[Ffmpeg-devel-irc] ffmpeg-devel.log.20121104

[burek]

[00:01] <ubitux> new fate run available soon
[00:01] <ubitux> fate coverage*
[00:02] <saste> who wants to debug a funny overflow bug?
[00:02] <saste> ffplay -f lavfi "aevalsrc=sin(400*t*2*PI):cos(399*t*2*PI),earwax"
[00:03] <ubitux> haha
[00:04] <durandal_1707> what happens?
[00:05] <durandal_1707> michaelni: so you are going to add more fate tests?
[00:06] <ubitux> new coverage available
[00:06] <michaelni> durandal_1707, dont wait for me, feel free to add all that are missing
[00:06] <michaelni> i have many other things to work on in ffmpeg ...
[00:06] <ubitux> http://lucy.pkh.me/ffmpeg-coverage-snapshots/1351982754/ ’ http://lucy.pkh.me/ffmpeg-coverage-snapshots/1351983968/
[00:06] <ubitux> i cleanup the history
[00:07] <ubitux> you will see the old version in that directory
[00:07] <ubitux> with something like 200 runs iirc
[00:07] <ubitux> max_snapshots=200
[00:08] <ubitux> i wonder why it still has some /x86 code reached
[00:08] <ubitux> (even with max_snapshots=200
[00:08] <ubitux> oups
[00:08] <ubitux> even with disable optims
[00:08] <michaelni> optims != asm
[00:09] <ubitux> oh, it switches to -O0 or similar
[00:09] <ubitux> ok
[00:10] <michaelni> asm should be no problem for coverage but optims might be if it ends up optimizing some code out it can AFAIK no longer know if the code was executed
[00:10] <michaelni> would have executed
[00:10] <durandal_1707> is lavf/avc.c file used by anything?
[00:11] <ubitux> yeah right, i used -O0 explicitely previously in optflags
[00:11] <ubitux> --disable-optimizations is indeed better
[00:12] <ubitux> anyway, config: http://lucy.pkh.me/fate-configs/coverage.cfg, hook script: http://lucy.pkh.me/coverage.sh, history of the coverage: http://lucy.pkh.me/ffmpeg-coverage-snapshots/, and last one: http://lucy.pkh.me/ffmpeg-coverage/
[00:12] <ubitux> everything accessible through the index file anyway
[00:18] <cone-979> ffmpeg.git 03Stefano Sabatini 07d034847f7ab4: lavfi/earwax: add missing NULL check in filter_samples()
[00:44] <durandal_1707> lol i get this: 
[00:44] <michaelni> saste, coverity just found a new issue in vf_drawbox
[00:44] <durandal_1707> NULL @ 0x2a4d3c00] The maximum value for lowres supported by the decoder is 0
[00:44] <durandal_1707> NULL @ 0x2a4d3c00] The maximum value for lowres supported by the decoder is 3
[00:47] <durandal_1707> heh they are actually comming from two different decoders
[00:47] <durandal_1707> if i use -1 their names are displayer and if i use too big value, say 10 their names are not displayed
[01:04] <cone-979> ffmpeg.git 03Michael Niedermayer 07c9e455432988: cook: fix use of uninitialized variable
[01:04] <cone-979> ffmpeg.git 03Michael Niedermayer 075f531386210e: movenc: remove unneeded check
[01:34] <durandal_1707> michaelni: this wavpack demuxer does nonsense stuff, like recreating/injectiing extra bytes into packets
[01:34] <durandal_1707> i do not see logic behind this at all
[01:36] <cone-979> ffmpeg.git 03Michael Niedermayer 074a2599b0e6b8: transform: give avfilter_transform() a return value for returning an error
[03:06] <durandal_1707> i want to use avio to read from buffer in libavformat and at same time be able to use avio to write to url - is this possible?
[03:15] <michaelni> you can have multiple avios and its possible to have ones backed by a buffer, there should be examples somewheer that use a seperate avio for handling zlib compression or to build a header in memory before sending it to a url backed avio
[03:22] <durandal_1707> i look for function which reads buffer from one avio and writes it to another one
[03:59] <durandal_1707> michaelni: i fixed wvenc multi channels muxing, it looks extremly ugly like demuxer, will clean it up tommorow
[04:49] <durandal_1707> michaelni: i get strange thing with ffmpeg -i <input> -acodec pcm_s24be out.voc
[04:57] <michaelni> seems no regression
[04:58] <durandal_1707> solution is obvious
[05:02] <cone-979> ffmpeg.git 03Paul B Mahol 071c5b56eec7bb: vocenc: abort if unsupported codec is used
[05:26] <michaelni> durandal_1707, make: *** [fate-lavf-voc] Error 1
[05:27] <michaelni> unsupported codec
[05:36] <durandal_1707> what???
[05:47] <cone-979> ffmpeg.git 03Paul B Mahol 0750d3a1a962d5: vocenc: fix 10l regression
[11:00] <cone-190> ffmpeg.git 03Michael Niedermayer 074e6fa5be3ce5: Revert "cmdutils: allow specifying the file for -report"
[11:00] <cone-190> ffmpeg.git 03Nicolas George 07e021eeb9f06f: lavu/opt: parse key into a mallocated buffer.
[11:00] <cone-190> ffmpeg.git 03Nicolas George 073bdf4971bab4: lavu/opt: implement av_opt_get_key_value().
[11:00] <cone-190> ffmpeg.git 03Nicolas George 071fa47f8deafe: cmdutils: allow to set the report file name.
[11:55] <cone-190> ffmpeg.git 03Stefano Sabatini 07617d0d41a2cf: doc: add libavcodec.texi
[13:24] <cone-190> ffmpeg.git 03Michael Niedermayer 07bbd7b46b94ee: sab_diamond_search: add assert to double check the minima_count
[13:24] <cone-190> ffmpeg.git 03Michael Niedermayer 07a04698c4a9b6: ffserver: zero dest_addr
[15:18] <divVerent> does anyone know whether .mp4 really allows 64bit timestamps?
[15:18] <divVerent> because, iOS 6.0.1 (and probably all previous versions) only support them up to 32bit
[15:18] <divVerent> when the time stamp exceeds that, video playback stops and seeking beyond that point is impossible
[15:19] <divVerent> so, either it's invalid, and the mp4 muxer should bail out then...
[15:19] <divVerent> or it's valid, Apple deserves being sued to death for not properly supporting mp4, and ffmpeg would need an option to limit the time base denominator
[15:20] <durandal_1707> mp4/mov, etc is same thing
[15:21] <kierank> divVerent: apple devices not following specs
[15:21] <kierank> who would have thought...
[15:25] <durandal_1707> divVerent: how big are that files?
[15:27] <divVerent> durandal_1707: 25min
[15:27] <divVerent> and it guessed a huge timebase, of like 10000k
[15:27] <divVerent> how do I override the time base in ffmpeg cmdline BTW?
[15:27] <divVerent> -time_base option did not work
[15:31] <divVerent> currenly trying -vsync vfr -r 120000
[15:32] <divVerent> but that sounds like breaking x264's VBV stuff
[15:33] <divVerent> or does x264's vfr mode stay active with that?
[15:37] <divVerent> apparently works fine
[15:38] <divVerent> does still anyone have a working usage of -time_base? the option itself is in the manpage but without any text
[15:43] <cone-190> ffmpeg.git 03Paul B Mahol 07b0428741c633: indeo3: move get_buffer() above decode_plane() calls
[17:35] <cone-190> ffmpeg.git 03Carl Eugen Hoyos 07b5d496f3a4c3: Support new transparency colour-spaces with libopenjpeg.
[17:50] <cone-190> ffmpeg.git 03Michael Niedermayer 07a15adabdd32c: aacsbr: fix integer overflow
[17:54] <cone-190> ffmpeg.git 03Paul B Mahol 07c4dd4f19f335: smush: improve probing
[17:58] <cone-190> ffmpeg.git 03Paul B Mahol 071b3dbe3f75a6: lcldec: fix decoding of uncompressed rgb24 sample
[18:02] <cone-190> ffmpeg.git 03Stefano Sabatini 071ba2459fbb5e: lavfi/scale: declare the filter private class
[18:02] <cone-190> ffmpeg.git 03Stefano Sabatini 07008ace352469: doc/libavcodec: apply misc minor typo fixes
[18:15] <cone-190> ffmpeg.git 03Stefano Sabatini 0774a50e75a895: lavu: drop disabled FF_API_OLD_TC_ADJUST_FRAMENUM code
[19:28] <cone-190> ffmpeg.git 03Michael Niedermayer 07712287ef4577: ff_ac3_apply_rematrixing: set flags to NULL
[19:28] <cone-190> ffmpeg.git 03Michael Niedermayer 075eaeb4237b88: ivi_common: dc_transform is needed for intra
[20:12] <cone-190> ffmpeg.git 03Paul B Mahol 076e625d41988e: fate: add some missing dependencies
[20:21] <durandal_1707> michaelni: probetest results are not deterministic
[20:37] <michaelni> durandal_1707, i get the same result between 2 runs
[20:51] <durandal_1707> michaelni: i dont get it? why lvf fails but tak does not?
[20:52] <durandal_1707> if i add tak magic at start of random file it gets misdetected
[20:53] <durandal_1707> if i try same with flac it is not
[20:56] <michaelni> 4 x uppercase has 457k cases, 4x mixed case is 7m cases, so the tak probe is quite a bit harder to fail on random strings
[20:59] <durandal_1707> ok i improved cdxl probing enought that probetest does not show it anymore
[21:04] <durandal_1707> michaelni: is there option to test only single size?
[21:06] <michaelni> i dont think so
[21:18] <michaelni> saste, theres a coverity issue in ffprobe and 2 in libavdevice/lavfi.c
[21:21] <cone-190> ffmpeg.git 03Paul B Mahol 07992a6a53efba: cdxl: add read_probe function
[22:15] <saste> michaelni, lavfi: should be already fixed
[22:16] <saste> ffprobe: should be a false positive
[22:18] <michaelni> saste, can you mark the issues accordingly or should i do it ?
[22:19] <michaelni> it important so others know the issues are dealt with already ...
[22:27] <saste> ok
[22:41] <michaelni> saste,  thx
[22:41] <michaelni> btw, in segment.c oc is used before its initialized in line 389
[22:42] <michaelni> thats CID732220
[22:42] <michaelni> i think you worked on segment.c, want to fix it ?
[23:24] <Daemon404> is there any way to get libavcodec/libavformat to "skip" un-displayable frames
[23:24] <Daemon404> i have a mpeg2 stream with leadign b-frames that rely on a previous GOP that isnt available
[23:24] <Daemon404> and lavc kind explodes and destroys valid frames that come after it too
[23:25] <durandal_1707> looks like bug
[23:25] <Daemon404> there might be a secret im missing
[23:25] <Daemon404> there are a *lot* of secrets around mpeg*.c
[23:28] <ubitux> Daemon404: did you have a look to the mov/amerge crash?
[23:28] <Daemon404> ive kidna been without power for a week :P
[23:28] <Daemon404> (no)
[23:28] <ubitux> raw raw fight the powa~
[23:29] <Daemon404> not to mention the box that happens on is in my work building
[23:29] <Daemon404> which is... not functional post-hurricane
[23:29] <durandal_1707> no power?
[23:29] <cone-190> ffmpeg.git 03Paul B Mahol 07bf8f4db4135c: fate: add MxPEG test
[23:30] <Daemon404> their electrical system got destroyed
[23:30] <Daemon404> by 9 feet of water
[23:30] <ubitux> http://d24w6bsrhbeh9d.cloudfront.net/photo/5718493_700b.jpg ?
[23:31] <Daemon404> lol
[23:31] <Daemon404> more like
[23:31] <Daemon404> http://archrecord.construction.com/news/2012/10/images/slideshow/121031/3.jpg
[23:31] <Daemon404> (real pic of it)
[23:31] <ubitux> nice swimming pool
[23:36] <cbsrobot> Daemon404: did you try with -ec ?
[23:36] <Daemon404> cbsrobot, i am using the api
[23:36] <Daemon404> not teh cli util
[23:37] <cbsrobot> k
[23:37] <Daemon404> i dunno how to do that via C api
[23:41] <Daemon404> hmm
[23:51] <mike_f> I'm wondering if anyone can help me. I'm using ffmpeg to decode some NVIDIA hardware encoder plugin that I'm writing. Unlike with the 3 other plugins I've already written, I'm not seeing any output from ffmpeg on the resulting H264 data. Should I download the source to ffmpeg and debug into it, or is there a way to get to the error without debugging into the source. All I get at the moment 
[23:51] <mike_f> is a -1 return result from avcodec_decode_video2(...).
[23:54] <dmon_> sorry to jump channels, but everyone on the main irc is talking about linux distros and not talking about Video or Audio encoding at all.... I'm hoping someone here may know about LOG-C to REC709 Look-Up Table colour conversion?
[23:55] <dmon_> I've got a lot of ARRI Alexa footage that I would prefer not having to encode with Davinic or Scratch Assimilate.
[00:00] --- Mon Nov  5 2012