[Ffmpeg-devel-irc] ffmpeg-devel.log.20181128

burek burek021 at gmail.com
Thu Nov 29 03:05:03 EET 2018

[01:13:17 CET] <cone-105> ffmpeg 03Mark Thompson 07master:b9aff7a53d59: hwcontext_opencl: Only release command queue if it exists
[02:48:58 CET] <vincent42> hi all, I've some MKV produced from live DVB recording (tvheadend), when trying to remux them (ffmpeg in codec copy mode) I often get wrong DTS  / PTS warnings and the result play with sluttering. 
[02:50:11 CET] <vincent42> However if I transcode them, there's no such problem, so the input stream is correct, it just seems that ffmpeg in copy mode (H264) does not extract all timing informations it could, is there an option to do that (to look deeper in the H264 ES, even though we don't want to re-encode)
[04:12:36 CET] <BradleyS> vincent42: sounds like a question for #ffmpeg
[04:12:55 CET] <BradleyS> probably want to include your ffmpeg --version
[04:13:14 CET] <BradleyS> also, lots of people are sleeping at this hour, so be patient for replies
[04:42:18 CET] <vincent42> BradleyS: noted, thanks
[10:26:39 CET] <cone-155> ffmpeg 03Paul B Mahol 07master:18aea7bdd96b: avcodec/opus: set skip_samples
[12:20:14 CET] <cone-155> ffmpeg 03Peter Ross 07master:533ba0b6b5a5: configure: detect djgpp libc
[12:20:15 CET] <cone-155> ffmpeg 03Peter Ross 07master:dea1224754fd: additional math.h functions for djgpp
[12:20:16 CET] <cone-155> ffmpeg 03Peter Ross 07master:8b7a1c22274e: configure: memalign is broken on djgpp 2.05
[12:20:17 CET] <cone-155> ffmpeg 03Peter Ross 07master:331715534a5b: avpriv_tempfile: add djgpp fallback
[18:57:48 CET] <devinheitmueller> Good afternoon.  Anyone interested in doing some consulting to address some bugs in libavcodecs HEVC decoder?
[19:04:09 CET] <JEEB> durandal_1707: beep ^
[19:04:20 CET] <JEEB> I thought you were interested in work WRT FFmpeg
[19:07:03 CET] <j-b> 'morning
[19:14:43 CET] <kurosu> devinheitmueller, is that for a feature or a bug? For the later, having it on the bugtracker would be nice
[19:15:01 CET] <kurosu> (bug or bugs :)
[19:15:27 CET] <devinheitmueller> It relates to properly decoding interlaced streams, so I guess its a bug, or a feature if you consider supporting interlaced video a feature.  :-)
[19:15:39 CET] <nevcairiel> more like an anti-feature
[19:15:42 CET] <nevcairiel> :)
[19:15:54 CET] <kurosu> ah, that rings a bell, where the thing went nowhere
[19:15:54 CET] <devinheitmueller> Hey, if I could get rid of all interlaced content from the world, I would.  :-)
[19:16:13 CET] <nevcairiel> It would help to not make new content in new codecs :-D
[19:16:46 CET] <devinheitmueller> I wouldnt have objected if the HEVC people said deinterlace your damn content before encoding it.
[19:17:08 CET] <nevcairiel> we almost made it to that
[19:17:18 CET] <nevcairiel> but last minute someone put the damn interlaced metadata in
[19:17:29 CET] <nevcairiel> at least there is no special interlaced coding
[19:17:51 CET] <devinheitmueller> True.
[19:18:05 CET] <devinheitmueller> OTOH, I have yet to see good open source deinterlacing that doesnt crush the CPU.
[19:18:30 CET] <philipl> There's good open source deinterlacingthat crushes the GPU too.
[19:18:43 CET] <devinheitmueller> yadif 2x does a decent job, but wow does it take CPU cycles.
[19:19:48 CET] <devinheitmueller> All you have to do it tune to a channel with a news ticker at the bottom to see how terrible they are.
[19:21:23 CET] <JEEB> I still haven't checked if the mix-mash deinterlacer that mixes yadif and w3f is any good
[19:21:33 CET] <philipl> bwdif?
[19:21:36 CET] <JEEB> yea
[19:21:42 CET] <philipl> It seems to be better than yadif.
[19:22:10 CET] <philipl> I've been trying to write a cuda version but I can't get it to pass a basical visual equivalence test. I'm doing something wrong...
[19:23:13 CET] <devinheitmueller> Anybody have any idea if the vaapi deinterlacing is any good?
[19:23:51 CET] <philipl> Seemed reasonable when I've briefly looked at it in the past.
[19:25:49 CET] <devinheitmueller> Im sure Ill get to poke at it soon enouhg.
[20:05:16 CET] <nevcairiel> durandal_1707: your opus change broke things
[20:10:28 CET] <durandal_1707> fix it
[20:10:56 CET] <JEEB> durandal_1707: also when you get the time read the back-log
[20:26:58 CET] <durandal_1707> nevcairiel: i not gonna fix others bugs, one of those issues happen outside of ffmpeg
[20:27:15 CET] <nevcairiel> you should run fate before pushing, if you dont, then you broke it =p
[20:29:58 CET] <cone-041> ffmpeg 03Paul B Mahol 07master:68f289dacddc: avcodec/opus: check if internal is available
[22:51:57 CET] <cone-041> ffmpeg 03Carl Eugen Hoyos 07master:271a00043ef3: lavf/rawdec: Do not mark streams from raw subtitle demuxers as data streams.
[00:00:00 CET] --- Thu Nov 29 2018

More information about the Ffmpeg-devel-irc mailing list