[Ffmpeg-devel-irc] ffmpeg-devel.log.20190427

[burek]

[00:03:03 CEST] <cone-150> ffmpeg 03Paul B Mahol 07master:bf15dcc5c8d8: avfilter/vf_stack: use time_base from framesync
[00:06:59 CEST] <Lynne> durandal_1707: have you tested your chained opus patch with this stream?
[00:07:01 CEST] <Lynne> http://listen.42fm.ru:8000/stealkill-opus
[00:08:04 CEST] <Lynne> I also have a patch to fix chained opus decoding which works with both the test sample and that stream, just haven't cleaned it up to send it
[00:10:33 CEST] <Lynne> I also made it check page CRCs to reduce errors, and because of how bad my other CRC patches were taken on the mailing list I didn't know whether to remove it or not
[00:11:55 CEST] <durandal_1707> well CRC checking is not improving decoding much :)
[00:13:04 CEST] <durandal_1707> Lynne: my patch produces same number of samples as opusdec tool but with some small artifacts, i dunno what is source of that one
[00:14:11 CEST] <jamrial> durandal_1707: did you try decoding with libopus instead of the internal decoder?
[00:15:53 CEST] <Lynne> the demuxer sets the preskip so it doesn't matter which one is used
[00:16:20 CEST] <durandal_1707> jamrial: nope
[00:16:56 CEST] <Lynne> durandal_1707: the source is a recent icecast version, that russian radio station is the only chained opus stream I've found
[00:17:41 CEST] <Lynne> it was hard to get both it and the sample file to decode properly, I asked on #opus which one was incorrect but I wasn't told
[00:24:04 CEST] <cone-150> ffmpeg 03Paul B Mahol 07master:c2f305ca17e3: avfilter: add audio soft clip filter
[00:41:06 CEST] <durandal_1707> Lynne: what you use to play that stream?
[00:57:11 CEST] <Lynne> durandal_1707: normal ffmpeg opus decoder, with my patch
[00:57:53 CEST] <durandal_1707> Lynne: i mean ffplay/vlc/mpv ?
[00:58:33 CEST] <Lynne> Oh, wget to a file, wait until there's a switch, then ffmpeg
[01:17:54 CEST] <durandal_1707> Lynne: send your work/patch without unrelated stuff to ML
[01:40:17 CEST] <cone-150> ffmpeg 03Lynne 07master:4b7166c9d57d: x86/opusdsp: replace loads with shuffles
[09:50:11 CEST] <cone-138> ffmpeg 03Jun Zhao 07master:d6489ddb7a6c: lavf/hls: Remove HLSContext.strict_std_compliance field
[12:48:10 CEST] <cone-138> ffmpeg 03Timo Rothenpieler 07master:23ed147e8fc2: avcodec/nvenc: only unregister input resources when absolutely needed
[12:48:11 CEST] <cone-138> ffmpeg 03Timo Rothenpieler 07master:2e254bb89747: avcodec/nvenc: fix indentation
[12:59:48 CEST] <cone-138> ffmpeg 03Paul B Mahol 07master:6347146e3d3d: avformat/subtitles: ignore extra '\r' at line endings
[12:59:49 CEST] <cone-138> ffmpeg 03Paul B Mahol 07master:163bb087f81d: avformat/microdvddec: skip empty lines
[13:26:50 CEST] <cehoyos> michaelni: The pgm regression has real-life effect afaict.
[13:36:42 CEST] <michaelni> cehoyos, what pgm regression ?
[13:43:40 CEST] <durandal_1707> michaelni: was it seriously coded in such a way that PGM is reparsed from start after each combine frame?
[13:52:53 CEST] <cehoyos> michaelni:I sent an email to -devel after a pgm regression was reported on -user
[13:53:35 CEST] <cehoyos> (I already wonder if you will be patient enough to test the actual slowdown...)
[14:09:16 CEST] <michaelni> yes, thats messed up
[14:13:51 CEST] <durandal_1707> michaelni: i have fix
[14:14:49 CEST] <michaelni> i have too but didnt test yet
[14:16:45 CEST] <durandal_1707> https://pastebin.com/JtRywy9C
[14:39:08 CEST] <atomnuker1> durandal_1707: I've combined your patch and my patch - https://paste.ubuntu.com/p/jXnhYkC7vf/
[14:50:20 CEST] <durandal_1707> atomnuker1: is that some kind of joke?
[14:51:19 CEST] <durandal_1707> atomnuker1: ah this is another thing, please state it when saying something
[15:01:00 CEST] <durandal_1707> atomnuker1: there is still invalid argument erro on song switch
[15:17:50 CEST] <durandal_1707> atomnuker1: also why we can not get full metadata update on song switch?
[15:19:28 CEST] <durandal_1707> also for chained ogg, it would be best if each new component uses pts of last one, so seeking from one song to another one works - currently seeking works only in current song
[15:23:32 CEST] <durandal_1707> atomnuker1: with your patch ffplay can _not_ seek in chained ogg+opus
[15:33:33 CEST] <durandal_1707> Lynne: can you please show yours patch?
[21:08:39 CEST] <durandal_1707> ubitux: so how to proceed regarding lut3d's LUT size?
[21:12:07 CEST] <jamrial> durandal_1707: can you look at my scpr3 patch?
[21:12:24 CEST] <durandal_1707> jamrial: i consider such changes very trivial
[21:13:25 CEST] <jamrial> they are, but when it's not my code i usually prefer to not push without a review
[21:14:31 CEST] <durandal_1707> is there player that can seek without issues in chained ogg?
[21:15:39 CEST] <jamrial> foobar2000 i think can
[21:15:55 CEST] <cone-252> ffmpeg 03James Almer 07master:938cb783d40a: avcodec/scpr3: add missing check for decode_value3() return value
[21:34:06 CEST] <ubitux> durandal_1707: you can go for it if you think it's fine
[21:35:27 CEST] <durandal_1707> i doubt that embeded users do 3D LUT
[21:37:29 CEST] <ubitux> don't tell me what to do with my boards ;)
[21:38:17 CEST] <durandal_1707> video of what size VGA or QCIF?
[21:39:41 CEST] <durandal_1707> also i think doing indexes manually would give same code as it is currently?
[22:25:29 CEST] <ubitux> i don't plan to use it, it was half of a joke
[22:25:45 CEST] <ubitux> i believe gcc will do better offseting than us
[22:25:50 CEST] <ubitux> but feel free to check
[22:41:02 CEST] <cone-252> ffmpeg 03Michael Niedermayer 07master:9fc1031ac2e8: avcodec/pnm_parser: Remember the size of the image and do not reparse the header
[22:41:03 CEST] <cone-252> ffmpeg 03Michael Niedermayer 07master:801939588999: avcodec/pnm_parser: Remember the length already scanned for ascii images
[00:00:00 CEST] --- Sun Apr 28 2019