[Ffmpeg-devel-irc] ffmpeg-devel.log.20170920

burek burek021 at gmail.com
Thu Sep 21 03:05:02 EEST 2017

[00:04:07 CEST] <jkqxz> Is fate-lavf-mxf_dvcpro50 meant to be failing?
[00:11:55 CEST] <jamrial> jkqxz: not really
[00:13:48 CEST] <jkqxz> It does for me, but no idea why at all.
[00:13:52 CEST] <jkqxz> I've replied to the push on the ML to ask.
[00:14:22 CEST] <jamrial> jkqxz: doesn't fail for me on mingw-w64
[00:14:32 CEST] <jamrial> it's a new test using the internal encoder and muxer
[00:16:26 CEST] <cone-214> ffmpeg 03Mark Thompson 07master:375cf55fe9f5: vaapi: Disable deprecation warnings around use of old struct vaapi_context
[00:16:27 CEST] <cone-214> ffmpeg 03Mark Thompson 07master:22afa87a8e90: kmsgrab: Fix DRM format definitions
[00:16:28 CEST] <cone-214> ffmpeg 03Mark Thompson 07master:9f9625ea80ac: kmsgrab: Remove multiple-plane formats
[00:16:29 CEST] <cone-214> ffmpeg 03Mark Thompson 07master:b0ece2b31fdf: hwcontext_vaapi: Fix DRM format mapping
[00:16:30 CEST] <cone-214> ffmpeg 03Carl Eugen Hoyos 07master:f952edaa73ee: kmsgrab: Add more DRM plane formats
[00:16:48 CEST] <jkqxz> Submitted fate results show it working everywhere else, too.
[00:17:39 CEST] <jkqxz> Let me do a clean build and try again.  Cosmic rays seem unlikely, but I guess they're possible.
[00:18:36 CEST] <jamrial> jkqxz: https://0x0.st/7F_.mxf that's the "good" mxf output from the test
[00:18:56 CEST] <jamrial> in case you want to compare it with what it generates in your case
[00:19:53 CEST] <jamrial> i bet it's a byte or two somewhere in the header
[00:21:14 CEST] <jamrial> jkqxz: also, want to check my hwcontext_dxva2 patch?
[00:23:10 CEST] <jkqxz> Yep, looks like it <http://sprunge.us/hONZ>.
[00:25:37 CEST] <jamrial> are you trying in some local branch or in a clean master?
[00:25:55 CEST] <jamrial> it's really weird no fate client fails
[00:26:19 CEST] <jkqxz> That's clean master now.
[00:27:15 CEST] <jkqxz> Replied to the DXVA2 patch.
[00:28:34 CEST] <jamrial> look at the file's metadata. your mxf file seems to have a modification_date tag the "good" output doesn't
[00:30:14 CEST] <jkqxz> It passes on my laptop.  Wtf.
[00:31:32 CEST] <cone-214> ffmpeg 03James Almer 07master:18516d3e6959: avutil/hwcontext_dxva2: return an error when buffer allocation fails
[00:56:48 CEST] <jkqxz> Right, shared libraries with the same version got me.
[00:57:57 CEST] <cone-214> ffmpeg 03Carl Eugen Hoyos 07master:5cb70381d995: lavd/kmsgrab: Remove the mapping for AV_PIX_FMT_RGB8.
[00:57:58 CEST] <cone-214> ffmpeg 03Carl Eugen Hoyos 07master:2f3a3a7e3270: lavf/utils: Do not force chapter end time before chapter start.
[00:57:59 CEST] <iive> can we use DRI instead of DRM ?
[00:58:03 CEST] <jamrial> a reminder to always test with static libraries :p
[00:59:14 CEST] <jkqxz> Yeah.  Normally you get a big "oh noes you fucked up!" message when you do that, but if they are exactly the same version then it doesn't notice.
[01:00:50 CEST] <jamrial> debian stable is 3.2 it seems. I'd say a lot more tests should fail with a git master fate suite
[01:00:51 CEST] <jkqxz> iive:  What do you want to do with DRI?
[01:01:22 CEST] <iive> just use that name
[01:01:48 CEST] <jkqxz> Yeah.  It was a copy I installed a day or two ago, so it was pretty much right except for that one test.
[01:02:10 CEST] <iive> drm have a more popular meaning, that is still relevant in multimedia context.
[01:02:28 CEST] <jkqxz> A meaning which should be subverted as much as possible.
[01:02:44 CEST] <iive> that's not going to happen
[01:02:56 CEST] <iive> you are just going to spread confusion.
[01:03:24 CEST] <jkqxz> This is definitely DRM though, using libdrm and whatnot.  You could kindof talk about DRI around it, but I would generally only use that to refer to the X11 protocols.
[01:04:38 CEST] <jkqxz> I did write that frame-grabber originally on DRI2 rather than KMS, but it was much less useful in that form.
[01:06:11 CEST] <iive> they did that mistake long ago... they are stuck with it. we can avoid it.
[01:06:30 CEST] <iive> if you want you can use KMS instead...
[01:11:07 CEST] <cone-214> ffmpeg 03Carl Eugen Hoyos 07master:b4b02477bd80: lavfi/stereo3d: Set SAR for every output frame.
[01:13:28 CEST] <jkqxz> KMS is only one part of the DRM API.
[02:14:40 CEST] <iive> jkqxz: yes, it is. it stands for kernel model setting and does little more than that. Yet it also signifies major API change.
[02:15:52 CEST] <iive> the X does have modesetting DDX (2D driver), that can actually do a lot more than software framebufer. it does use glamor in order to provide 2D acceleration though opengl-like api.
[02:16:59 CEST] <iive> so.. there will be no confusion if you use KMS.
[02:19:43 CEST] <jkqxz> But I can only use KMS to refer to things which actually use KMS, such as kmsgrab.  And that still returns DRM objects.
[02:21:30 CEST] <cone-214> ffmpeg 03Carl Eugen Hoyos 07master:ca72cd137d3b: lavf/mpegts: Consider stream_type 0x0f just a hint towards AAC.
[02:46:31 CEST] <wm4> I guess I acked a patch that was already merged
[02:49:07 CEST] <iive> well, there used to be a custom, to send reply saying something has been committed, in order to avoid such confusion.
[02:49:32 CEST] <iive> and no harm done. it would have been a problem if you rejected it ;)
[03:01:11 CEST] <cone-214> ffmpeg 03Michael Niedermayer 07master:5480e82d7777: avcodec/pngdec: Clean up on av_frame_ref() failure
[03:01:12 CEST] <cone-214> ffmpeg 03Thomas Mundt 07master:9777836670ae: MAINTAINERS: add myself for bwdif and (t)interlace
[03:01:13 CEST] <cone-214> ffmpeg 03Kaustubh Raste 07master:f4ba85dc82f5: avcodec/mips: Fixed rnd_val variable to 6 in hevc uni mc msa functions
[03:01:14 CEST] <cone-214> ffmpeg 03Kaustubh Raste 07master:e428e5ded626: avcodec/mips: preload data in hevc sao edge 0 degree filter msa functions
[03:34:32 CEST] <cone-214> ffmpeg 03Carl Eugen Hoyos 07master:b4093e60c51a: lavf/caf: Support demuxing Opus.
[04:08:23 CEST] <cone-214> ffmpeg 03Anton Khirnov 07master:5a3b602acda6: avformat/cafdec: reject multichannel Opus streams
[14:10:41 CEST] <bove> where in the source can I find the bitstream filters?
[14:12:54 CEST] <JEEB> bove: find . -iname '*bsf*.c'
[14:13:03 CEST] <JEEB> they're all in libavcodec as far as I see
[14:13:17 CEST] <bove> thanks
[16:34:59 CEST] <kierank> atomnuker: meet at RO at 7?
[16:40:05 CEST] <bjorn__> Hey all! Im working on this: https://trac.ffmpeg.org/ticket/4443
[16:40:15 CEST] <bjorn__> Ive got it working so that I can generate a transparent gif just fine  but only a single frame.
[16:40:28 CEST] <bjorn__> Multiple frames cause problems with the gif encodings oprtimizations, so I cant successfully create an animated gif with transparencies.
[16:44:17 CEST] <bjorn__> I am pretty familiar with teh Gif format, but Im confused about how FFMPeg handles incoming transparencies.
[16:48:37 CEST] <bjorn__> Is libavcodec/gif.c even designed to handle this situation?
[18:18:31 CEST] <atomnuker> kierank: yeah, I'll be there
[18:51:45 CEST] <atomnuker> kierank: sorry, can't go, got a meeting at 19:00 and I need to sleep right after it (messed up sleeping cycle for a month now)
[18:52:54 CEST] <atomnuker> also got to finish my slides, send a patch for the opus psy thing and if I find the time eat
[18:53:03 CEST] <atomnuker> we'll drink in paris on friday
[18:53:13 CEST] <kierank> No Louie though
[18:53:36 CEST] <kierank> It was meant to be louie's leaving party
[19:11:59 CEST] <kierank> atomnuker: are you going to Disneyland
[19:16:39 CEST] <atomnuker> no
[21:46:08 CEST] <durandal_170> comming to disneyland with guns
[21:50:02 CEST] <JEEB> no, you just have fun there
[21:52:28 CEST] <Compn> nerf needs to do a nerf gun theme park
[22:37:14 CEST] <gnafu> Compn: Wow, why haven't they already?
[22:37:28 CEST] <gnafu> As soon as I read that...
[22:37:32 CEST] <gnafu> Never thought of it before.
[23:51:40 CEST] <Compn> gnafu : i am an expert consultant, you can have my idea for $500k
[23:53:01 CEST] <j-b> you are cheap
[23:53:27 CEST] <Compn> i was going to say $1m but i know gnafu does not have this on hand.
[23:53:47 CEST] <wm4> regarding VDD, will there be a meeting between ffmpeg and libav
[23:54:08 CEST] <JEEB> I don't think many from FFmpeg will be going
[23:54:13 CEST] <Compn> there have been meetings at previous VDD for this
[23:54:53 CEST] <iive> haven't there been meeting almost every year?
[23:55:05 CEST] <iive> s
[23:55:12 CEST] <JEEB> I think after that one co-meeting there was mostly just project-specific meetings
[23:55:27 CEST] <Compn> iive : yes
[23:55:42 CEST] <Compn> there are also project specific meetings.
[23:56:20 CEST] <Compn> is this the year j-b will lay down the new plan? :P
[23:56:25 CEST] <JEEB> lol
[23:56:42 CEST] <JEEB> "everyone sucks, I'm going home" I guess ;)
[23:56:44 CEST] <j-b> nope
[23:56:46 CEST] <iive> wan't the new plan like 2 years ago?
[23:57:02 CEST] <j-b> yeah, I offered some ideas and help
[23:57:06 CEST] <j-b> my life has moved on
[23:57:17 CEST] <JEEB> makes sense
[23:57:17 CEST] <Compn> j-b : good , no reason to stress over this nonsense :)
[23:57:42 CEST] <j-b> I'll make more money and have less pain than fight for other people to get rich
[23:57:47 CEST] <Compn> i also offered help and ideas. but i also gave up :)
[23:59:00 CEST] <Compn> https://www.youtube.com/watch?v=91edAkvQCxI
[23:59:09 CEST] <Compn> australia solved the piracy problem!
[23:59:39 CEST] <wm4> huh so all reunification ideas are dead
[23:59:48 CEST] <wm4> and just merging libav commits is also dead
[00:00:00 CEST] --- Thu Sep 21 2017

More information about the Ffmpeg-devel-irc mailing list