[Ffmpeg-devel-irc] ffmpeg-devel.log.20120130

[burek]

[00:07] <michaelni> btw, if someone has an idea about the openbsd/sh segfault ...
[00:07] <michaelni> it seems working if bash is used ...
[00:07] <michaelni> but we should get it working by default
[00:08] <michaelni> the commit that "caused" it was the addition of the yuv4 test
[00:08] <michaelni> reverting that on HEAD doesnt fix it though
[00:13] <durandal_1707> probably need reverting v308 test too
[00:13] <durandal_1707> wich points that we can not add more tests
[00:14] <iive> is it the sh that segfaults?
[00:18] <Daemon404> [18:07] <@michaelni> it seems working if bash is used ... <-- well considering it doesnt even have bash by default... :P
[00:21] <ohsix> it probably doesn't have any of the libraries & headers you need to build ffmpeg either
[00:21] <ohsix> by default
[00:23] <michaelni> iive its /bin/sh which seems provided by ksh
[00:23] <cbsrobot> ksh is the default shell on obsd
[00:24] <Daemon404> ohsix, and it also likely has an ancient gcc version in its base
[00:24] <ohsix> awesome
[00:25] <iive> it may be better to report the segfault to openbsd before than trying to workaround it.
[00:25] <ohsix> could be a security issue ;]
[00:25] <durandal_1707> you will be flamed if it is fixed in newer version
[00:26] <Daemon404> he will be flamed regardless
[00:27] <ohsix> it is openbsd afterall
[00:30] <durandal_1707> fourcc A64L is our rgba64le
[01:33] <durandal_1707> wtf r210 still encodes crap on valgrind
[01:34] <ubitux> keep in mind it takes ~6-7 hours to make a run
[01:35] <durandal_1707> ahh, you are right it still compiles old revision
[01:36] <durandal_1707> michaelni: one of openbsd recovered, did you changed something?
[01:40] <durandal_1707> hmm, openbsd machines are compiling unknown revision....
[01:41] <durandal_1707> that is michaelni playing
[01:42] <ubitux> durandal_1707: also, the box will certainly get down in the night so i'll have to restart the run manually tomorrow ;)
[01:49] <michaelni> durandal_1707, i changed the first to bash
[01:50] <durandal_1707> michaelni: but other have same revision and are still red
[01:59] <michaelni> the other 2 run with /bin/sh
[02:00] <michaelni> i dont really want to depend on bash
[02:47] <burek> :/
[02:47] <burek> I still can't get rid of this stupid error: http://pastebin.com/rex228L7
[02:48] <burek> it compiles everything just fine, but linking in the end fails..
[02:57] <funman> burek: hm
[02:58] <funman> why does it pull libfontconfig ?
[02:58] <funman> ah freetype
[02:58] <funman> pkg-config --libs freetype2
[02:59] <burek> # pkg-config --libs freetype2
[02:59] <burek> -L/usr/local/lib -lfreetype
[02:59] <burek> I've compiled both freetype and fontconfig with
[02:59] <burek> --enable-static=yes --enable-shared=no
[03:17] <funman> then your freetype2.pc should have -lfontconfig
[03:17] <funman> which version?
[03:18] <burek> version of fontconfig?
[03:18] <funman> freetype
[03:18] <burek> 2.4.8
[03:19] <funman> yeah it's buggy
[03:19] <funman> hm oh well
[03:19] <funman> fontconfig depends on freetype it seems, not the opposite
[03:20] <burek> fontconfig is from latest git
[03:20] <funman> where does the -lfontconfig come from ?
[03:21] <funman> any way pkg-config --libs fontconfig
[03:21] <burek> let me check
[03:22] <burek> maybe libass
[03:23] <funman> my libass.pc has Requires: fontconfig >= 2.2.0
[03:23] <burek> mine too
[03:24] <burek> and freetype2 >= 9.6.3
[03:26] <funman> hm weird
[03:28] <burek> since "XML_GetCurrentLineNumber" is part of libexpat, it seems like libexpat is missing somewhere..
[03:28] <burek> but I've compiled that too
[03:28] <funman> from fontconfig
[03:28] <funman> .pc
[03:29] <burek> yes
[03:29] <burek> hmh..
[05:29] <CIA-108> ffmpeg: 03Aneesh Dogra 07master * rfef3771778 10ffmpeg/libavcodec/sunrast.c: 
[05:29] <CIA-108> ffmpeg: sunrast: Replace magic number by a macro.
[05:29] <CIA-108> ffmpeg: Signed-off-by: Aneesh Dogra <lionaneesh at gmail.com>
[05:29] <CIA-108> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[05:29] <CIA-108> ffmpeg: 03Aneesh Dogra 07master * r4ffb8e2c14 10ffmpeg/libavcodec/sunrast.c: 
[05:29] <CIA-108> ffmpeg: sunrast: Remove if (unsigned int < 0) check.
[05:29] <CIA-108> ffmpeg: Note: This fixes the following GCC warning :-
[05:29] <CIA-108> ffmpeg: libavcodec/sunrast.c:94: warning: comparison of unsigned expression < 0 is always false.
[05:29] <CIA-108> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[05:29] <CIA-108> ffmpeg: 03Aneesh Dogra 07master * r415f358a1f 10ffmpeg/libavcodec/sunrast.c: 
[05:29] <CIA-108> ffmpeg: sunrast: Cosmetics
[05:29] <CIA-108> ffmpeg: Signed-off-by: Aneesh Dogra <lionaneesh at gmail.com>
[05:29] <CIA-108> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[05:29] <CIA-108> ffmpeg: 03Ronald S. Bultje 07master * re92003514d 10ffmpeg/libavcodec/ (9 files in 2 dirs): png: move DSP functions to their own DSP context.
[05:29] <CIA-108> ffmpeg: 03Ronald S. Bultje 07master * r331e7c4cb3 10ffmpeg/libavcodec/x86/imdct36_sse.asm: 
[05:29] <CIA-108> ffmpeg: imdct36: mark SSE functions as using all 16 XMM registers.
[05:29] <CIA-108> ffmpeg: On x86-64, it indeed uses all 16 registers (and on x86-32, this gets
[05:30] <CIA-108> ffmpeg: Merge remote-tracking branch 'qatar/master'
[05:30] <CIA-108> ffmpeg: * qatar/master:
[05:30] <CIA-108> ffmpeg:  png: add missing #if HAVE_SSSE3 around function pointer assignment.
[05:30] <CIA-108> ffmpeg:  imdct36: mark SSE functions as using all 16 XMM registers.
[05:30] <CIA-108> ffmpeg:  png: move DSP functions to their own DSP context.
[05:30] <CIA-108> ffmpeg:  sunrast: Add a sample request for TIFF, IFF, and Experimental Rastfile formats.
[05:30] <CIA-108> ffmpeg: 03Aneesh Dogra 07master * rf9708e9a0e 10ffmpeg/libavcodec/sunrast.c: 
[05:30] <CIA-108> ffmpeg: sunrast: Add a sample request for TIFF, IFF, and Experimental Rastfile formats.
[05:30] <CIA-108> ffmpeg: Signed-off-by: Aneesh Dogra <lionaneesh at gmail.com>
[05:30] <CIA-108> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[11:37] <CIA-108> ffmpeg: 03Carl Eugen Hoyos 07master * r6838df0488 10ffmpeg/libavcodec/ (imgconvert.c version.h): 
[11:37] <CIA-108> ffmpeg: Add pix_fmt_info for yuva444.
[11:37] <CIA-108> ffmpeg: Reviewed-by: Paul B Mahol
[11:37] <CIA-108> ffmpeg: 03Carl Eugen Hoyos 07master * r36436a4032 10ffmpeg/libavcodec/ (pgssubdec.c version.h): Add option forced_subs_only for Bluray subtitles.
[11:37] <CIA-108> ffmpeg: 03Martin Lambers 07master * r6fe12c7ff2 10ffmpeg/libavformat/img2.c: Detect jps and pns files as (3D) jpg and png images.
[11:37] <CIA-108> ffmpeg: 03Carl Eugen Hoyos 07master * r5f27374c1b 10ffmpeg/libavformat/isom.c: Cosmetics: Add bit per component information to v308 comment.
[12:46] <weekee> anyone here who could possibly do a little modification to the .MOV muxing in ffmpeg for a bounty?
[12:46] Action: av500 would prefers a Snickers
[12:47] <weekee> hehe :)
[12:47] <weekee> man now i want a bounty too
[14:40] <ubitux> broken video: http://dewy.fem.tu-ilmenau.de/CCC/24C3/mp4/24c3-2355-de-tracker_fahrn-COMPATIBLE.mp4
[14:41] <ubitux> it looks like something is wrong with aac
[14:41] <ubitux> i'll open an issue today if i don't forget
[15:29] <ubitux> michaelni: could you comment on http://lists.ffmpeg.org/pipermail/ffmpeg-devel/2012-January/120146.html ?
[16:49] <Compn> http://mpg123.org/beating_mp3lib_in_mplayer/
[16:49] <Compn> anyone with 3dnowex skills want to help the mpg123 dev? :P
[16:49] <Compn> ehe
[16:53] <av500> duron?
[16:54] <av500> k6-3?
[16:54] <teratorn> computers
[16:54] <nevcairiel> ...very old computers
[16:54] <av500> were they 32bit already?
[16:54] <teratorn> yep
[16:55] <Compn> 300mhz > 1ghz ?
[16:55] <Compn> roughly
[16:55] <nevcairiel> but not 686
[16:55] <nevcairiel> or wait, the k6-3 might've been
[16:55] <teratorn> they've been 32-bit since the 386 in 1985
[16:57] <teratorn> k6 was 686 i think
[16:57] <teratorn> k5 was last 586?
[16:57] <teratorn> hm, or not
[16:58] <nevcairiel> i think it was 686
[16:59] <nevcairiel> although wikipedia says its "P5 Pentium-class", which is 586
[16:59] <nevcairiel> oh well
[16:59] <nevcairiel> who cares!
[17:00] <teratorn> but then wiki always lists K5 and K6 as PentiumPro/6th-gen competitors - http://en.wikipedia.org/wiki/Pentium_Pro
[17:01] <Compn> why is amd stuff on pentium page ?
[17:01] <nevcairiel> shows them under competitors
[17:01] <teratorn> in any case, 586 and 686 are the same architecture
[17:01] <Compn> x86
[17:01] <teratorn> with the same extensions
[17:01] <nevcairiel> 686 must have something that 586 doesnt
[17:01] <Compn> more cpu opts
[17:02] Action: Compn doesnt have a clue about this stuff
[17:02] <teratorn> MMX
[17:03] <teratorn> only diff would be weather it was AMD with 3dNow or Intel with SSE
[17:03] <teratorn> I forget when it was that intel and amd adopted/licensed each others 3dn/sse extensons
[17:04] <teratorn> nevcairiel: it's a different, more advanced microarchitecture
[17:04] <teratorn> but from software it's the same thing
[17:05] <nevcairiel> intel never cared to license 3dnow, unless i missed that bit over the years
[17:06] <teratorn> hmm, I guess not
[17:07] <Compn> 3dnow is dead, i think 
[17:07] <nevcairiel> i remember AMD saying they are removing it
[17:08] <teratorn> yeah
[17:08] <teratorn> according to wiki they are
[17:08] <nevcairiel> also found an official post about it
[17:17] <ubitux> michaelni: thx
[18:39] <CIA-108> ffmpeg: 03Reimar Döffinger 07master * r6166bf3cbe 10ffmpeg/libavcodec/huffman.c: (log message trimmed)
[18:39] <CIA-108> ffmpeg: Deobfuscate ff_huff_build_tree.
[18:39] <CIA-108> ffmpeg: I have no idea what the idea was behind the original code,
[18:39] <CIA-108> ffmpeg: but the new code is equivalent to it.
[18:39] <CIA-108> ffmpeg: In that loop that places the new node nodes[j] contains
[18:39] <CIA-108> ffmpeg: always the data of the new node (since the steps are always
[18:39] <CIA-108> ffmpeg: in order: FFSWAP copies node[j] to node[j-1], j is decremented).
[18:39] <CIA-108> ffmpeg: 03Reimar Döffinger 07master * r2ab5feafe1 10ffmpeg/libavcodec/huffman.c: 
[18:39] <CIA-108> ffmpeg: huffman: use a simple assignment instead of FFSWAP.
[18:39] <CIA-108> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[18:42] <sagitter> Hello !
[18:43] <durandal_1707> hello
[18:47] <sagitter> i'm trying to build ffmpeg in Fedora 16 
[18:48] <sagitter> in make of ffmpeg, obtain "collect2: ld returned 1 exit status"
[18:48] <sagitter> at this point: libavcodec/libavcodec.a(libx264.o): In function `X264_init':
[18:48] <sagitter> can you help me please ?
[18:53] <relaxed> sagitter: that is a #ffmpeg question
[18:53] <sagitter> ok, thank you
[18:53] <sagitter> sorry
[20:47] <CIA-108> ffmpeg: 03Michael Niedermayer 07master * r9729f140ae 10ffmpeg/libavcodec/diracdec.c: 
[20:47] <CIA-108> ffmpeg: diracdec: Fix integer overflow leading to out of global array read.
[20:47] <CIA-108> ffmpeg: Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
[20:47] <CIA-108> ffmpeg: Reviewed-by: Paul B Mahol <onemda at gmail.com>
[20:47] <CIA-108> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:48] <CIA-108> ffmpeg: 03Michael Niedermayer 07master * rd2a47c9473 10ffmpeg/tests/fate.sh: 
[20:48] <CIA-108> ffmpeg: fate: add support for specifying the shell used to run configure
[20:48] <CIA-108> ffmpeg: Reviewed-by: Paul B Mahol <onemda at gmail.com>
[20:48] <CIA-108> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:13] <Daemon404> michaelni, poke (because i have a question about a file you wrote)
[22:13] <durandal_1707> what file?
[22:13] <Daemon404> libavutil/x86_cpu.h
[22:14] <Daemon404> it;s teh only file that is stopping libpostproc from being standalone
[22:14] <Daemon404> (i.e. it is not installed during make install of ffmpeg)
[22:14] <Daemon404> i mean it's just a bunch of defines... so...
[22:14] <Daemon404> i might just package it with postproc
[22:15] <Daemon404> yeah ill do that
[22:16] <Daemon404> and remove config.h
[22:21] <ubitux> michaelni: your proposition seems to work (http://pastie.org/3284781), but i'd like to understand the cur_dts check; if cur_dts is AV_NOPTS_VALUE or any other value`0, what does this mean?
[22:22] <ubitux> i have very hard time understanding what's going on in lavu/utils.c
[22:40] <durandal_1707> would it have sense to make emboss filter or this is already doable with current code?
[22:43] <michaelni> vf_geq could probably be used but a dedicated filter would be faster & easier to use
[22:43] <michaelni> ubitux, cur_dts!=0 pretty much means there was a packet with known timestamp or duration
[22:43] <michaelni> thats also why mp3 failed
[22:44] <michaelni> it needs to demux & parse a few packets to determine duration
[22:44] <ubitux> can't this known timestamp be 0?
[22:44] <michaelni> i can
[22:44] <michaelni> it can
[22:44] <michaelni> but the first packet with mp3 is often some header liek id3
[22:44] <ubitux> and cur_dts will be ` in that case? :/
[22:44] <michaelni> and their information like duration /sample rate can be random
[22:44] <ubitux> (`0 i mean)
[22:45] <ubitux> also, how come cur_dts==AV_NOPTS_VALUE means the timestamp or duration is known?
[22:45] <michaelni> cur_dts will be set to the timestamp if tehres one and will count upward if a duration can be found
[22:45] <durandal_1707> michaelni: i wonder why that neat filter is not yet ported
[22:47] <michaelni> ubitux, theres no code that sets cur_dts to AV_NOPTS_VALUE unless there was a known timestamp before
[22:47] <ubitux> is it reliable to assume this will always be true?
[22:49] <michaelni> durandal_1707, it would be great if someone ported it ...
[22:50] <michaelni> ubitux, you could exclude AV_NOPTS_VALUE from the check
[22:51] <ubitux> ok, last question: anything in particular i need to check when doing modification in that code?
[22:51] <ubitux> i'm not sure "passing fate" is enough :
[22:51] <ubitux> :p
[22:52] <michaelni> improve fate :)
[22:52] <ubitux> hehe sure, but what kind of checks should i add? :)
[22:57] <michaelni> dunno, but anything that failes and is fixed could get a fate test in theory
[22:58] <michaelni> that is a fate that for every ticket we fix from 1 to where we are now :)
[22:58] <michaelni> s/that/test/
[22:58] <ubitux> sounds like a good GSoC ;)
[22:59] <ubitux> a bit boring maybe :)
[23:32] <CIA-108> ffmpeg: 03Clément BSsch 07master * r7c0d30b57b 10ffmpeg/libavcodec/ass.c: ass: use av_asprintf() instead of a temporary stack buffer.
[23:32] <CIA-108> ffmpeg: 03Clément BSsch 07master * rcc7342f712 10ffmpeg/libavcodec/assdec.c: assdec: fix wrong alloc pointer check.
[23:32] <CIA-108> ffmpeg: 03Clément BSsch 07master * r6a3cc21242 10ffmpeg/libavformat/mp3dec.c: 
[23:32] <CIA-108> ffmpeg: mp3dec: fix truncating packet error message while decoding a complete MP3.
[23:32] <CIA-108> ffmpeg: The MP3 demuxer split the data in packets of 1024B which are later split
[23:32] <CIA-108> ffmpeg: in MP3 frames by the MPEG audio parser. The last read is "truncated",
[23:32] <CIA-108> ffmpeg: but this should not raise any error.
[23:32] <CIA-108> ffmpeg: Solution-by: Michael Niedermayer
[23:54] <CIA-108> ffmpeg: 03Michael Bradshaw 07master * rf0d6828dbe 10ffmpeg/libavcodec/libopenjpegdec.c: Supoort decoding of gray16 images with libopenjpeg.
[00:00] --- Tue Jan 31 2012