[Ffmpeg-devel-irc] ffmpeg-devel.log.20160228

[burek]

[01:37:57 CET] <philipl>    
[02:03:48 CET] <cone-833> ffmpeg 03Michael Niedermayer 07master:b70e9b4906a3: avutil/imgutils: Assert that the 2nd av_image_fill_linesizes() call in av_image_fill_linesizes() still succeeds
[03:20:29 CET] <michaelni> j-b, any news about the hwaccel-mt case ? I really would like you to remove the check that prevents ffmpeg from being used with vlc 2.2 or whatever versions are affected. 
[03:21:43 CET] <michaelni> also if there is anything i can do, like rewording the warning or adding a flag to turn the warning off or otherwise, then please say so
[03:32:10 CET] <cone-833> ffmpeg 03Andreas Cadhalpun 07master:15708f13477a: configure: add direct detection of libopencv
[04:26:36 CET] <cone-833> ffmpeg 03Syed Andaleeb Roomy 07master:b4dcd351ec50: movenc: Timecode in MP4 Although MP4 does not have a concrete specification to store timecode information, the following technical note from Apple describes a way to achieve this via timecode track, similar to how it is done for MOV files.
[04:26:37 CET] <cone-833> ffmpeg 03Michael Niedermayer 07master:5d18dc379539: tests/lavf-regression: Add mp4 timecode test
[05:33:50 CET] <Timothy_Gu> michaelni: next time you apply a patch could you please make sure that the commit message is doesn't contain such a long line? thanks
[09:31:36 CET] <funman> michaelni: i am looking if i can test vdpau hwaccel iwth vlc
[10:03:21 CET] <funman> http://git.videolan.org/?p=vlc.git;a=commitdiff;h=b8869f97ea66ac7ec9912a74c2e8b5e15daf7752;hp=bfca1ad036f4740b2a01fe0e64541c604c119449
[12:09:15 CET] <michaelni> funman, thanks alot!
[13:37:58 CET] <cone-054> ffmpeg 03Reimar Döffinger 07master:0f199f0ad01e: mss2: Fix buffer overflow.
[13:37:58 CET] <cone-054> ffmpeg 03Reimar Döffinger 07master:45fa03b1f9b0: mjpegdec: Do not assume unused plane pointer are NULL.
[13:37:58 CET] <cone-054> ffmpeg 03Reimar Döffinger 07master:4dd4d5353130: Document and validate AVFrame plane pointers.
[14:10:47 CET] <kierank> atomnuker: ping
[14:11:19 CET] <atomnuker> kierank: pong?
[14:11:37 CET] <atomnuker> what's up?
[14:19:40 CET] <kierank2> kieran at ubuntu:~/mpv-build$ ./mpv/build/mpv ~/ffmpeg/trailer.avi -lavfi-complex='[aid1] asplit [t1] [ao] ; [t1] showvolume [t2] ; [vid1] [t2] overlay [vo]'
[14:19:59 CET] <kierank2> I get
[14:19:59 CET] <kierank2> Error parsing option lavfi-complex (option not found) Setting command line option '--lavfi-complex=[aid1] asplit [t1] [ao] ; [t1] showvolume [t2] ; [vid1] [t2] overlay [vo]' failed.
[14:21:04 CET] <durandal_170> what mpv version is that?
[14:21:14 CET] <durandal_170> old release?
[14:21:33 CET] <kierank2> mpv 0.15.0-git-d1d6257 (C) 2000-2016 mpv/MPlayer/mplayer2 projects  built on Sun Feb 28 13:09:14 GMT 2016
[14:21:53 CET] <kierank> mpv 0.15.0-git-d1d6257 (C) 2000-2016 mpv/MPlayer/mplayer2 projects
[14:21:53 CET] <kierank>  built on Sun Feb 28 13:09:14 GMT 2016
[14:22:12 CET] <durandal_170> i guess you need git one, latest version
[14:29:19 CET] <kierank> ok it works with git
[14:44:43 CET] <cone-054> ffmpeg 03Roman Ryltsov 07master:af2568a675a7: avformat/Makefile: Fixed missing rawutils.o reference (required by movenc.c)
[14:52:58 CET] <Daemon404> michaelni, https://github.com/golang/go/issues/14522 this image decodes wrong in ffmpeg too
[14:53:03 CET] <Daemon404> in case you are interested
[14:53:24 CET] <Daemon404> im not sure if its the jpeg's fault, or if it uses some obscure feature only libjpeg and ms's jpeg decoder can decode (and maybe apple's)
[14:58:10 CET] <cone-054> ffmpeg 03Derek Buitenhuis 07master:1c9215e580b6: lavf/mp3: Properly check return values of seeks and reads while reading the header
[15:15:00 CET] <kierank> durandal_170: can you make volume like this? https://wiki.videolan.org/File:VLC-visualeffect.png/
[15:19:02 CET] <atomnuker> kierank: showfreqs does more or less that: https://0x0.st/8O0.jpg
[15:19:21 CET] <kierank> I mean with the green and yellow colours
[15:19:24 CET] <kierank> at the moment it's a fixed colour on the bars
[15:28:02 CET] <durandal_170> kierank: why you need that? I guess it just picks some random frequencies to show or it averages them somehow
[15:28:48 CET] <kierank> I mean the colours
[15:29:09 CET] <kierank> maybe I will send a patch to use the actual colours
[15:34:00 CET] <durandal_170> i you mean volume colour to be like gradient?
[15:34:09 CET] <durandal_170> kierank: ^
[16:17:09 CET] <kierank> Yes
[17:19:16 CET] <cone-054> ffmpeg 03Timothy Gu 07master:222e6da605ea: x86/vf_blend: Add SSE2 optimization for divide
[17:20:24 CET] <cone-054> ffmpeg 03Timothy Gu 07master:bbf7500df99c: LICENSE: Thorough editing
[19:01:21 CET] <cone-054> ffmpeg 03Michael Niedermayer 07master:c6f4720b8664: avcodec/mjpegdec: Fix decoding slightly odd progressive jpeg
[19:14:36 CET] <Daemon404> michaelni, what exactly was wrong/odd with the file?
[19:15:45 CET] <michaelni> that not all passes reached Al=0, but i dont think jpeg requires that
[19:16:05 CET] <Daemon404> interesting
[20:05:18 CET] <cone-054> ffmpeg 03Paul B Mahol 07master:2ad1c87bb260: avfilter/vf_vectorscope: add color5 mode, mode like color but with higher saturation
[20:06:43 CET] <cone-054> ffmpeg 03Rostislav Pehlivanov 07master:5cc53c2e530d: vc2enc: cache bits per quantizer, calculate wasted bits
[20:06:44 CET] <cone-054> ffmpeg 03Rostislav Pehlivanov 07master:fc1d3cbfdc66: vc2enc: use 32 bits for quantized coefficients LUT
[20:06:45 CET] <cone-054> ffmpeg 03Rostislav Pehlivanov 07master:0a2adf0f47ce: vc2enc: carry over quantization index across frames as a starting point
[20:06:46 CET] <cone-054> ffmpeg 03Rostislav Pehlivanov 07master:6e5c6d61bdda: vc2enc: clip and warn when user bitrate set too low
[20:06:47 CET] <cone-054> ffmpeg 03Rostislav Pehlivanov 07master:2f19583911eb: 2enc: clip and warn when user bitrate set too low
[20:06:48 CET] <cone-054> ffmpeg 03Rostislav Pehlivanov 07master:bbcd5e99c303: vc2enc: allocate the DWT context with the current plane size
[20:06:49 CET] <cone-054> ffmpeg 03Rostislav Pehlivanov 07master:e7345abe052b: vc2enc: redistribute leftover bytes
[20:08:00 CET] <atomnuker> grr, why does copying commit messages always stip the first 2 chars
[20:09:07 CET] <ubitux> forgot to enter insert mode?
[20:09:55 CET] <ubitux> 'c' seems to trigger insert mode when in visual mode
[20:10:28 CET] <atomnuker> seems I did
[20:10:57 CET] <atomnuker> oh well, it's in the middle of the commits, anyone with basic pattern recognition skills knows what it touches
[20:26:23 CET] <jamrial> atomnuker: make sure to backport the commits that fix crashes or wrong output to the 3.0 branch
[20:31:25 CET] <atomnuker> jamrial: it'll be a bit difficult to do it directly since 2f195839 needs 2 more arguments (NULLs), but I'll try
[20:35:14 CET] <kierank> durandal_170: ping
[20:41:09 CET] <ubitux> michaelni: i have random mb glitches with msmpeg4 video: http://b.pkh.me/msmpeg4-mb-glitch.png
[20:41:20 CET] <ubitux> i don't know if it's an encoding error though
[20:41:52 CET] <ubitux> does it ring a bell or i should open a ticket?
[20:43:06 CET] <michaelni> ubitux, probably encoding error (too low qscale) but i might miss-guess
[20:43:38 CET] <ubitux> how can i check that?
[20:44:02 CET] <atomnuker> looks like the DC coefficient is incorrect
[20:45:03 CET] <michaelni> ubitux, are there any warnings during encoding ?
[20:45:15 CET] <ubitux> i didn't encode the file
[20:45:57 CET] <michaelni> are there any errors during decoding and how does it look with the binary decoder vfw or whatever?
[20:46:00 CET] <ubitux> i uncommented ERROR_DETAILS in libavcodec/msmpeg4dec.c but it didn't raise anything either
[20:46:20 CET] <michaelni> it all sounds like too low QP
[20:46:37 CET] <michaelni> msmpeg4 cannot increase the QP like other codecs can
[20:46:57 CET] <michaelni> so when theres a contrasty edge bad things happen at low QP and intra MBs
[20:47:16 CET] <michaelni> you could check the QP used 
[20:47:21 CET] <michaelni> -debug 1 probably shows it
[20:48:01 CET] <ubitux> qp:2
[20:48:19 CET] <michaelni> thats too low IIRC for some patterns
[20:49:15 CET] <ubitux> ok; it's curious that this file was distributed, because these patterns are kind of common in the whole video
[20:49:31 CET] <ubitux> it often happens in very high contrast part of the video
[20:49:38 CET] <ubitux> like black on white (eyes etc)
[20:49:49 CET] <michaelni> yes thats the too low qp issue you describe
[20:49:56 CET] <ubitux> ok
[20:56:05 CET] <ubitux> http://ubitux.fr/pub/pics/msmpeg4-mb-glitch-0.png http://ubitux.fr/pub/pics/msmpeg4-mb-glitch-1.png http://ubitux.fr/pub/pics/msmpeg4-mb-glitch-2.png http://ubitux.fr/pub/pics/msmpeg4-mb-glitch-3.png
[20:56:26 CET] <durandal_170> kierank: pong, ping durandal_1707 next time, that's my phone
[20:57:26 CET] <kierank> ok
[20:57:30 CET] <durandal_170> wm4: can ass draw random antialiased lines?
[20:57:41 CET] <kierank> durandal_1707: how does synchronisation between video and overlay work in lavfi?
[20:57:46 CET] <kierank> where does the logic happen to match pts
[20:58:12 CET] <durandal_170> kierank: inside framesync.c
[20:58:19 CET] <durandal_170> is there something wrong?
[21:00:11 CET] <durandal_170> framesync will sync to same pts, it may duplicate frames, or keep them in memory until it finds one it needs
[21:00:27 CET] <kierank> that's what I'm worried about
[21:01:00 CET] <kierank> I don't understand how buffering will work
[21:02:02 CET] <durandal_170> framesync itself will drop frames when its increases internal queue
[21:02:38 CET] <kierank> that scares me
[21:03:18 CET] <durandal_170> you should not worry if audio pts match video
[21:03:33 CET] <kierank> I am just doing video + subtitle
[21:03:49 CET] <kierank> but I don't think it'll do what I want
[21:06:20 CET] <kierank> Let's say I have video frame with PTS n and subtitle with PTS n
[21:06:24 CET] <kierank> then I continue with video n+1, n+2 but I don't get a subtitle until n+5
[21:06:37 CET] <kierank> what will it do?
[21:07:30 CET] <durandal_170> you mean with subtitle filter and overlay?
[21:08:00 CET] <kierank> no I have my own bitmaps
[21:08:40 CET] <kierank> the dvbsub decode goes via my own pipeline
[21:10:55 CET] <durandal_170> if you do not give another frame to overlay filter it will wait for bitmap
[21:11:31 CET] <kierank> you have no way of knowing latency
[21:15:04 CET] <durandal_170> by default it will keep up to 64 frames in queue
[21:15:43 CET] <durandal_170> if nothing appears, all frames are dropped iirc
[21:16:11 CET] <durandal_170> until right one appears
[21:28:45 CET] <wm4> durandal_170: ass can draw lines, yes
[21:28:59 CET] <wm4> at least I'm pretty sure it does
[21:29:06 CET] <durandal_170> wm4: function?
[21:29:16 CET] <wm4> durandal_170: why does vf_overlay need a queue???
[21:29:29 CET] <wm4> it doesn't have a line draw function, but a polygon rasterizer
[21:30:54 CET] <durandal_170> wm4: it have queue because framesync have it, you can setup framesync to not do any pts syncing
[21:31:11 CET] <wm4> I mean why would syncing need a queue
[21:31:21 CET] <wm4> you'd only need 2 buffered images per input?
[21:31:34 CET] <wm4> (at most)
[21:32:11 CET] <durandal_170> ask your friend Nicolas :-P
[21:35:10 CET] <kierank> problem is the latency is undefined
[21:35:25 CET] <kierank> so you have no idea when to present your frame
[21:36:23 CET] <kierank> is there any way I can present frames together to libavfilter?
[21:36:29 CET] <kierank> so I can say overlay X onto Y
[21:37:07 CET] <durandal_170> if they have same pts, sure
[21:37:20 CET] <kierank> yes but I have gaps
[21:37:23 CET] <kierank> so how do I stop that
[21:37:37 CET] <kierank> I just want data x overlaid on data y
[21:37:47 CET] <kierank> no buffering, no pts syncing and all that stuff
[21:38:31 CET] <wm4> didn't vf_overlay originally strictly sync to the first input
[21:38:55 CET] <durandal_170> it still does, iirc
[21:39:01 CET] <wm4> so it'd output the frame from input 1, overlayed with whatever happens to be in the other ones
[21:39:41 CET] <kierank> pfftg
[21:39:45 CET] <kierank> might have to write my own overlay then
[21:39:47 CET] <durandal_170> overlay needs 2 frames
[21:39:59 CET] <kierank> yes for no reason
[21:40:05 CET] <kierank> I have data x that needs to overlay on data y
[21:40:08 CET] <kierank> why should I buffer
[21:40:56 CET] <durandal_170> i will look can it be done with current framesync, currently busy writting antialiased lines drawing
[21:41:12 CET] <wm4> frame scheduling could become very complicated if no buffering has to be guaranteed, because how would it know the next frame on input 2 has the required pts, or a higher pts
[21:41:23 CET] <kierank> and that's the problem with vfr
[21:41:49 CET] <wm4> well most vfr things actually come with frame durations
[21:41:54 CET] <wm4> just not ffmpeg in general
[21:42:04 CET] <wm4> which makes vfr much harder
[21:42:23 CET] <durandal_170> lavfi does have frame durations i used it
[21:42:33 CET] <kierank> is there a way to manually push frames to the filter
[21:42:35 CET] <wm4> (libavformat's guessed and wrong frame durations don't make ti easier)
[21:43:21 CET] <durandal_170> av_frame_get_pkt_duration(out)
[21:43:58 CET] <durandal_170> kierank: directly, without buffersrc? no
[21:44:12 CET] <kierank> can I avoid this unnecessary framesync?
[21:44:26 CET] <kierank> it adds latency for no reason
[21:44:42 CET] <kierank> and worse still latency I can't measure
[21:49:08 CET] <durandal_170> try experimenting setting sync to 0 for one of streams in libavfilter/dualinput.c:87
[21:49:29 CET] <durandal_170> actually 59th line
[21:50:06 CET] <kierank> hmmm
[21:51:11 CET] <kierank> I guess I have to write a function that guesses the latency
[21:51:46 CET] <kierank> but I still don't know how to simulate missing teletext
[21:53:28 CET] <kierank> durandal_170: so does it expect ABABABABAB
[21:53:43 CET] <kierank> assuming A and B have same PTS
[21:54:44 CET] <durandal_170> yes, i think so
[21:55:10 CET] Action: rcombs makes stream with one frame in the middle and then nothing for the whole rest of the duration
[21:55:20 CET] <rcombs> (this frame lasts 2 seconds)
[21:55:26 CET] <rcombs> pls make sure you don't fail on that case
[21:55:39 CET] <rcombs> (I've been bitten by that sort of thing before)
[22:01:52 CET] <cone-054> ffmpeg 03Rodger Combs 07master:3617e69d50dd: lavf/mov: fix sidx with edit lists
[22:01:53 CET] <cone-054> ffmpeg 03Rodger Combs 07master:22dbc1caaf13: lavf/mov: downgrade sidx errors to non-fatal warnings; fixes trac #5216
[22:25:01 CET] <kierank> durandal_1707: does vf_showvolume support arbitrary channel maps?
[22:25:14 CET] <kierank> i.e does it work if I just provide 12 channels or whatever?
[22:25:20 CET] <kierank> does it care what the map is?
[22:26:31 CET] <durandal_170> currently it accepts only known channel layouts, but i can change that
[22:27:37 CET] <wm4> can't see a reason why it'd care about the layout...
[22:28:18 CET] <durandal_170> to draw channel names
[22:28:47 CET] <wm4> ah
[22:43:18 CET] <ubitux> so... how do we proceed for the merge of codecpar?
[22:45:30 CET] <kierank> ubitux: you don't let morons like carl ruin the project
[22:45:34 CET] <kierank> and bring derek back
[22:45:53 CET] <ubitux> ?
[22:46:17 CET] <ubitux> Daemon404 is here
[22:50:20 CET] <ubitux> anyway, i'm happy to help if someone drives the merge and we are at least 2 working on the merge itself
[22:50:43 CET] <ubitux> i can allocate time for this, so if anyone is up to...
[22:50:50 CET] <durandal_170> its same as most boring thing ever
[22:52:58 CET] <ubitux> durandal_170: well, there are many fate changes
[22:53:49 CET] <durandal_170> can't it be done with sed?
[22:53:59 CET] <ubitux> wanna try?
[22:55:14 CET] <durandal_170> ubitux: wanna finish nlmeans?
[22:55:20 CET] <ubitux> if you look at the commit, you'll see that the codec->time_base field is replaced by setting the st->avg_frame_rate
[22:55:47 CET] <ubitux> and i wonder how much things this will influence
[22:55:55 CET] <ubitux> especially given that we still have r_framerate
[22:56:02 CET] <ubitux> durandal_170: yes
[22:56:53 CET] <durandal_170> tell others what branch on github we can use?
[22:58:41 CET] <cone-054> ffmpeg 03Raymond Hilseth 07master:20e4863ab14a: avformat/file: enable file_move() without unistd.h
[22:58:42 CET] <cone-054> ffmpeg 03Raymond Hilseth 07master:c66a6369e488: avformat/dashenc: Enable dash output to work when the output isn't a local file
[23:04:55 CET] <cone-054> ffmpeg 03Marton Balint 07master:4840effe4222: avformat/mov: merge mov_read_custom functions
[23:04:56 CET] <cone-054> ffmpeg 03Marton Balint 07master:e22bd239c046: avformat/mov: do not leak memory on ffio_read_size failure
[23:17:27 CET] <cone-054> ffmpeg 03Michael Niedermayer 07master:149250720878: ffmpeg_vdpau: Remove unused variable
[23:18:41 CET] <cone-054> ffmpeg 03Paul B Mahol 07master:65cc3915db66: avfilter/avf_showvolume: support unknown channel layouts too
[23:40:34 CET] <cone-054> ffmpeg 03Michael Niedermayer 07master:3e42c1128fe7: avcodec/libzvbi-teletextdec: Remove unused variable
[23:53:41 CET] <durandal_170> kierank trolling smpte :)
[00:00:00 CET] --- Mon Feb 29 2016