[Ffmpeg-devel-irc] ffmpeg-devel.log.20120203

[burek]

[00:43] <ubitux> durandal_1707: you are Paul right?
[00:44] <durandal_1707> for you yes
[00:46] <ubitux> just a little suggestion about the apedec patch you sent a few minutes ago; it's generally a good idea to have more expressive message about the error: for instance you could specify what is the unsupported bps
[00:56] <CIA-40> ffmpeg: 03Clément BSsch 07master * r9477fa094b 10ffmpeg/libavutil/timecode.c: lavu/timecode: use @see for URL in doxygen.
[00:57] <CIA-40> ffmpeg: 03Clément BSsch 07master * r174880f580 10ffmpeg/libavfilter/vf_thumbnail.c: 
[00:57] <CIA-40> ffmpeg: lavfi/thumbnail: rename @url to @see.
[00:57] <CIA-40> ffmpeg: @url isn't a valid Doxygen directive.
[01:04] <durandal_1707> some thing is strange when seeking in ape
[01:32] <ubitux> nevcairiel: it seems at least the AAC decoder is changing the allocated extradata
[01:32] <ubitux> sorry i'm still on that ;)
[01:41] <dalecurtis> Hi all, I merging upstream ffmpeg into chromium and noticed an issue with videos w/o sound. It's reproducible with tot ffplay as well. Essentially the last 0.5 secs of video is being dropped. Has anyone else ran into this issue?
[01:41] <dalecurtis> ffplay cli:  ffplay -autoexit -i sync0.mp4 , test case: http://dl.dropbox.com/u/242784/sync0.mp4
[01:42] <dalecurtis> Repros with mp4, ogv, webm.
[01:42] <ubitux> is it a regression?
[01:42] <dalecurtis> Yes, ffmpeg from 6 mo ago doesn't show this issue.
[01:42] <ubitux> could you git bisect and find the problematic commit?
[01:43] <dalecurtis> Sure, let me look.
[01:59] <durandal_1707> michaelni: are you sure your ape changes are correct? I get strange problem when decoding 24bit samples
[01:59] <durandal_1707> by changes I mean from first ape commit to today
[02:01] <durandal_1707> i can give you sample and patch to support 8 and 24 bit
[02:07] <michaelni> durandal_1707, does reverting the commit fix it ?
[02:07] <michaelni> and all patches are always welcome
[02:08] <durandal_1707> will start reverting commits, because seeking is broken too
[02:20] <dalecurtis> ubitux: Looks like the bad commit is 4478e9d8db65ca827f2b3ef3ef6ee806bffdba45 , which is sadly a merge commit from qatar/master.
[02:20] <dalecurtis> Came in on Dec 31.
[02:20] <durandal_1707> dalecurtis: that is threads
[02:20] <durandal_1707> force threads to 1, does issue still happens?
[02:23] <dalecurtis> Yes, -threads=1 solves the problem.
[02:30] <durandal_1707> michaelni: i cant reproduce crashes with png
[02:31] <dalecurtis> durandal_1707: Should I file an issue for this? Or is this known?
[02:31] <durandal_1707> dalecurtis: it is known issue
[02:31] <michaelni> png is a misalignment issue with linesize % 8 != 0 ithink but i didnt debug further
[02:31] <michaelni> or better said that is the reason for the segfault i think
[02:31] <michaelni> i dont know what causes the vis artifacts
[02:32] <dalecurtis> durandal_1707, ubitux: Thanks.
[02:45] <durandal_1707> michaelni: i think your changes actually removed >16bit stuff
[02:45] <michaelni> durandal_1707, which changes ?
[02:46] <durandal_1707> tables
[02:46] <michaelni> which commit are we talking baout ?
[02:46] <durandal_1707> 8d4bef64b4ab7e66cf6ca1fb2bc762103390950a
[02:47] <michaelni> march 2008 :)
[02:48] <michaelni> now if i could remember ...
[02:48] <michaelni> but if that entries are usefull i can put them back ?
[02:48] <durandal_1707> thats for are logs
[02:48] <durandal_1707> you should write in commit message all stuff
[02:48] <michaelni> i know
[02:49] <michaelni> log messages are something i always was bad at
[02:51] <durandal_1707> i hope that I would not be forced to remove bunch of commits, exp ones from Loren
[02:54] <ubitux> michaelni: consider your commits a blog post entry instead of a tweet :)
[02:54] <ubitux> i'm sure you'll try to be a bit more descriptive
[02:59] <durandal_1707> some filese caused apedec to crash, now I would need to search for them
[03:06] <durandal_1707> this is going to be harder
[03:28] <sassan> hi all. I'm writing a protocol plugin and due to the nature of the system it connects to, the protocol receives the metadata seperately, rather than as part of the encoded data. Is there a way for the protocol plugin to make ffmpeg aware of this metadata?
[03:33] <michaelni> sassan, easy way is to throw it into the data stream if thats easy possible
[03:33] <sassan> it's an mp3 stream. not sure how easy that would be
[03:39] <michaelni> what metadata is it ?
[03:39] <sassan> artist, title, etc
[03:44] <michaelni> you could put it in a id3 at the start of the stream
[03:45] <sassan> hmmm, I'll give that a try
[03:45] <sassan> thanks :)
[03:51] <Compn> sassan : like icyhttp / icecast ?
[03:51] <Compn> i think it sends metadata seperately
[03:51] <Compn> somehow
[03:52] <ohsix> in http headers :]
[03:53] <sassan> Compn, last.fm
[04:03] <Compn> you are working on last.fm ?
[04:08] <sassan> Compn, yep
[04:08] <sassan> Compn, well, I'm ripping apart an existing player to turn it into an ffmpeg protocol plugin, yes
[04:11] <Compn> so... working against last.fm's player then :)
[04:11] <Compn> what was that other site that merged with last.fm before last.fm got bought
[04:11] <Compn> something audioscrobbler 
[04:12] <durandal_1707> someone with bunch of 24bit ape files?
[04:14] <durandal_1707> or wav files, MAC.exe does not want to encode wavs generated with ffmpeg
[04:14] <Compn> possibly kostya or some ape site, or whoever did the 24bit
[04:14] <Compn> try using mplayer -ao pcm :P
[04:16] <CIA-40> ffmpeg: 03Justin Ruggles 07master * r83ce51cc7d 10ffmpeg/libavcodec/mpc7.c: 
[04:16] <CIA-40> ffmpeg: mpc7: use av_fast_padded_malloc()
[04:16] <CIA-40> ffmpeg: Avoids doing malloc/free for each frame.
[04:16] <CIA-40> ffmpeg: Also fixes valgrind errors due to use of uninitialized padding bytes.
[04:16] <CIA-40> ffmpeg: Based on a patch by Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[04:16] <CIA-40> ffmpeg: 03Justin Ruggles 07master * r3c432631e9 10ffmpeg/libavcodec/pcm.c: 
[04:16] <CIA-40> ffmpeg: pcmenc: Do not set avpkt->size.
[04:16] <CIA-40> ffmpeg: It is already the correct size as set by ff_alloc_packet().
[04:16] <CIA-40> ffmpeg: 03Justin Ruggles 07master * r236a550c3f 10ffmpeg/libavcodec/x86/dsputil_yasm.asm: 
[04:16] <CIA-40> ffmpeg: Fix a typo in the x86 asm version of ff_vector_clip_int32()
[04:16] <CIA-40> ffmpeg: Specifies the correct number of xmm registers used so that they can be saved
[04:16] <CIA-40> ffmpeg: and restored on Win64 if necessary.
[04:16] <CIA-40> ffmpeg: 03Justin Ruggles 07master * r1a670973a7 10ffmpeg/libavcodec/ (internal.h utils.c): 
[04:16] <CIA-40> ffmpeg: ff_alloc_packet: modify the size of the packet to match the requested size
[04:16] <CIA-40> ffmpeg: This will simplify encoders which use this function to request the exact
[04:16] <CIA-40> ffmpeg: packet size rather than the maximum size.
[04:16] <CIA-40> ffmpeg: 03Paul B Mahol 07master * r41dd77bdbc 10ffmpeg/doc/general.texi: 
[04:16] <CIA-40> ffmpeg: doc: decoding Forward Uncompressed is supported
[04:16] <CIA-40> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[04:16] <CIA-40> ffmpeg: Signed-off-by: Justin Ruggles <justin.ruggles at gmail.com>
[04:16] <CIA-40> ffmpeg: 03Mans Rullgard 07master * r2c98f407c8 10ffmpeg/ (5 files in 4 dirs): 
[04:16] <CIA-40> ffmpeg: fate: make acodec-ac3_fixed test output raw AC3
[04:16] <CIA-40> ffmpeg: There is no point in this test using the RM format.
[04:16] <CIA-40> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[04:16] <CIA-40> ffmpeg: 03Janne Grunau 07master * rfcc518a024 10ffmpeg/doc/APIchanges: APIchanges: add missing commit hashes
[04:16] <CIA-40> ffmpeg: 03Ronald S. Bultje 07master * rde53b9068a 10ffmpeg/libswscale/x86/ (input.asm swscale_mmx.c): swscale: implement MMX, SSE2 and AVX functions for RGB32 input.
[04:17] <CIA-40> ffmpeg: Do not use AVStream's duration for dts generation since it contains in
[04:17] <CIA-40> (33 lines omitted)
[05:00] <durandal_1707> looks like issue is in ape demuxer and not in decoder, ... sigh*
[05:01] <bcoudurier> durandal_1707, where is your patch for ape ?
[05:01] <bcoudurier> I don't see it
[05:01] <durandal_1707> on libav ml
[05:02] <bcoudurier> oh right
[05:03] <Compn> durandal_1707 enjoys the pain of sending patches to two lists :D
[05:04] <bcoudurier> he didn't send this one so I was wondering
[05:04] <Compn> hes not sending same patches to both lists 
[05:04] <Compn> just to add in confuuuusion
[05:04] <Compn> and less merges i guess
[05:04] <bcoudurier> :)
[05:05] <durandal_1707> libav bikesheed more - i like that
[05:06] <bcoudurier> let's troll !
[05:06] <ohsix> kind of makes me wonder if people know they can reply off-list
[05:28] <sassan> michaelni, adding an id3 tag worked, and easier than I thought it would be :) Thanks :)
[13:59] <ubitux> flashsv2 fate is broken or it's me?
[16:49] <michaelni> ubitux, fate.ffmpeg.org is green
[17:12] <durandal_1707> lets make party
[18:01] <ubitux> michaelni: i was wondering if fate.ffmpeg.org wasn't dead
[18:03] <ubitux> i may be just smoking weed, forget this
[18:10] <michaelni> anyone has an idea how to make darwin llvm green too ?
[18:19] <ubitux> mmh i still have vsynth1-flashsv2 failing locally&
[18:20] <ubitux> i'm not sure fate.ffmpeg.org is taking into account the last reports
[18:23] <ubitux> i have no idea what's going on actually
[18:24] <ubitux> my last commit is indeed the d77294c merge, just like what fate reports
[18:28] <ubitux> http://pastie.org/3310367
[18:28] <ubitux> this is the failure i have
[18:29] <ubitux> maybe a zlib update?
[18:59] <CIA-40> ffmpeg: 03Ramiro Polla 07master * rc9e5acad61 10ffmpeg/libavdevice/dshow.c: 
[18:59] <CIA-40> ffmpeg: dshow: fix AVInputFormat declaration after ABI breakage
[18:59] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:59] <CIA-40> ffmpeg: 03Ramiro Polla 07master * r1f0fa80bfa 10ffmpeg/libavdevice/dshow.c: 
[18:59] <CIA-40> ffmpeg: dshow: use DirectShow information to set time_base
[18:59] <CIA-40> ffmpeg: Fixes build after removal of AVFormatParameters.
[18:59] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:53] <ubitux> michaelni: ok i think a zlib update or something broke it
[19:53] <ubitux> http://fate.ffmpeg.org/report.cgi?time=20120203183521&slot=x86_64-archlinux-gcc-enableshared
[19:53] <ubitux> i just did the updates on the box
[19:58] <ubitux> ok, confirmed, downgrading to zlib 1.2.5 fixes the issue
[19:59] <ubitux> something changed in 1.2.6
[20:06] <CIA-40> ffmpeg: 03Vincent Torri 07master * red3c0fe85d 10ffmpeg/libavformat/ (network.h os_support.h): 
[20:06] <CIA-40> ffmpeg: Remove some warnings when compiling with mingw (mingw-w64, 32 bits)
[20:06] <CIA-40> ffmpeg: Reviewed-by: Ramiro Polla <ramiro.polla at gmail.com>
[20:06] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:42] <CIA-40> ffmpeg: 03Reimar Döffinger 07master * rda1ba4e88b 10ffmpeg/libavcodec/x86/rv40dsp.asm: 
[20:42] <CIA-40> ffmpeg: Fix NASM compilation.
[20:42] <CIA-40> ffmpeg: movd needs explicit register size prefix for NASM.
[20:42] <CIA-40> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[23:13] <CIA-40> ffmpeg: 03Reimar Döffinger 07master * r58dabf7bf2 10ffmpeg/libavcodec/ (pngdsp.h x86/pngdsp.asm): (log message trimmed)
[23:13] <CIA-40> ffmpeg: Fix png decoding on x86.
[23:13] <CIA-40> ffmpeg: Line sizes are only 8-byte aligned, so use unaliged loads
[23:13] <CIA-40> ffmpeg: for add_bytes_l2 pointers.
[23:13] <CIA-40> ffmpeg: Increasing the alignment requirement to 16 seemed a bit extreme
[23:13] <CIA-40> ffmpeg: (png may be used for rather small sizes).
[23:13] <CIA-40> ffmpeg: Also fix a mov that had its arguments swapped, leading
[23:34] <danverde> I am writing a program that is using libavfilters.  I currently setup the filter chain with avfilter_graph_parse.
[23:35] <danverde> I now need dynamic arguments for my filters
[23:35] <danverde> is there an example of setting up the graph and parameters a different way
[23:35] <danverde> for example I want to dynamically change a text overlow and scale value of output video frames
[23:49] <michaelni> danverde, theres some code in ffmpeg.c that can dynamically change a limited set of things
[23:50] <michaelni> see avfilter_graph_send_command()
[00:00] --- Sat Feb  4 2012