[Ffmpeg-devel-irc] ffmpeg-devel.log.20120915

burek burek021 at gmail.com
Sun Sep 16 02:05:02 CEST 2012


[00:00] <michaelni> did you check if its encoder or decoder issue ?
[00:00] <Daemon404> compare md5s to a known good encode?
[00:00] <michaelni> well you can try the msvc encoded png with linux decoder and the other way around
[00:01] <Daemon404> true
[00:01] <Daemon404> well
[00:01] <Daemon404> the md5 of the encoded mpng files are identical
[00:01] <Daemon404> on msvc and on linux
[00:01] <Daemon404> so it's a decoder issue
[00:02] <michaelni> how does the decoder output look (to a human) ?
[00:02] <Daemon404> it looks normal to me...
[00:03] <Daemon404> i can decode to rgb and binary diff tho
[00:03] <Daemon404> or upload both
[00:14] <Daemon404> michaelni, looks like the mpng decoder decodes frame 9 as frame 8
[00:14] <Daemon404>  ebeee94d24a1bda5aa581e17f57dec84 *07.png
[00:14] <Daemon404> -770ec5af429d91ed3b88bc5cb38f0a17 *08.png
[00:14] <Daemon404> +4819315201834d17289c4f441995a904 *08.png
[00:14] <Daemon404>  4819315201834d17289c4f441995a904 *09.png
[00:14] <Daemon404> diffing the frame md5s
[00:27] <iive> and #08 is the same as #09
[00:27] <Daemon404> [18:14] <@Daemon404> michaelni, looks like the mpng decoder decodes frame 9 as frame 8
[00:27] <Daemon404> ^ what i said ;)
[00:29] <saste> uhm should I quote the "," in filters.texi examples?
[00:32] <iive> Daemon404: frame 9 is decoded as frame 9.
[00:32] <Daemon404> yes
[00:33] <Daemon404> but frame is is decode as frame 9
[00:33] <Daemon404> and i dont know where to look
[00:33] <Daemon404> :D
[00:33] <Daemon404> s/frame/frame 8/
[00:49] <Daemon404> Error while decoding stream #0:0: Error number -1 occurred
[00:49] <Daemon404> hmm
[00:49] <Daemon404> useful -_-\
[00:53] <saste> Daemon404, no strerror_r available?
[00:54] <ubitux> he'll get "permission denied" anyway :)
[00:54] <Daemon404> ?
[00:54] Action: Daemon404 doesnt follow
[00:54] <ubitux> -1 is permission denied
[00:54] <Daemon404> yes
[00:54] <Daemon404> its also generic as fuck
[00:55] <ubitux> and the message "error number -1 occured" seems to be because there is no strerror_r
[00:55] <Daemon404> i dont even know where to look
[00:55] <Daemon404> this failure doesnt happen in lobav
[00:55] <Daemon404> ;libav*
[00:55] <ubitux> look at the given demuxer?
[00:55] <saste> Daemon404, most demuxers/codecs return -1 lame error codes
[00:56] <Daemon404> this is png
[00:56] <Daemon404> there is no demuxer
[00:56] <saste> that was the usual practice, before we reworked the error codes a few years ago
[00:56] <Daemon404> or maybe there is a failure in reading the files
[00:56] <Daemon404> since it happens with mpng too
[00:56] <Daemon404> as i showed above
[00:56] <Daemon404> (which is in avi)
[00:56] <saste> also sadly mingw/windows lacks a strerror_r
[00:56] <Daemon404> this is MSVC
[00:56] <Daemon404> :P
[00:59] <Daemon404> mpng also as that -1 error
[00:59] Action: Daemon404 is convinced it is in pngdec.c
[01:00] Action: ubitux gets crazy trying to understand oggdec again :(
[01:00] <Daemon404> man i dunno how to tarck down a stupid -1 code
[01:02] <Daemon404> i dont even know if it comes from lavc or lavcf
[01:02] <Daemon404> s/cf/f/
[01:02] <saste> Daemon404, gdb is the answer
[01:02] <Daemon404> i think you mean msvc debugger ;)
[01:02] <saste> Daemon404, yeah :)
[01:02] <Daemon404> and i still dont know how i would, besides guessing
[01:02] <Daemon404> no clear place to put a bp
[01:03] <ubitux>     struct ogg_state *ost =
[01:03] <ubitux>         av_malloc(sizeof (*ost) + (ogg->nstreams-1) * sizeof (*ogg->streams));
[01:03] <ubitux> wait. WHAT
[01:03] <Daemon404> :D
[01:03] <Daemon404> well it's definitely happening in pngdec.c
[01:04] <Daemon404> confirmed that
[01:04] Action: Daemon404 looks for logless goto fails
[01:07] <Daemon404> found it
[01:08] <Daemon404> png_decode_idat fails
[01:09] <saste> zlib
[01:09] <saste> more inter-library debugging fun
[01:11] CTCP re:  from CIA-57 (CIA-57!cia at cia.vc) to #ffmpeg-devel
[01:11] <CIA-57> ffmpeg: 03Jemy Tran 07master * r75d34864d1 10ffmpeg/ (doc/filters.texi libavfilter/version.h libavfilter/vf_hue.c): 
[01:11] <CIA-57> ffmpeg: lavfi/hue: add dynamic expression evaluation support
[01:11] <CIA-57> ffmpeg: Signed-off-by: Stefano Sabatini <stefasab at gmail.com>
[01:12] <Daemon404> ... i wonder....
[01:12] Action: Daemon404 disables zlib asm
[01:14] <Daemon404> nope.
[01:19] Action: Daemon404 is stumped
[01:26] <Compnn> png demuxer is image2 demuxer ...
[01:26] <Compnn> rite ?
[01:27] <Daemon404> Compn, its an error in pngdec
[01:27] <Compn> ah
[01:29] <Daemon404> cp -hmm
[01:29] <Daemon404> hmm*
[01:29] <Daemon404> if i cp over libav's pngdec.c
[01:30] <Daemon404> same error
[01:33] <llogan> those aren't supposed to breed.
[01:35] <Daemon404> :|
[01:36] <Compn> are they even different ?
[01:36] Action: Compn asks dumb question
[01:36] <Compn> did you make bugreport yet ?
[01:37] <Daemon404> its msvc-specific
[01:37] <Daemon404> no i havent
[01:37] <Compn> ah
[01:37] <Daemon404> its not img2dec...
[01:38] <Compn> i thought msvc was easier to debug
[01:38] Action: Compn trolls
[01:38] <Daemon404> i dont know WHERE to debug
[01:38] <Daemon404> i dont know where tahn dam nerror is happening
[01:38] <Compn> start putting in printfs
[01:38] <Daemon404> already found where it errors out
[01:38] <Daemon404> i dont know why
[01:38] <Compn> oh
[01:38] <Compn> paste debug report then ?
[01:38] <Daemon404> debug report?
[01:38] <Compn> like gdb output
[01:38] <Compn> but whatever msvc uses
[01:39] <Daemon404> its nto a crash!@
[01:39] <Daemon404> theres no "outut"
[01:39] <Daemon404> output*
[01:39] <Compn> oh 
[01:39] <Compn> make it crash then :P
[01:39] Action: Compn not helpful
[01:39] <Compn> you shouldnt talk to me
[01:39] <llogan> lol
[01:40] <Daemon404> how do i get debug output from pngdec
[01:40] <Daemon404> -v debug doesnt seem to work
[01:40] <Compn> good question
[01:41] Action: Daemon404 uncomments debug define in pngdec
[01:42] <Daemon404> aaaand not useful
[01:42] <Compn> From: "Ronald S. Bultje" <rsbultje at gmail.com>
[01:42] <Compn> This conforms to C99, but requires Windows >= XP.
[01:42] <Compn> ugh
[01:42] <Daemon404> fuck you
[01:42] Action: Compn looks at his win2k box
[01:42] <Daemon404> people who run win2k in 2012
[01:42] <Daemon404> should rot in hell
[01:42] <Daemon404> <_<
[01:43] Action: Daemon404 is tired of this BS about support unsupported oses
[01:43] <Compn> people who break compatability should rot in hell!
[01:43] <Daemon404> it is an unsupported OS by its own fuckign creator
[01:43] <Compn> for 'design choices' lol
[01:43] <Daemon404> and supporting it actively harms ffmpeg
[01:43] <Compn> chrome does that not-working-on-win2k shit too
[01:43] <Compn> really annoying
[01:44] <Daemon404> youre the annoying one
[01:44] <Daemon404> >_>
[01:44] <Compn> ehe, even Zeranoe / KSHawkEye 's ffmpeg builds dont work on win2k
[01:44] <Daemon404> "bawwww my old unsupported and security-hole ridden os isnt supported by new software"
[01:44] <Compn> because iirc hes linking against xp dlls
[01:45] <Daemon404> anyway
[01:45] <Daemon404> heading home while a debug build builds
[01:45] <Daemon404> michaelni, maybe you can help me after :/
[01:45] Action: Daemon404 is off
[01:45] <Compn> my os is so old, security exploits have stopped caring about it :D
[01:46] <Compn> no one writes exploits for dos! you guys are missing out on being completely unexploitable! :P
[01:46] Action: Compn laughs to himself
[01:48] <Compn> ubitux : if you are asking about ogg demuxer, i think everyone gave up on it. including mru
[01:49] <ubitux> well what i'm doing here is not much complicated
[01:49] <ubitux> i'm just trying to get a clue a little about it
[01:49] <ubitux> so i can fix the issue :p
[01:51] Action: Compn sees a bunch of new code being added
[01:51] <Compn> yeah. sure.
[01:52] <Compn> good luck, anyways
[01:54] Action: Compn should go do something more useful than troll
[01:57] <CIA-57> ffmpeg: 03Ben Jackson 07release/0.11 * rf1ec792ae3 10ffmpeg/libavcodec/pthread.c: (log message trimmed)
[01:57] <CIA-57> ffmpeg: pthread: Avoid crashes/odd behavior caused by spurious wakeups
[01:57] <CIA-57> ffmpeg: pthread_wait_cond can wake up for no reason (Wikipedia: Spurious_wakeup).
[01:57] <CIA-57> ffmpeg: The FF_THREAD_SLICE thread mechanism could spontaneously execute jobs or
[01:57] <CIA-57> ffmpeg: allow the caller of avctx->execute to return before all jobs were complete.
[01:57] <CIA-57> ffmpeg: This adds tests to both cases to ensure the wakeup is real.
[01:57] <CIA-57> ffmpeg: Signed-off-by: Ben Jackson <ben at ben.com>
[01:57] <CIA-57> ffmpeg: 03Michael Niedermayer 07release/0.11 * r9f1e01c991 10ffmpeg/libavcodec/escape124.c: 
[01:57] <CIA-57> ffmpeg: escape124: fix integer overflow leading to excessive memory allocation
[01:57] <CIA-57> ffmpeg: Fixes Ticket1629
[01:57] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:57] <CIA-57> ffmpeg: (cherry picked from commit 3d7817048cb387de87600f2152075f78b37b60a6)
[01:57] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:57] <CIA-57> ffmpeg: 03Michael Niedermayer 07release/0.11 * r55b3e408fa 10ffmpeg/libavcodec/faxcompr.c: 
[01:57] <CIA-57> ffmpeg: (cherry picked from commit 8a57ca5c6a1c0ad28afa7ea6f824981e6761cce1)
[01:57] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:57] <CIA-57> ffmpeg: 03Michael Niedermayer 07release/0.11 * rb5e72345a6 10ffmpeg/ (Doxyfile RELEASE VERSION): 
[01:57] <CIA-57> ffmpeg: Update for 0.11.2
[01:57] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:57] <CIA-57> ffmpeg: 03Michael Niedermayer 07release/0.11 * ra5069caecd 10ffmpeg/libavutil/arm/intmath.h: 
[01:57] <CIA-57> ffmpeg: arm/intmath: disable av_clip_uintp2_arm with config_small
[01:57] <CIA-57> ffmpeg: The code otherwise breaks compilation (gcc 4.5 tested but probably any compiler)
[01:57] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:57] <CIA-57> ffmpeg: (cherry picked from commit 2942431a0f94ecb77744a772c164ab71711ffe19)
[01:57] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:57] <CIA-57> ffmpeg: 03Michael Niedermayer 07release/0.11 * r450e4b1a60 10ffmpeg/libavcodec/sp5xdec.c: 
[01:57] <CIA-57> ffmpeg: sp5xdec: fix off by 1 error causing a crash
[01:57] <CIA-57> ffmpeg: Fixes Ticket1633
[01:57] <CIA-57> ffmpeg: Found-by: Piotr Bandurski <ami_stuff at o2.pl>
[01:57] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:57] <CIA-57> ffmpeg: (cherry picked from commit f0896a6bd94e5b45447c7d640c8e8aa95d860d7a)
[01:58] <ohsix> wikipedia, seriously?
[01:58] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:58] <CIA-57> ffmpeg: 03Michael Niedermayer 07release/0.11 * rd721cb009d 10ffmpeg/libavcodec/bmv.c: 
[01:58] <CIA-57> ffmpeg: bmv_videodec: fix out of array read
[01:58] <CIA-57> ffmpeg: Fixes Ticket1373
[01:58] <CIA-57> (3 lines omitted)
[03:56] <CIA-57> ffmpeg: 03Akihiro Tsukada 07master * rc3c646a868 10ffmpeg/libavcodec/ (aac.h aacdec.c avcodec.h version.h): (log message trimmed)
[03:56] <CIA-57> ffmpeg: aacdec: add support for dual mono in Japanese DTV
[03:56] <CIA-57> ffmpeg: Japanese DTV uses some non standard extensions in AAC audio.
[03:56] <CIA-57> ffmpeg: One example is 'dual mono', which combines two independent
[03:56] <CIA-57> ffmpeg: audio into one stereo stream, storing them in left and right channels
[03:56] <CIA-57> ffmpeg: respectively. Historically, dual mono audio has been used for
[03:56] <CIA-57> ffmpeg: multi-lingual audio, one for local/native language, and another for english,
[04:08] <Daemon404> michaelni, do you understand ffmpeg.c and pals very well?
[04:08] <Daemon404> maybe ubitux?
[04:10] <Compn> you know michael has maintained that code for years, right ?
[04:10] <Compn> (i read it in wikipedia! :P)
[04:10] <Daemon404> ;p
[04:15] <michaelni> Daemon404, what exactly is the question ?
[04:16] <Daemon404> michaelni, how exactly does "1 dup!" get tripped?
[04:16] <Daemon404> im still trying to solve that png problem
[04:16] <Daemon404> and i noticed on the failing build, it says 1 dup
[04:16] <michaelni> if its just duplciate then maybe theres something with the timestamps (float rounding m,aybe)
[04:17] <Daemon404> well its failing in pngdec too
[04:17] <Daemon404> but the code in pngdec.c is not at fault
[04:18] <michaelni> what is failing ?
[04:18] <michaelni> and what is at fault
[04:18] <michaelni> ?
[04:18] <Daemon404> i have no idea what is at fault
[04:18] <michaelni> so what do you know ?
[04:18] <Daemon404> im having quite a hard time figuring out what is causing the two png tests to go:
[04:19] <Daemon404> Error while decoding stream #0:0: Error number -1 occurred
[04:19] <Daemon404> [png @ 00FF3AF0] row_size=1056 crow_size =1057
[04:19] <Daemon404> *** 1 dup!
[04:19] <Daemon404> that -1 is because inflate returns an error in png_decode_idat
[04:19] <Daemon404> all of this only happens on MSVC...
[04:20] <michaelni> hmm, lacking a better idea quickly id suggest you dump all that is input and output from zlib and compare between linux and msvc
[04:20] <michaelni> that is with same encoded input
[04:21] <michaelni> does the input to zlib then differe first or the output
[04:21] <Daemon404> im better input differs, but ill check
[04:22] <michaelni> if its input then question is where what exactly ?
[04:23] <Daemon404> yes... its all very tedious and confusing to track
[04:25] <michaelni> also if its the zlib input the the question is if the png decoder gets the same input (data/size/pts/dts/flags)
[04:28] <Daemon404> zlib is returning Z_DATA_ERROR
[04:28] <Daemon404> gonna check input...
[04:48] <Daemon404> input is the same
[04:48] <Daemon404> @ michaelni 
[04:48] <Daemon404> really weird.
[04:48] <Daemon404> at leats teh first 8 bytes are
[04:58] <Daemon404> [22:57] < Daemon404> "incorrect data check"
[04:58] <Daemon404> [22:57] < Daemon404> "inflate:   check matches trailer"
[04:58] <Daemon404> er, latter is not relevant
[05:04] <michaelni> Daemon404, so input to zlib is ok and output is not? if thats so then zlib is broken or theres some mem corruption or zlib is used incorrectly (undefined behavior)
[05:05] <michaelni> does zlib have some regression tests itself ?
[05:05] <Daemon404> not as far as i know
[05:05] <Daemon404> same zlib works fine with libav
[05:05] <michaelni> or maybe theres some known good binary zlib somewhere ?
[05:05] <Daemon404> it's really weird
[05:06] <Daemon404> ive tried multiple zlibs
[05:06] <Daemon404> well i only tested teh first 8 bytes of input
[05:06] <Daemon404> it might get cut off or something
[05:16] <Daemon404> humm... input is the same
[05:16] <Daemon404> :|
[05:20] <michaelni> if you pass exactly the same in all the zlib functions and an error comes out well ...
[05:21] <michaelni> time to debug zlib :)
[05:21] <Daemon404> :<
[05:21] <michaelni> is there some valgrind like tool on windows ?
[05:21] <Daemon404> im sure there is
[05:21] <michaelni> also maybe try compiling without optimizations and without any asm
[05:21] <Daemon404> i dont know what it is though
[05:21] <Daemon404> i have no asm atm
[05:23] <ohsix> does it fail the same way with wine?
[05:24] <Daemon404> good question
[05:24] <Daemon404> lets try
[05:28] <Daemon404> ohsix, sure does
[05:29] <ohsix> free valgrind
[05:30] <Daemon404> can valgrind run on wine stuff?
[05:31] <ohsix> yea
[05:31] <ohsix> just means a bit more noise to suppress
[05:32] <ohsix> http://wiki.winehq.org/Wine_and_Valgrind
[05:32] <michaelni> btw just found: http://stackoverflow.com/questions/413477/is-there-a-good-valgrind-substitute-for-windows
[05:34] <michaelni> does address sanitizer work on windows ?
[05:34] <ohsix> that winehq wiki article also has a list of some alternatives
[05:34] <ohsix> and links to that question :p
[05:35] <Daemon404> michaelni, i dont know what it does so i dont know
[05:36] <michaelni> address sanitizer ?
[05:36] <michaelni> its like valgrind
[05:36] <michaelni> just faster and it misses some things and it finds some that valgrind misses
[05:36] <michaelni> but it wont work with msvc :(
[05:36] <michaelni> iam an idiot
[05:37] <Daemon404> lol
[05:37] <michaelni> time to go to bed i think ...
[05:38] <ohsix> emulators ftw
[09:50] <CIA-57> ffmpeg: 03Michael Niedermayer 07master * rab94d1be7a 10ffmpeg/libavcodec/bintext.c: 
[09:50] <CIA-57> ffmpeg: bintext: mark hscroll() as unused
[09:50] <CIA-57> ffmpeg: This fixes a warning about a unused function
[09:50] <CIA-57> ffmpeg: Alternatively the function could be removed
[09:50] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[10:48] <CIA-57> ffmpeg: 03Luca Barbato 07master * r4d1f443c3d 10ffmpeg/libavformat/utils.c: 
[10:48] <CIA-57> ffmpeg: avformat: simplify avformat_close_input
[10:48] <CIA-57> ffmpeg: avio_close checks by itself for NULL condition.
[10:48] <CIA-57> ffmpeg: 03Luca Barbato 07master * r44272c1ccc 10ffmpeg/libavformat/utils.c: 
[10:48] <CIA-57> ffmpeg: avformat: refactor avformat_close_input
[10:48] <CIA-57> ffmpeg: Do not crash if the input format is not allocated yet.
[10:48] <CIA-57> ffmpeg: 03Mans Rullgard 07master * r692dd8ed26 10ffmpeg/configure: 
[10:48] <CIA-57> ffmpeg: configure: x86: improve ebp availability check
[10:48] <CIA-57> ffmpeg: Some compilers are extra strict about register usage in main(),
[10:48] <CIA-57> ffmpeg: disallowing ebp in inline asm there while allowing it elsewhere.
[10:48] <CIA-57> ffmpeg: This change makes the test better reflect actual usage.
[10:48] <CIA-57> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[10:48] <CIA-57> ffmpeg: 03Mans Rullgard 07master * r07e87858d4 10ffmpeg/tests/fate/ac3.mak: 
[10:48] <CIA-57> ffmpeg: fate: ac3: add 4.0 and downmix tests
[10:48] <CIA-57> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[10:48] <CIA-57> ffmpeg: 03Luca Barbato 07master * r714508bcb9 10ffmpeg/libavcodec/vorbisdec.c: 
[10:48] <CIA-57> ffmpeg: vorbisdec: ensure FASTDIV denominator is never 1
[10:48] <CIA-57> ffmpeg: In both usages of FASTDIV the denominator might be 1.
[10:48] <CIA-57> ffmpeg: Using a branch could make the function slower than using a normal
[13:14] <burek> is it possible to use ffmpeg in such a way that the input video stream is being copied (-c:v copy) to the output up until a certain moment (a defined frame or point in time with -ss or something) from which it will start encoding frames instead of copying
[13:14] <burek> does that seem like a lot of work to implement?
[13:15] <nevcairiel> seems like a odd requirement
[13:16] <burek> it's based on the discussion taking place on #ffmpeg right now
[13:16] <burek> when people want to have a specific precise length of the output and they would like to keep the quality
[13:38] <Compn> just have them create two clips and merge them together
[13:39] <Compn> well
[13:39] <Compn> yeah, not even mencoder can do that :P
[13:39] <Compn> since it cant take multiple -ovc commands
[13:39] <Compn> iirc
[16:24] <CIA-57> ffmpeg: 03Nicolas George 07master * r1b9a251438 10ffmpeg/ffmpeg.c: 
[16:24] <CIA-57> ffmpeg: ffmpeg: call sub2video_update for end packets.
[16:24] <CIA-57> ffmpeg: Without that change, all subtitles will stay until the next one.
[16:24] <CIA-57> ffmpeg: 03Nicolas George 07master * r9dd8272431 10ffmpeg/libavcodec/utils.c: 
[16:24] <CIA-57> ffmpeg: lavc: set AVSubtitle.pts if possible.
[16:24] <CIA-57> ffmpeg: If the packet has a PTS and the corresponding time base is known,
[16:24] <CIA-57> ffmpeg: set the pts field of the decoded subtitle structure before the
[16:24] <CIA-57> ffmpeg: call to the decoder. The decoder is still allowed to change the
[16:24] <CIA-57> ffmpeg: PTS if necessary.
[16:24] <CIA-57> ffmpeg: 03Nicolas George 07master * r37bbc9eb8b 10ffmpeg/libavcodec/pgssubdec.c: 
[16:24] <CIA-57> ffmpeg: lavc/pgssubdec: use the PTS from the presentation segment.
[16:24] <CIA-57> ffmpeg: According to the sample for trac ticket #1722, PGS subtitles
[16:24] <CIA-57> ffmpeg: are decoded from several packets at the same DTS and varying PTS.
[16:24] <CIA-57> ffmpeg: The PTS from the presentation segment seem to be the valid one;
[16:25] <CIA-57> ffmpeg: 03Nicolas George 07master * ra9e4817716 10ffmpeg/ffmpeg.c: ffmpeg: use PTS from the AVSubtitle structure for sub2video.
[16:25] <CIA-57> ffmpeg: 03Nicolas George 07master * r6d8b32633e 10ffmpeg/ (ffmpeg.c ffmpeg.h): ffmpeg: use PTS from the AVSubtitle structure.
[16:26] <CIA-57> ffmpeg: 03Andrey Utkin 07master * raa7c88df50 10ffmpeg/doc/protocols.texi: 
[16:26] <CIA-57> ffmpeg: Document tcp proto 'timeout' parameter behavior
[16:26] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:26] <CIA-57> ffmpeg: 03Michael Niedermayer 07master * r47277c4153 10ffmpeg/libavcodec/x86/v210.asm: 
[16:26] <CIA-57> ffmpeg: x86/v210: fix xmm clobbers
[16:26] <CIA-57> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:07] Action: Daemon404 thinks fflogger needs to log less
[20:08] Action: ubitux thinks it's fine
[20:22] <nevcairiel> I have it on ignore because its f'ing annoying :p
[20:22] <Daemon404> yea
[20:23] <Daemon404> nevcairiel, i actually like its FATE reporting when stuff fails
[20:23] <Daemon404> but not teh trac sttuff
[20:23] <nevcairiel> thats the only useful thing in it
[20:26] <ohsix> do people actually pay attention to trac without it?
[20:26] <Daemon404> yes
[20:26] <ohsix> oh noes i don't
[20:29] <ubitux> monitoring the trac is a pain for alerts
[20:29] <ubitux> there is a rss stream though
[20:30] <ubitux> but i kind of like the trac activity in real time :)
[20:30] <ohsix> yea
[20:30] <ohsix> there was a whole hubub about getting people and things on irc not too long ago
[20:34] <Daemon404> michaelni, careful when merging c1c8fdab46414e3518400d6ee18916d2444a06d7, as it differs from the patch ffmpeg is getting
[20:52] <CIA-54> ffmpeg: 03Ben Jackson 07master * r23503cadd1 10ffmpeg/libavcodec/ (vp56.c vp56.h): 
[20:52] <CIA-54> ffmpeg: lavc/vp56: Refactor ff_vp56_init into ff_vp56_init/ff_vp56_init_context
[20:52] <CIA-54> ffmpeg: Pave the way for per-thread context initialization.
[20:52] <CIA-54> ffmpeg: Signed-off-by: Ben Jackson <ben at ben.com>
[20:52] <CIA-54> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:52] <CIA-54> ffmpeg: 03Ben Jackson 07master * r6e86d6af6b 10ffmpeg/libavcodec/vp6.c: 
[20:52] <CIA-54> ffmpeg: lavc/vp6: Refactor vp6_decode_init into vp6_decode_init/vp6_decode_init_context
[20:52] <CIA-54> ffmpeg: Pave the way for per-thread context initialization.
[20:52] <CIA-54> ffmpeg: Signed-off-by: Ben Jackson <ben at ben.com>
[20:52] <CIA-54> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:52] <CIA-54> ffmpeg: 03Ben Jackson 07master * reebc11ebc8 10ffmpeg/libavcodec/ (vp5.c vp56.c vp56.h vp6.c): 
[20:52] <CIA-54> ffmpeg: lavc/vp56: Move golden_frame into VP56Context
[20:52] <CIA-54> ffmpeg: Makes golden_frame more like other frame data, paves way for threading
[20:52] <CIA-54> ffmpeg: alpha channel decode.
[20:52] <CIA-54> ffmpeg: lavc/vp6: Implement "slice" threading for VP6A decode
[20:52] <CIA-54> ffmpeg: The YUV channels of VP6 are encoded in a highly linear fashion which does
[20:52] <CIA-54> ffmpeg: not have any slice-like concept to thread. The alpha channel of VP6A is
[20:52] <CIA-54> ffmpeg: fairly independent of the YUV and comprises 40% of the work. This patch
[20:52] <CIA-54> ffmpeg: uses the THREAD_SLICE capability to split the YUV and A decodes into
[20:52] <CIA-54> ffmpeg: separate threads.
[21:06] <Daemon404> oh, nevcairiel, there is a sole (only one!) last thing to figure out for msvc/ffmpeg ... :)
[21:06] <Daemon404> but it's a hard one
[21:06] Action: Daemon404 is almost there...
[21:07] <nevcairiel> whats that?
[21:07] <Daemon404> png decode error
[21:07] <Daemon404> i cannot, for the life of me, figure it out
[21:07] <nevcairiel> oh right
[21:07] <Daemon404> i found exactly where it fails
[21:07] <Daemon404> but it's like.. inexplicable
[21:09] <Daemon404> i was wondering if you could reproduce it
[21:09] <ohsix> you mean the _ihdr thing or inside libpng/libz?
[21:09] <Daemon404> idat
[21:09] <Daemon404> zlib is returning a failure
[21:09] <Daemon404> but its getting teh exact same input
[21:11] <ohsix> well, the input is readable ... i wouldn't want to do it, but it's not a dead end
[21:11] <Daemon404> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/pngdec.c;h=67225bb7af1295272e313f52e8d434bb391236dc;hb=HEAD#l372
[21:11] <Daemon404> this is where it fails
[21:19] <nevcairiel> i suppose i can give it a try
[21:20] <nevcairiel> just have to find all the patches i have to apply :d
[21:20] <Daemon404> nevcairiel, there are only 4
[21:21] <Daemon404> most of which will be pushed
[21:21] <Daemon404> today-ish
[21:21] <Daemon404> let me link you
[21:24] <Daemon404> nevcairiel, pm'd
[21:24] <Daemon404> really wish patchwork wasnt a pain to set up
[21:25] <nevcairiel> thanks, applied them, lets see
[21:26] <nevcairiel> was bored trying to figure out avcodecs subtitle support anyway
[21:26] <Daemon404> lol
[21:27] <Daemon404> i already did most of the tedious work with this bug...
[21:27] <Daemon404> finding what returned -1 in ffmpeg is kind of a massive PITA
[21:27] <nevcairiel> usually i would try to find all places and put av_logs in them
[21:27] <nevcairiel> guess now i can put breakpoints
[21:28] <ohsix> ltrace :>
[21:28] <Daemon404> thats what i did
[21:28] <Daemon404> ohsix, this is windows
[21:28] <ohsix> apitrace
[21:29] <ohsix> err, it's been taken over on google, let me find it
[21:29] <nevcairiel> i think i found the wrong apitrace, it seems to be for tracing opengl calls :P
[21:30] <ohsix> http://www.rohitab.com/apimonitor
[21:35] <nevcairiel> why does a make -j10 on this machine only use around 40% cpu, something is wrong
[21:37] <ohsix> something called a disk
[21:38] <nevcairiel> i should move it over to the ssd and try again =p
[21:38] <ohsix> iotop, latencytop will tell you the big users & waiters
[21:38] <ohsix> the higher the -j number themore nonsensical the io pattern, it's not a free lunch
[21:40] <nevcairiel> Daemon404: i get a build error in tiff.c, apparently the converter has a problem with a designated initializer there, line 81
[21:40] <CIA-54> ffmpeg: 03Derek Buitenhuis 07master * r5a627c14b5 10ffmpeg/mt-work/ (7 files): 
[21:40] <CIA-54> ffmpeg: Remove mt-work directory
[21:40] <CIA-54> ffmpeg: It is all dead and rotting code.
[21:40] <CIA-54> ffmpeg: Also, move yuvcmp.c to tools. Its license is non-existent, however,
[21:40] <CIA-54> ffmpeg: and should be clarified.
[21:40] <CIA-54> ffmpeg: Signed-off-by: Derek Buitenhuis <derek.buitenhuis at gmail.com>
[21:41] <Daemon404> nevcairiel, you need a new ver
[21:41] <Daemon404> sec
[21:41] <nevcairiel> meh :)
[21:41] <Daemon404> tha twas fixed weeks ago
[21:41] <Daemon404> or so
[21:41] <Daemon404> lol
[21:41] <nevcairiel> this one shouldnt be all that old
[21:41] <nevcairiel> well, from the 7th
[21:42] <nevcairiel> how i hate github and their display of "authored 8 days ago", as if i feel like doing math to figure out which date that was exactly
[21:43] <nevcairiel> show me the real date, stupid page
[21:43] <nevcairiel> hm, it was fixed on the 7th, guess my build was from right before that
[21:46] <Daemon404> nevcairiel, uploading
[21:46] <Daemon404> https://www.dropbox.com/s/okh9zzfwqmv7lny/convert-laytest-nev.7z
[21:47] <Daemon404> >laytest
[21:47] <Daemon404> oh well. i still claim to not be illiterate.
[21:47] <nevcairiel> this time i dont even know what its meant to say
[21:48] <Daemon404> latest
[21:48] <nevcairiel> oh
[21:48] <nevcairiel> also, this av_restrict warning is annoying
[21:48] <nevcairiel> stupid mplayer
[21:49] <Daemon404> yea
[21:51] <nevcairiel> strip: command not found .. guess it needs another small patch to get rid of that
[21:51] <nevcairiel> not that it blocks now,i got the exe
[21:51] <Daemon404> yeah
[21:53] <Daemon404> nevcairiel, im sending one now
[21:55] <nevcairiel> i'm running fate now
[21:57] <Daemon404> nevcairiel, check for vsynth2-mpng and lavf-png
[21:58] <nevcairiel> yeah vsynth2-mpng failed
[22:16] <CIA-54> ffmpeg: 03Andrey Utkin 07master * raefed6ca87 10ffmpeg/libavformat/udp.c: 
[22:16] <CIA-54> ffmpeg: udp: pick params from AVOptions also
[22:16] <CIA-54> ffmpeg: This updated version does not deviate from previous behavior on default value of 'buffer_size'
[22:16] <CIA-54> ffmpeg: I skipped porting 'sources', 'block' options for now as they're parsed seriously. So i added TODO remarks.
[22:16] <CIA-54> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:16] <CIA-54> ffmpeg: 03Michael Niedermayer 07master * r0da0b5e0f8 10ffmpeg/libavcodec/pngdec.c: 
[22:16] <CIA-54> ffmpeg: pngdec: av_log() zlib errors
[22:16] <CIA-54> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:17] <Daemon404> lul
[22:54] <nevcairiel> something odd is going on with my debugger, it seems to get confused by something, screwing with its display of variable values
[22:54] <Daemon404> humm
[22:55] <Daemon404> what do you mena
[22:55] <nevcairiel> i put a breakpoint at the start of png_decode_idat, and it  parked there just fine. But when i then step over the first line, suddenly all the variables are broken and can't be inspected anymore
[22:56] <nevcairiel> i guess line information is screwed up because of the pre-processing by the converter
[22:57] <Daemon404> oh
[22:57] <Daemon404> maybe
[22:57] <Daemon404> :<
[22:57] <Daemon404> i thought it would use debug symbols
[22:57] <Daemon404> or store the src in the pdb
[22:57] <Daemon404> so it doesnt get affected
[22:57] <nevcairiel> i dont think the pdb contains full sources
[22:58] <Daemon404> bleh
[22:58] <nevcairiel> maybe it can be made to
[23:05] <CIA-54> ffmpeg: 03Alexander Strasser 07master * rbac1b31bf6 10ffmpeg/libavformat/segment.c: 
[23:05] <CIA-54> ffmpeg: lavf/segment: Simplify CSV field quoting code
[23:05] <CIA-54> ffmpeg: Should also be faster (though I doubt that hardly ever matters
[23:05] <CIA-54> ffmpeg: for the usage here).
[23:05] <CIA-54> ffmpeg: Also remove the pointer copy. Since we do not need to reset the
[23:05] <CIA-54> ffmpeg: pointer to the start of the string, it is not needed anymore.
[23:05] <CIA-54> ffmpeg: Signed-off-by: Alexander Strasser <eclipse7 at gmx.net>
[23:05] <CIA-54> ffmpeg: 03Alexander Strasser 07master * r3245be95ac 10ffmpeg/libavformat/segment.c: 
[23:05] <CIA-54> ffmpeg: lavf/segment: Rename variable in print_csv_escaped_str()
[23:05] <CIA-54> ffmpeg: Rename a local variable to make the code more self-explanatory.
[23:05] <CIA-54> ffmpeg: Signed-off-by: Alexander Strasser <eclipse7 at gmx.net>
[23:05] <CIA-54> ffmpeg: 03Alexander Strasser 07master * r49f3b331c7 10ffmpeg/libavformat/segment.c: 
[23:05] <CIA-54> ffmpeg: lavf/segment: Use distinct IDs for list type ext & csv
[23:05] <CIA-54> ffmpeg: Remove a bogus warning when using segment_list_type csv.
[00:00] --- Sun Sep 16 2012


More information about the Ffmpeg-devel-irc mailing list