: a 4% advantage for ALT in FLAC decoding. There is thus no (longer)
[02:23] <CIA-49> ffmpeg: 03Mans Rullgard 07master * rb8909cb364 10ffmpeg/libavcodec/get_bits.h: 
[02:23] <CIA-49> ffmpeg: get_bits: remove useless pointer casts
[02:24] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r552ec4c9fd 10ffmpeg/: (log message trimmed)
[02:24] <CIA-49> ffmpeg: Merge remote-tracking branch 'qatar/master'
[02:24] <CIA-49> ffmpeg: * qatar/master:
[02:24] <CIA-49> ffmpeg:  get_bits: remove A32 variant
[02:24] <CIA-49> ffmpeg:  avconv: support stream specifiers in -metadata and -map_metadata
[02:24] <CIA-49> ffmpeg:  wavpack: Fix 32-bit clipping
[02:24] <CIA-49> ffmpeg:  wavpack: Clip samples after shifting
[02:24] <CIA-49> ffmpeg: 03Mans Rullgard 07master * r11b1db2759 10ffmpeg/libavcodec/arm/ (rv40dsp_init_neon.c rv40dsp_neon.S): 
[02:24] <CIA-49> ffmpeg: rv40: NEON optimised weak loop filter
[02:24] <CIA-49> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[02:24] <CIA-49> ffmpeg: 03Derek Buitenhuis 07master * r365e1ec26d 10ffmpeg/libavcodec/wavpack.c: 
[02:24] <CIA-49> (11 lines omitted)
[02:30] <Daemon404> 'cept we dont have an encoder for that lol
[02:31] <kierank> no point adding pcm_bluray
[02:31] <Daemon404> why not?
[02:31] <kierank> because blu-ray muxing will almost certainly never be spec compliant
[02:31] <Daemon404> true
[02:32] <Daemon404> ill just forward michealni's large frame fix to libav
[02:32] <Daemon404> and hope i dont get flamed
[02:32] <Daemon404> it's still relevant to otehr stuff
[02:32] <Daemon404> and bye :V
[02:47] <Daemon404> michaelni, are you here quickly
[02:49] <michaelni> Daemon404, sure, how can i helpo
[02:50] <Daemon404> michaelni, can you check my Yet Another WavPack fix on irc quickly
[02:50] <Daemon404> it's 2 lines
[02:50] <michaelni> i can try
[02:50] <michaelni> but iam kinda tired today so dunno, i might miss something
[02:51] <Daemon404> this should be a pretty obvious fix
[02:51] <Daemon404> just want a sanity check before i send it to the ML
[02:51] <michaelni> sure, show the lines ...
[02:52] <Daemon404> http://pastebin.com/nSWhUYYG
[02:53] <michaelni> Daemon404, looks buggy
[02:54] <Daemon404> hmm
[02:54] <Daemon404> how so?
[02:54] <michaelni> i think
[02:54] <michaelni> i think it should clip to 0x7FFFFFFF but here it clips to maybe 0x7FFFFFFC
[02:55] <Daemon404> oh
[02:55] <Daemon404> i see what you mean
[02:55] <Daemon404> michaelni, ill make a 32-bit test .wv that needs clipping and work on it tomorrow
[02:56] <Daemon404> and verify the results against upstream
[02:56] <michaelni> yes, thats a good idea
[02:57] <Daemon404> anyway, ive got to run out. ill look at it tomorrow.
[02:57] <Daemon404> sleep well, michaelni.
[02:57] <michaelni> i still have work to do :(
[02:57] <Daemon404> well... sleep-code well, michaelni 
[02:57] Action: Daemon404 goes out to eat dinner
[03:36] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r9af6abdc17 10ffmpeg/libavcodec/atrac3.c: 
[03:36] <CIA-49> ffmpeg: atrac3: Fix crash in tonal component decoding.
[03:36] <CIA-49> ffmpeg: Fixes Ticket780
[03:36] <CIA-49> ffmpeg: Bug Found by: cosminamironesei
[03:36] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:36] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r53b6503184 10ffmpeg/ (5 files in 3 dirs): 
[03:36] <CIA-49> ffmpeg: avienc: Fix rawvideo 32bit
[03:36] <CIA-49> ffmpeg: Bug found by: Rik Maes <rik.maes54 at gmail.com>
[03:36] <CIA-49> ffmpeg: Change based on suggestion by Rik Maes
[03:36] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:16] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r4d70023a2a 10ffmpeg/libavcodec/h264_mvpred.h: 
[05:16] <CIA-49> ffmpeg: h264: fix init of topleft ref/mv.
[05:16] <CIA-49> ffmpeg: Fixes Ticket778
[05:16] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:16] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * rf37b2d5a68 10ffmpeg/libavcodec/mpeg4audio.c: 
[05:16] <CIA-49> ffmpeg: aacdec: Fix null pointer crash
[05:16] <CIA-49> ffmpeg: Fixes bug170
[05:16] <CIA-49> ffmpeg: Bug found by: Gautam Gupta
[05:16] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:45] <durandal_1707> ffmpeg small task are extremly hard task without samples
[06:12] <michaelni> durandal_1707, for what do you need a sample ?
[06:12] <michaelni> i meant what codec/format ...
[06:19] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r68230eb375 10ffmpeg/libavutil/utils.c: 
[06:19] <CIA-49> ffmpeg: pixfmts: Selfcheck against insertions / deletions in the enum.
[06:19] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:19] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r75f2cc5134 10ffmpeg/libavutil/utils.c: 
[06:19] <CIA-49> ffmpeg: avutil: check sampleformats enum, pict type enum and media type enum against insert/remove mistakes.
[06:19] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:19] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r83b46713f0 10ffmpeg/libavcodec/utils.c: 
[06:19] <CIA-49> ffmpeg: lavc: Check codec_ids against insertion/removial typos
[06:19] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[07:06] <durandal_1707> michaelni: xbox: xwav, xvas, xmu, wavm .. the one vgmstream already supports
[14:26] <Cosmin_> michaelni: where can i find files to test with ffmpeg, like the fate system which libav has?
[14:30] <Cosmin_> michaelni: ?
[14:31] <buzz_> read fate.txt in docs folder
[14:32] <buzz_> make fate-rsync SAMPLES=path/to/samples/
[14:33] <Cosmin_> but is there any source with other files?
[14:33] <Cosmin_> because i have the libav fate files
[14:33] <buzz_> mplayer has stuff
[14:34] <Cosmin_> how can i get them/
[14:36] <buzz_> try a google search
[14:37] <buzz_> http://samples.mplayerhq.hu/ - but i don't see why I have to the google search for you.
[14:39] <ohsix> then why did you do it? :[
[14:40] <buzz_> no idea.
[16:15] <michaelni> anyone alive?
[16:16] <michaelni> iam searching for the svg of our logo
[16:16] <michaelni> there was a svg i think ...
[16:16] <michaelni> or some other source of higher resolution
[16:24] <ubitux> https://en.wikipedia.org/wiki/File:Ffmpeg_logo.svg this one maybe?
[16:25] <ubitux> but that's the old one.
[16:32] <michaelni> ubitux, i need the new
[16:32] <michaelni> its for GCI
[16:32] <michaelni> http://www.google-melange.com/gci/task/view/google/gci2011/7197471
[16:50] <CIA-49> ffmpeg: 03Clément BSsch 07master * rc0bea18411 10ffmpeg/doc/developer.texi: doc: fix "reference to nonexistent node `Regression Tests'" warning.
[16:50] <CIA-49> ffmpeg: 03Clément BSsch 07master * r6aeba3fd6b 10ffmpeg/doc/fate.texi: doc: fix "Unknown command `fate'" warning.
[18:54] <ubitux> how is *{buffer,null}sink supposed to be used?
[18:54] <ubitux> s/is/are/
[18:56] <ubitux> ffplay -f lavfi nullsrc is ok, but nullsrc,nullsink or any other src just fails
[18:57] <ubitux> i guess the usage is a bit special given the "programmatically usage"
[20:27] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r56bf2c2a19 10ffmpeg/libavcodec/motion_est.c: 
[20:27] <CIA-49> ffmpeg: motion_est: avoid generating motion vectors that point between widthxheight and mb_widthxmb_height.
[20:27] <CIA-49> ffmpeg: No difference in PSNR or bitrate in the printed precission with the matrix lobby scene at 322x242
[20:27] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:46] <CIA-49> ffmpeg: 03Nicolas George 07master * r9cb6a39c46 10ffmpeg/libavformat/sbgdec.c: 
[22:46] <CIA-49> ffmpeg: sbgdec: replace EOVERFLOW with ENOMEM.
[22:46] <CIA-49> ffmpeg: This should fix track issue #781, regarding compilation with
[22:46] <CIA-49> ffmpeg: i686-w64-mingw32-gcc.
[23:23] <redbal> has anyone gotten ffmpeg working on oracle solaris 11?
[23:27] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * re09ffa4524 10ffmpeg/libavcodec/msvideo1enc.c: 
[23:27] <CIA-49> ffmpeg: msvideo1enc: Check that dimensions are a multiple of 4.
[23:27] <CIA-49> ffmpeg: Fixes Ticket784
[23:27] <CIA-49> ffmpeg: Found by: ami_stuff
[23:27] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:29] <michaelni> redbal, whats the problem ?
[23:31] <redbal> I'm "playing around" with solaris 11 on a lenovo laptop and I can't install ffmpeg, there's no pre-compiled version from oracle and there doesn't seem to be any effort on this os thus far
[23:33] <michaelni> there are 2 solaris-something boxes that regularly test ffmpeg 
[23:34] <michaelni> just install git, clone the latest ffmpeg and try to build it
[23:34] <michaelni> if it fails ask in #ffmpeg
[23:35] <redbal> rg, have posted, but no response  I have downloaded and attempted to install FFmpeg-552ec4c
[23:37] <michaelni> well what goes wrong when you try to compile & install ?
[23:37] <redbal> awk: syntax error near line 1, illegal statement near line 1    yasm not found
[23:38] <redbal> so, I disable-yasm and retried
[23:40] <redbal> that results in a make error make: Fatal error in reader: config.mak, line 3: Unexpected end of line seen
[23:40] <michaelni> gmake
[23:41] <redbal> that produced a long string of errors...last one ./libavutil/mathematics.h:79: confused by earlier errors, bailing out
[23:45] <redbal> i'm downloading asm and associated system headers and will retry
[23:46] <michaelni> make also sure it uses gcc and not some sun compiler
[23:46] <redbal> gcc version 3.4.3 (csl-sol210-3_4-20050802)
[23:47] <michaelni> that should work
[23:47] <michaelni> gmake --version
[23:48] <redbal> GNU Make 3.81
[23:48] <redbal> This program built for i386-pc-solaris2.11
[23:49] <michaelni> ok
[23:49] <ubitux> did anyone contact baptiste to fix fate?
[23:49] <michaelni> maybe you need bash/gawk but iam a bit guessing here, ffmpeg just needs posix tools 
[23:51] <redbal> I have both of those...asm is no longer available for this version of solaris, "obsolete"
[23:54] <michaelni> having them is good, but make sure configure also uses them and not some non posix awk/sh
[23:56] <redbal> after adding the core c and c++ system headers, I get a different set of errors: most of them are compile warning errors about static declarations followed by non-static declaration.
[23:57] <redbal> the first "error" occurs in libavdevice/alldevices.c in function `avdevice_register_all';
[23:57] <redbal> libavdevice/alldevices.c:41: error: `CONFIG_ALSA_OUTDEV' undeclared (first use in this function)
[23:58] <michaelni> awk --version 
[23:59] <michaelni> if it doesnt say GNU awk something then this might be the problem
[00:00] --- Sun Dec 18 2011