[Ffmpeg-devel-irc] ffmpeg-devel.log.20180930

[burek]

[01:24:08 CEST] <cone-843> ffmpeg 03Marton Balint 07master:b577153231ee: avcodec/libzvbi-teletextdec: add formatted ass output
[02:22:23 CEST] <kierank> atomnuker: I agree with you but I want av1 to succeed and if dav1d being bsd and out of lavc then it's a necessary evil
[03:43:41 CEST] <TD-Linux> also dav1d is nearly faster than libaom already
[03:44:59 CEST] <TD-Linux> don't really see the argument for sticking with libaom longer tbh
[03:45:03 CEST] <TD-Linux> even if you're going to copypaste it eventually
[04:00:08 CEST] <jamrial> TD-Linux: much like with vpx, we're not going to remove the aom decoder wrapper. it can be used for easy decoding comparisons and such
[04:04:44 CEST] <TD-Linux> jamrial, oh yeah for sure
[04:04:58 CEST] <TD-Linux> I just mean vs not merging the dav1d wrapper at all
[04:21:00 CEST] <BBB> yes please do keep the aom wrapper, I use it a fair bit :)
[04:21:05 CEST] <BBB> (for conformance testing)
[04:23:53 CEST] <philipl> BtbN: https://github.com/FFmpeg/nv-codec-headers/pull/3
[13:02:34 CEST] <BBB> nevcairiel: so, technically speaking, if we wanted to hwaccel in dav1d and consider doing that without hard-forking dav1d, possibly even as a library. how hard is it to expose enough details and callbacks in the library so hw decoders can work correctly?
[14:12:02 CEST] <JEEB> hmm, in the API did we have any info on "demuxer" seekability?
[19:25:04 CEST] <j-b> 'morning
[19:27:43 CEST] <durandal_1707> 'evening
[19:28:11 CEST] <j-b> what's up?
[19:29:50 CEST] <jamrial> sunday :p
[19:46:25 CEST] <durandal_1707> j-b: videolan wiki bounty page is unchanged
[19:46:33 CEST] <j-b> indeed
[19:46:39 CEST] <j-b> I'm waiting for koda for samples
[19:47:15 CEST] <durandal_1707> i'm afraid he can not share his samples
[19:58:20 CEST] <JEEB> did we have an annex b probing function somewhere_
[19:58:31 CEST] <JEEB> or which could give you a hint of AVCc vs annex b?
[20:42:21 CEST] <RiCON> jamrial_: libdav1d wrapper doesn't work in configure, in static mingw x86_64
[20:42:49 CEST] <RiCON> though it might be because libdav1d is only 28KB and missing every symbol except dav1d_thread_entrypoint
[20:43:47 CEST] <cone-213> ffmpeg 03Haihao Xiang 07master:4c14a25281bb: lavc/hevc: Don't parse NAL unit for a dummy buffer
[20:43:47 CEST] <cone-213> ffmpeg 03Michael Niedermayer 07master:449b1dcd7d52: avcodec/hevc_parser: very minor simplification in hevc_parse()
[20:51:42 CEST] <durandal_1707> how to debug motion vectors issues in codec? it appears some blocks are done with idct add and other with idct put, and there is no bit to signal it iiuc
[20:55:52 CEST] <RiCON> jamrial_: nevermind, same issue as https://github.com/mesonbuild/meson/pull/3939
[21:10:32 CEST] <cone-213> ffmpeg 03Karthick Jeyapal 07master:c0ee4e0ac2c4: avdevice/decklink_dec: add option to align capture start time
[21:28:13 CEST] <kierank> durandal_1707: which codec
[21:28:51 CEST] <durandal_1707> kierank: very important one
[21:29:15 CEST] <kierank> ah ok I don't work on important stuff
[21:31:06 CEST] <BtbN> philipl, ping me again about it on sunday if I forget. I probably won't have a time until then.
[21:43:41 CEST] <durandal_1707> michaelni: libavcodec/gdv.c:77:9: warning: unused variable 'i' [-Wunused-variable]
[21:43:53 CEST] <durandal_1707> michaelni: please fix what you broke
[22:06:27 CEST] <kierank> durandal_1707: what samples you need
[22:08:05 CEST] <durandal_1707> kierank: IMM4,IMM5,IMM6,RGBB,RFBW,AGM2,AGM3,....
[22:09:29 CEST] <JEEB> AGM I think can be gotten from http://t-ishii.la.coocan.jp/download/AGDRec.html
[22:09:40 CEST] <JEEB> so you can use most likely some vfw thing to create samples
[22:09:44 CEST] <JEEB> like virtualdub
[22:10:16 CEST] <durandal_1707> i cant read japanese, and there are variants tool does not write
[22:10:29 CEST] <JEEB> yea, it's probably the older thing
[22:10:33 CEST] <JEEB> also there's a linux player
[22:10:34 CEST] <JEEB> funky
[22:11:40 CEST] <JEEB> converter http://t-ishii.la.coocan.jp/download/AGMConvU.html , player http://t-ishii.la.coocan.jp/download/AGMPlayerU.html
[22:11:55 CEST] <JEEB> also there is the older and newer version there from 2013 and 2017/8
[22:13:48 CEST] <JEEB> seems like his old ubuntu version required java o_O
[22:25:40 CEST] <cone-213> ffmpeg 03xiaofeng 07master:e78016b04a71: ffmpeg: break loop when dts_heuristic check done
[22:25:41 CEST] <cone-213> ffmpeg 03Michael Niedermayer 07master:2076e11839b3: avcodec/gdv: Remove unused variable
[22:25:41 CEST] <cone-213> ffmpeg 03Michael Niedermayer 07master:a8a98ba9eeed: avcodec/wcmv: Fix integer overflows
[22:25:42 CEST] <cone-213> ffmpeg 03Michael Niedermayer 07master:d70276921348: avcodec/scpr: Skip frames which change nothing
[22:25:58 CEST] <michaelni> durandal_1707, unused variable removed
[22:26:55 CEST] <kierank> durandal_1707: you should learn japanese
[22:36:06 CEST] <durandal_1707> JEEB: there are also AGM0,AGM1,AGM4,AGM5,AGM6,AGM7 dunno how to get them
[22:37:06 CEST] <JEEB> and the player doesn't handle some of those (the new and old versions)
[22:37:08 CEST] <JEEB> ?
[22:37:39 CEST] <durandal_1707> dunno, i do not have files with those numbers
[22:39:23 CEST] <durandal_1707> kierank: you should learn vulkan
[22:39:52 CEST] <kierank> I already speak vulkan
[22:40:59 CEST] <durandal_1707> then write something up
[22:51:13 CEST] <JEEB> oh wow, IVF demuxer uses ff_codec_bmp_tags and if you add a HEVC tag there does it mean we also inadvertedly add support for HEVC in AVI >_<
[22:52:38 CEST] <durandal_1707> multimedia sucks
[22:58:50 CEST] <kierank> JEEB: i would really mention that asap!
[23:03:37 CEST] <JEEB> I almost did but then removed that line because I added an "where's my <INSERT ALCOHOL> bottle now?" reference to it
[23:03:56 CEST] <JEEB> I just hope people notice the mention of ff_codec_bmp_tags
[23:05:58 CEST] <jamrial> JEEB: carl attempted to add hevc to riff.c some time ago, but was told not to
[23:06:25 CEST] <jamrial> easy to solve, in any case. just make ivf use a custom AVCodecTag list
[23:06:30 CEST] <JEEB> yes
[23:09:29 CEST] <JEEB> jamrial: thanks for noting that previous precedent, and thus I added a comment
[23:35:35 CEST] <JEEB> jamrial: I was mostly noting that if there was a helper that was checking for the Annex B start codes that should probably be used instead of having the checks there
[23:35:40 CEST] <JEEB> probably worded myself badly
[23:36:12 CEST] <JEEB> of course I had already seen that movenc/matroskaenc just seem to be doing the same thing, so I started mentioning a helper :P
[23:36:31 CEST] <JEEB> in the hopes that a helper would grow out of this (wink wink, nudge nudge)
[23:40:19 CEST] <jkqxz> Was there an answer as to why H.26[45] in IVF is of any use to anyone?
[23:40:39 CEST] <JEEB> absolutely not
[23:41:18 CEST] <JEEB> I got a private response from the guy since he wasn't on the ML then which just said "At Google we have some old software that expect H264 packed into IVF container. It would be great if FFmpeg supported transmuxing H264 into IVF."
[23:41:30 CEST] <jkqxz> Then why are people only nitpicking the patch rather than asking that?
[23:41:53 CEST] <JEEB> I actually asked that as the very first thing
[23:42:14 CEST] <JEEB> also someone on the ML noted that if Google specified it, we should support it
[23:42:17 CEST] <jkqxz> Yeah, but there wasn't a reply and the discussion has moved on as if that were answered.
[23:43:04 CEST] <jkqxz> Supporting it for demuxing is fine if there are files in the wild with it.  Supporting it for muxing wants a real use-case since it's clearly not a sensible thing to do.
[23:47:18 CEST] <jamrial> jkqxz: ivf is owned by google, so if they define h264/hevc for it, then what can you do?
[23:47:24 CEST] <jamrial> i agree we could just say no to muxing it
[23:47:35 CEST] <jamrial> and only implement it in the demuxer side
[23:50:16 CEST] <jamrial> then again, they seem to care mainly about creating files their own software can then use afterwards, rather than have ffmpeg read files they already have
[23:58:14 CEST] <jkqxz> If that's the case then it should probably be gated behind some additional option, to stop people from accidentally making files that nothing but some google-internal thing can read.
[00:00:00 CEST] --- Mon Oct  1 2018