[Ffmpeg-devel-irc] ffmpeg-devel.log.20150105

[burek]

[00:18] <ubitux> saste: it would be nice to have a sample accurate "aselect" :p
[00:18] <ubitux> with eval it might be way too slow though
[00:25] <wm4> why is asample _not_ sample accurate?
[00:26] <wm4> err aselect
[00:26] <wm4> seems like an achievement
[00:26] <wm4> does it select by frames?
[00:27] <ubitux> yes, afaik
[00:27] <wm4> but frames have an arbitrary size
[00:27] <ubitux> if you want to do it sample accurate you need to evaluate the expression per sample
[00:27] <kierank> ubitux: can a libavfilter filter generate both audio and video at the same time?
[00:27] <ubitux> kierank: yes
[00:27] <kierank> next q
[00:28] <ubitux> i think ebur128 actually does this
[00:28] <wm4> (especially funny because ffmpeg.c/libavformat apparently makes PCM frames in some files arbitrarily huge)
[00:28] <kierank> is it possible then to have a chain of filters and output an exact (but variable) number of samples per frame?
[00:28] <ubitux> i guess so?
[00:28] <ubitux> asetnsamples?
[00:29] <kierank> I mean for each video frame I need n samples?
[00:29] <kierank> basically I want to build my test pattern filter at some point
[00:29] <kierank> with a lipsync test
[00:29] <ubitux> i think you could do that
[00:29] <ubitux> look how asetnsamples does it
[00:30] <kierank> I need it to be variable though (for ntsc)
[00:30] <ubitux> the filter calls ff_get_audio_buffer
[00:30] <ubitux> to which you pass the number of sample you want
[00:30] <ubitux> outsamples = ff_get_audio_buffer(outlink, nb_out_samples);
[00:31] <kierank> the difficulty I have is how to sync my video and audio in ntsc-land
[00:31] <kierank> because you don't sync by video frame in ntsc-land
[00:31] <kierank> s/don't//g
[00:33] <ubitux> well, you could probably write a "sync" filter
[00:33] <ubitux> with an ntsc mode to start with
[00:33] <kierank> it would be easy with a standalone filter, yes
[00:33] <kierank> but chaining filters (to use drawbox etc) is hard
[00:34] <ubitux> why?
[00:34] <ubitux> also, do you need the video stream for this audio pattern?
[00:34] <kierank> yes
[00:34] <kierank> because I need to sync the drawing of a box or whatever with the sound
[00:35] <ubitux> i mean, isn't what you want just a special cut?
[00:36] <ubitux> like, you just apply a certain pattern of frame segmentation to the audio, and that's it?
[00:36] <kierank> it's not that simple
[00:36] <ubitux> typically, can't you hack something in asetnsamples with an eval expression? :p
[00:36] <ubitux> (or just as a PoC)
[00:37] <ubitux> otherwise yeah, feel free to make a filter that output both
[00:37] <kierank> ubitux: afaik you can't do frame accurate drawbox on the cli
[00:38] <kierank> at the moment in my code I have a moving message with sin(k*t)
[00:38] <kierank> but that's not frame accurate
[00:39] <ubitux> well you can add a "n" paramter in eval
[00:39] <ubitux> it's a 2-3 line diff
[00:40] <ubitux> it won't work with seeking (so you can't really test with ffplay), but it will work for a transcode
[00:40] <ubitux> git grep 'VAR_N[^A-Z]' is your friend
[00:42] <llogan> saste: Postprocessing wiki page doesn't exist
[01:21] <KGB> [13FFmpeg] 15michaelni pushed 1 new commit to 06master: 02http://git.io/mca6TQ
[01:21] <KGB> 13FFmpeg/06master 14907101e 15Reynaldo H. Verdejo Pinochet: ffserver: reflow compute_status()...
[02:07] <KGB> [13FFmpeg] 15michaelni pushed 12 new commits to 06master: 02http://git.io/CJ50XA
[02:07] <KGB> 13FFmpeg/06master 149e55130 15Reynaldo H. Verdejo Pinochet: ffserver: reindent compute_status()...
[02:07] <KGB> 13FFmpeg/06master 14196bc03 15Reynaldo H. Verdejo Pinochet: ffserver: drop redundant else clause...
[02:07] <KGB> 13FFmpeg/06master 14ac1940b 15Reynaldo H. Verdejo Pinochet: ffserver: reflow rtsp_cmd_setup()...
[02:18] <KGB> [13FFmpeg] 15michaelni pushed 2 new commits to 06master: 02http://git.io/QmI_tA
[02:18] <KGB> 13FFmpeg/06master 143e16065 15Johan Andersson: cmdutils: update copyright year to 2015....
[02:18] <KGB> 13FFmpeg/06master 141d41571 15Michael Niedermayer: Merge commit '3e160652219ff4da433f5672ae1e5f4956abb815'...
[02:26] <KGB> [13FFmpeg] 15michaelni pushed 2 new commits to 06master: 02http://git.io/1tE5PA
[02:26] <KGB> 13FFmpeg/06master 14f4d4e66 15Martin Storsjö: configure: Remap -L to -libpath for msvc...
[02:26] <KGB> 13FFmpeg/06master 141ed5912 15Michael Niedermayer: Merge commit 'f4d4e66a24a5c9497a5b6d3c089ac58089a87428'...
[02:34] <KGB> [13FFmpeg] 15michaelni pushed 2 new commits to 06master: 02http://git.io/Ls6VLQ
[02:34] <KGB> 13FFmpeg/06master 14c4aa0f8 15Martin Storsjö: libavformat: Remove a misleading/incorrect comment...
[02:34] <KGB> 13FFmpeg/06master 14dd4de4a 15Michael Niedermayer: Merge commit 'c4aa0f865fb2c3b28a309d1bb27fd1c24b75ba28'...
[02:44] <KGB> [13FFmpeg] 15michaelni pushed 2 new commits to 06master: 02http://git.io/JKlKuw
[02:44] <KGB> 13FFmpeg/06master 149f810a9 15Martin Storsjö: libavformat: Check for malloc failures in avformat_new_stream...
[02:44] <KGB> 13FFmpeg/06master 14e7521d9 15Michael Niedermayer: Merge commit '9f810a9b374e0ff8e2a19fd8a7347afe2933e229'...
[04:19] <KGB> [13FFmpeg] 15michaelni pushed 2 new commits to 06master: 02http://git.io/9TklZA
[04:19] <KGB> 13FFmpeg/06master 1446808fd 15Martin Storsjö: movenc: Enable editlists by default if delay_moov is enabled...
[04:19] <KGB> 13FFmpeg/06master 14fb22e5a 15Michael Niedermayer: Merge commit '46808fdf04ab113df374157b90b506eb3110daf2'...
[05:03] <KGB> [13FFmpeg] 15michaelni pushed 1 new commit to 06master: 02http://git.io/eotSfQ
[05:03] <KGB> 13FFmpeg/06master 14beedeb4 15Werner Robitza: doc/ffmpeg: mention both ffpreset/avpreset in documentation, remove superfluous example...
[05:11] <KGB> [13FFmpeg] 15michaelni pushed 1 new commit to 06master: 02http://git.io/uCInAA
[05:11] <KGB> 13FFmpeg/06master 14c9151de 15wm4: avcodec/dvdsubdec: fix out of bounds accesses...
[06:42] <wm4> michaelni: would you still be interested in my sws_scale_frame() API patch I sent over a year ago?
[06:45] <rcombs> wait since when do repo logs come from russian intelligence
[06:47] <wm4> the main git repo is down
[06:47] <wm4> and apparently videolan's infrastructure?
[06:50] <rcombs> oh, so KGB is GitHub and work's being done directly on the mirror then
[06:56] <Tim_G> rcombs: from the README of KGB (assuming one is available): "CIA.vc is dead, long live KGB.vc!"
[06:57] <rcombs> if it works the way I assume it does, it's just GitHub's IRC hook on a custom nick?
[06:57] <Tim_G> rcombs: I guess
[06:58] <Tim_G> Libav is using https://github.com/ehamberg/simple-gitbot with the nick "lCIA"
[06:58] <Tim_G> but they are equally ugly IMO
[06:59] <rcombs> oh hah
[06:59] <rcombs> there's the joke
[07:04] <wm4> so is the KGB nick owned by someone related to ffmpeg, or is it some kind of shared service?
[07:13] <rcombs> seems to be owned by v1adimir?
[10:29] <ubitux> damn ffmpeg is so complex wrt timestamps
[10:36] <compn> rcombs : we also use cia bot, but since videolan is down, cia is not updating...
[10:37] <compn> easier to use kgb for github than reconfigure cia for github it seems
[10:44] <saste> there are issues with the git server today?
[10:44] <ubitux> saste: http://ffmpeg.org/pipermail/ffmpeg-devel/2015-January/167262.html
[10:45] <saste> ubitux, thanks
[12:27] <michaelni> wm4, do you have a link to the patch you speak of ? i dont exactly remember but i just woke up ...
[12:29] <wm4> this one https://ffmpeg.org/pipermail/ffmpeg-devel/2013-September/148824.html (that's the thread)
[12:30] <wm4> I didn't pursue it further, because I thought libavscale could actually happen
[12:31] <nevcairiel> You should know better by now
[12:46] <michaelni> wm4, yes, sure the patch looked/looks useful
[15:15] <cone-342> ffmpeg.git 03Paul B Mahol 07master:91bdb244009c: libavfilter/vf_blend: do not unconditionally compile both filters
[15:49] <durandal_1707> would it be okay to add (a)deinterleave filters?
[16:03] <cone-342> ffmpeg.git 03Johan Andersson 07master:3e160652219f: cmdutils: update copyright year to 2015.
[16:03] <cone-342> ffmpeg.git 03Martin Storsjö 07master:f4d4e66a24a5: configure: Remap -L to -libpath for msvc
[16:03] <cone-342> ffmpeg.git 03Martin Storsjö 07master:c4aa0f865fb2: libavformat: Remove a misleading/incorrect comment
[16:03] <cone-342> ffmpeg.git 03Martin Storsjö 07master:9f810a9b374e: libavformat: Check for malloc failures in avformat_new_stream
[16:04] <cone-342> ffmpeg.git 03Martin Storsjö 07master:46808fdf04ab: movenc: Enable editlists by default if delay_moov is enabled
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:9bff052b51f2: avfilter/vf_sab: fix filtering tiny images
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:0f931b29f7e9: Factorize avpriv_mirror() out
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:58721388b89b: avformat/flvenc: omit more metadata elements with specific meaning
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:dc265e2b6b6b: avformat/rawdec: Make ff_raw_data_read_header() non static
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:d1781fd786de: avformat/rawdec: Add FF_DEF_RAWSUB_DEMUXER
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:e313735d7d7d: avformat: add dvbsub demuxer
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:ce1e10aa24cc: avformat/utils: add dvbsub to set_codec_from_probe_data()
[16:04] <cone-342> ffmpeg.git 03Peter Ross 07master:f1098eb97d3c: avformat/aiffdec: only read codec tag when there is space in header
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:1d41571bbf0d: Merge commit '3e160652219ff4da433f5672ae1e5f4956abb815'
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:1ed5912e819d: Merge commit 'f4d4e66a24a5c9497a5b6d3c089ac58089a87428'
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:dd4de4a987f3: Merge commit 'c4aa0f865fb2c3b28a309d1bb27fd1c24b75ba28'
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:e7521d9771ea: Merge commit '9f810a9b374e0ff8e2a19fd8a7347afe2933e229'
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:fb22e5ae4585: Merge commit '46808fdf04ab113df374157b90b506eb3110daf2'
[16:04] <cone-342> ffmpeg.git 03Werner Robitza 07master:beedeb4399fe: doc/ffmpeg: mention both ffpreset/avpreset in documentation, remove superfluous example
[16:04] <cone-342> ffmpeg.git 03wm4 07master:c9151de7c425: avcodec/dvdsubdec: fix out of bounds accesses
[16:04] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:0e1c827e842c: Merge remote-tracking branch 'origin/master'
[17:41] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:3bf2b376e64b: avformat/dvbsub: check for 0xf0 earlier in probe, 30% faster
[17:41] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:b0826b832465: avformat/dvbsub: fix indent after previous commit
[17:41] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:d02d4d0457f9: avformat/dvbsub: increase probe score
[17:55] <durandal_1707> isn't first hunk of f929ab0569ff31ed5a59b0b0adb7ce09df3fca39 wrong?
[17:57] <ubitux> how so?
[18:19] <ubitux> rhaa ffs i'm so annoyed at these edit lists
[18:20] <JEEB> le edit lists, such a feature, much headache
[18:21] <ubitux> i started yet another approach to the problem: basically exporting the edts atom on demand
[18:22] <ubitux> and building the filtergraph in app side (so cmdutils, shared with ffplay/ffmpeg)
[18:22] <ubitux> (so people do not complain about the lavfi stuff in the libs)
[18:23] <ubitux> the problem is that the unit of time is not in the stream unit
[18:23] <ubitux> it's in the presentation unit, which is not exported
[18:23] <ubitux> ...i need to add a time_base field in the AVFormatContext...
[18:23] <ubitux> unless someone has a better idea
[18:23] <wm4> does anyone know why id3v2.c does this retarded dance around "extra metadata", and where to add custom tag parsers that just want to write to s->metadata?
[18:38] <wm4> kierank: now our problem is apparently that ffmpeg doesn't know how to deal with streams it's supposed to ignore (extra stuff like EPG etc.), any tips?
[18:38] <kierank> why is ffmpeg not ignoring EIT?
[18:39] <wm4> ignoring what?
[18:39] <kierank> EIT is the EPG table
[18:39] <kierank> in dvb
[18:41] <wm4> apparently such streams are detected as unknown streams (codec "none"), and they slow down probing
[18:41] <kierank> probably because of ffmpeg hacks
[18:41] <kierank> when there is no pmt
[18:41] <wm4> now there's a PMT (apparently)
[18:41] <wm4> I know nothing about dvb
[18:52] <ubitux> maybe exporting the filtergraph from the demuxer wasn't that a bad idea in the end
[18:52] <ubitux> rhaaa i'm so fucking annoyed
[18:54] <wm4> ubitux: be assured, it is
[18:54] <ubitux> i'll need to pass a fucking tb in all the per-stream functions of the app
[18:54] <ubitux> how insane is that
[18:54] <wm4> an incredibly unclean hack to make things somehow work for ffmpeg.c, and only ffmpeg.c
[18:54] <ubitux> wm4: i'm trying not to make a hack
[18:55] <ubitux> wm4: as i said, i was basically just exporting the edit list
[18:55] <ubitux> verbatim
[18:55] <ubitux> so the app can implement the timeline on its own
[18:55] <wm4> yes, that sounded ok
[18:55] <ubitux> except it's fucking annoying because as is, it's not usable since you need the presentation
[18:56] <ubitux> the time base* presentation
[18:56] <ubitux> i'm thinking of just putting the time base into that side data as a first field though
[18:56] <ubitux> won't be the exact atom, but it will relieve some pain for everyone
[18:59] <wm4> how does L-SMASH export this stuff?
[19:03] <wm4> also I'm going to implement this http://id3.org/id3v2.3.0#Synchronised_lyrics.2Ftext and everyone will hate me, including myself
[19:18] <compn> wm4 : what you dont like cdg ? do we have id3 lyrics samples ?
[19:18] <compn> oh wow that looks horrible :D
[20:30] <cone-342> ffmpeg.git 03James Almer 07master:08810a889517: x86/flacdsp: remove unneeded ifdeffery
[21:02] <cone-342> ffmpeg.git 03wm4 07master:bd7801040786: avformat/id3v1: strip trailing whitespace
[00:00] --- Tue Jan  6 2015