[Ffmpeg-devel-irc] ffmpeg-devel.log.20150520

[burek]

[01:37:04 CEST] <cone-748> ffmpeg 03Michael Niedermayer 07master:b8e7f2b277b8: avcodec/flacenc: Support Multi dimensional quantization
[01:43:23 CEST] <jamrial> michaelni: some basic documentation to explain what these new flacenc options do would be nice
[02:25:32 CEST] <Compn> 17:30] <cone-748> vlc Jean-Baptiste Kempf src/misc/fourcc.c: Add wIVG as MJPEG fourcc
[02:25:32 CEST] <Compn> [17:30] <cone-748> vlc Jean-Baptiste Kempf src/misc/fourcc.c: Add wMP4 fourCC for Mp4v
[02:25:35 CEST] <Compn> ugh
[02:31:58 CEST] <michaelni> jamrial, patch posted
[04:15:44 CEST] <BtbN> Let's see if this gets some movement in the nvenc license issue.
[04:29:41 CEST] <jamrial> http://pastebin.com/pypUA8ty that first line from the output is kinda annoying
[05:36:06 CEST] <cone-696> ffmpeg 03Niklesh 07master:b44a55ad2d18: Fix movtext crashes caused due to lack of proper bounds checking
[07:27:22 CEST] <jamrial> ah, nevermind, they removed that and other warnings in a commit right before tagging the 1.7 release
[11:15:05 CEST] <kierank> Anonymous 573462
[11:15:07 CEST] <kierank> interesting
[12:02:26 CEST] <durandal_1707> What?
[12:11:48 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:886ba9300903: doc/encoders: add flac
[12:22:24 CEST] <kierank> durandal_1707: bizzare email on the ml
[12:23:35 CEST] <wm4> "Sent from my iPhone"
[12:48:38 CEST] <cone-217> ffmpeg 03hSÇ 07master:0d41f1ffbcae: configure: disable mipsfpu for loongson3
[13:35:04 CEST] <wm4> is there any way to detect PTS discontinuities with ffmpeg in a generic way?
[13:35:57 CEST] <wm4> if the PTS jumps, it could be a discontinuity (often in .ts etc.), or some freaky VFR video (often in mp4 or mkv)
[13:36:33 CEST] <nevcairiel> i build some code for that, and i only enable it when the format has the flag set that it has discontinuities
[13:36:48 CEST] <nevcairiel> works fine mostly
[13:37:19 CEST] <nevcairiel> ie. AVInputFormat->flags & AVFMT_TS_DISCONT
[13:37:42 CEST] <nevcairiel> as generic as it gets, i guess
[13:38:01 CEST] <wm4> sounds like a potentially good idea
[13:38:13 CEST] <wm4> ideally ffmpeg would have a per packet or frame flag for this, though
[13:39:15 CEST] <wm4> how exactly do you detect it?
[13:42:03 CEST] <nevcairiel> just checking the difference in timestamps against an arbitrary constant
[13:42:08 CEST] <nevcairiel> i think i use 5 seconds right now
[13:43:11 CEST] <wm4> I have a ts stream with a 600ms discontinuity here
[13:43:37 CEST] <nevcairiel> i would probably just let those be, its not that long and wouldn't be extremely disruptive in playback
[13:44:20 CEST] <nevcairiel> since i do the checking before decoding, i do it on re-ordered pts, so have to be careful not to automatically trip on pts offsets
[13:44:33 CEST] <Compn> nevcairiel : obviously its messing up his beautiful HQ HD mpv opengl-hq playback :P
[13:44:37 CEST] <wm4> ah, I do it afterwards
[13:44:47 CEST] Action: Compn read some 4chan thread on mpv lol
[13:44:55 CEST] <Compn> those people are crazy.
[13:56:48 CEST] <ubitux> yes, fortunately here we are all very sane
[14:10:45 CEST] <thardin>  /g/ seems to have some very particular tastes when it comes to media players
[14:13:07 CEST] <wm4> it seems they like the madvr features (scaling etc.)
[14:13:28 CEST] <wm4> and also at least 1 mpv user and occasional contributor has been very active there
[14:13:37 CEST] <wm4> I could think of better user bases...
[14:48:49 CEST] <rcombs> http://i.4cdn.org/g/1432027327280.png kek
[14:58:04 CEST] <wm4> whatever
[15:00:46 CEST] <cone-217> ffmpeg 03Carl Eugen Hoyos 07master:eb0c08bc6938: lavc/vmnc: Fix 32bit colour-space, there is no transparency.
[15:03:19 CEST] <rcombs> well that was a good laugh, at least
[15:05:51 CEST] <Compn> what noise are they making?
[15:06:04 CEST] <Compn> my hirigana /katakana is weak.
[15:06:24 CEST] <Compn> as in i forgot everything i could about the japanese language.
[15:07:46 CEST] <rcombs> "amuamuamuamu"
[15:08:05 CEST] <Compn> huh
[15:08:17 CEST] <Compn> oh no there is a whole dedicated weekly thread to mpv , http://boards.4chan.org/g/thread/48052967/mpv-general
[15:09:12 CEST] Action: Compn ceases trolling and runs from the internet
[15:09:37 CEST] Action: rcombs gives chase
[15:09:46 CEST] <rcombs> oh no sunlight
[15:09:53 CEST] Action: rcombs steps back
[15:14:06 CEST] <ubitux> http://i.4cdn.org/g/1432120410148.png hey a useful post
[15:14:54 CEST] <Compn> wow
[15:14:59 CEST] <Compn> wonder if mplayer could steal that
[15:15:56 CEST] <wm4> ubitux: and of course I made this
[15:16:07 CEST] <ubitux> wm4: ah?
[15:16:07 CEST] <wm4> plus it sucks because it's hand-positioned
[15:16:31 CEST] <wm4> maybe I should give it more visibility, but I kind of hoped someone would make a better chart
[15:17:13 CEST] <ubitux> is it in the html man?
[15:18:25 CEST] <wm4> no
[16:24:41 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:50965e067371: avcodec: add mathops test
[16:24:42 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:611e1085f8d2: fate: Add fate-mathops test
[17:27:11 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:6bbb2f8f4da6: avformat/nutdec: Return error on EOF from get_str()
[17:49:24 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:b3496b4a33e8: avformat/nutdec: Fix recovery when immedeately after seeking a failure happens
[17:58:51 CEST] <cone-217> ffmpeg 03hSÇ 07master:f9d05786a82f: configure: add cpuflags for loongson3 series cpu
[18:22:45 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:6d1935d1b91c: avcodec/hevc: Remove unused variable
[19:01:51 CEST] <cone-217> ffmpeg 03Vittorio Giovara 07master:eafbc6716ced: vmnc: Delay pixel size check
[19:01:52 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:a4c7aeaf82de: Merge commit 'eafbc6716cede6d4a652f8bedf82f2901c68d06d'
[19:01:53 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:b6ca7bfc7c12: avcodec/vmnc: Simplify "24bit" support
[19:10:43 CEST] <cone-217> ffmpeg 03Andreas Cadhalpun 07master:37e679881d36: nutdec: fix infinite resync loops
[19:10:44 CEST] <cone-217> ffmpeg 03Andreas Cadhalpun 07master:fa7dec8cb00d: nutdec: stop skipping bytes at EOF
[19:23:57 CEST] <cone-217> ffmpeg 03Vittorio Giovara 07master:7046bd9bc9ce: lavfi: Move avcodec header to the only filter needing it
[19:23:58 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:6b783d2acea0: Merge commit '7046bd9bc9ce04521edf453c9b603d84d69c7512'
[19:27:10 CEST] <llogan> michaelni: you always beat me to those security updates
[19:27:40 CEST] <llogan> but this time of year i'm at the computer a lot less
[19:32:31 CEST] <cone-217> ffmpeg 03Vittorio Giovara 07master:265d884eb076: avcodec/vmnc: Add back adapted comment about 24bit
[19:55:13 CEST] <llogan> in git web how do i get a gitdiff of a file between two point releases? i used to know this...
[19:55:25 CEST] <llogan> or at least i imagine i did
[20:09:55 CEST] <michaelni> llogan, on github: https://github.com/FFmpeg/FFmpeg/compare/n1.2.3...n2.3
[20:11:31 CEST] <jamrial> llogan: also something like https://git.videolan.org/?p=ffmpeg.git;a=blobdiff;f=Changelog;hb=n2.6.3;hpb=n2.6.2
[20:12:27 CEST] <jamrial> just replace the file and point release tags with whatever you need
[20:13:49 CEST] <llogan> jamrial: thanks
[20:14:59 CEST] <llogan> better than the lazyass changelogs i was putting in tweets
[20:19:59 CEST] <llogan> jamrial: regarding #4448, my opinion is that your patch or a version of it should be committed, at least until/if a proper AVPacketList API is implemented. it would be better than the current situation.
[20:45:07 CEST] <cone-217> ffmpeg 03Andreas Cadhalpun 07master:bb23a15df507: nutdec: abort if EOF is reached in decode_info_header/read_sm_data
[21:47:12 CEST] <cone-217> ffmpeg 03wm4 07master:3efe0393e4f0: hevc: make avcodec_decode_video2() fail if get_format() fails
[21:47:13 CEST] <cone-217> ffmpeg 03Michael Niedermayer 07master:642355712b97: Merge commit '3efe0393e4f0eaa27bac8833edbd7d3cff4704e1'
[21:53:48 CEST] <Compn> llogan : want to post this to twitter/facebook? http://download.das-werkstatt.com/pb/fsfe/presentations/2015/the_ffv1_story.pdf
[21:54:02 CEST] <Compn> if its not already there
[21:58:01 CEST] <Daemon404> some of that slide doesnt really give with me (like how "it's clear" to use uncompress audio (no))
[22:00:18 CEST] <Compn> i cant remember any containers that work (well) with uncompressed audio
[22:00:22 CEST] <Compn> maybe mxf
[22:00:48 CEST] <Compn> otherwise you run into >interleaving
[22:06:26 CEST] <llogan> Compn: i'll try to take a look in a day or two, but i'm not the facebook overlord
[22:06:33 CEST] <wm4> Compn: that's probably because ffmpeg is retarded
[22:06:42 CEST] <wm4> AFAIK ffmpeg may attempt to write audio frames as large as a second
[22:06:57 CEST] <wm4> simply because PCM kind of allows arbitrary frame sizes
[22:09:04 CEST] <Daemon404> well i also meant, for archiving, having a format with checksums is kind of important
[22:12:56 CEST] <nevcairiel> wouldnt a lossless compression still be beneficial
[22:13:14 CEST] <nevcairiel> you usually get some sort of integrity control on a compression level, and its smaller..? :)
[22:19:27 CEST] <rcombs> wm4: OK, so how would you recommend breaking up tls.c
[22:19:43 CEST] <nevcairiel> carefully
[22:20:14 CEST] <nevcairiel> without looking at it, the obvious answer is common parts into one file, per-implementation parts into another
[22:20:20 CEST] <nevcairiel> :)
[22:22:22 CEST] <rcombs> well yeah
[22:22:26 CEST] <wm4> rcombs: the only actually shared code is the URL extraction code
[22:22:32 CEST] <wm4> which deals with proxies or something
[22:22:41 CEST] <wm4> actually I wanted to split it myself
[22:22:46 CEST] <wm4> I could do it tomorrow
[22:23:01 CEST] <wm4> except Libav will then require like a week to merge it
[22:23:19 CEST] <nevcairiel> you're being optimistic
[22:40:26 CEST] <rcombs> wm4: just sent new versions of my TLS patches; would appreciate if any splitting work was either done on top of them or in a manner that doesn't introduce too many conflicts
[22:42:26 CEST] <wm4> you can also have the patch merged quickly, and then michaelni needs to deal with the splitting patch incoming from Libav
[22:42:31 CEST] <rcombs> (changes fix a couple error reporting issues and problems in the configure script)
[22:42:38 CEST] <wm4> though that would be not so nice to him
[22:43:08 CEST] Action: rcombs looks at michaelni 
[22:43:17 CEST] <wm4> also that ifdef soup is really absurd
[22:43:33 CEST] <rcombs> I blame the previous authors
[22:43:35 CEST] <wm4> this is NOT how it should be done, but also shows nicely how it happens despite all contributors having the best intentions
[22:46:25 CEST] <Daemon404> wm4, for tls.c, there's not much of a better way, aside form makign a whole new "sub-api" for tls
[22:46:34 CEST] <Daemon404> tls libs = shit
[22:46:52 CEST] <wm4> the shared code is minimal
[22:47:12 CEST] <wm4> and most shared code is due to a bad idea: trying to make the TLS libs read/write from the sockets
[22:47:24 CEST] <wm4> while making them non-blocking or so, and compensating by using this poll thing
[22:47:25 CEST] <Daemon404> patches welcome to use curl
[22:47:54 CEST] <wm4> are you sure about "welcome"?
[22:48:10 CEST] <Daemon404> from me
[22:48:12 CEST] <rcombs> note how Secure Transport handles this (simple callbacks)
[22:48:17 CEST] Action: Daemon404 runs
[22:48:30 CEST] <wm4> rcombs: the other libs also provide callbacks
[22:49:08 CEST] <nevcairiel> SChannel works on a higher level, if i ever get to implement that ... it just works with data buffers and leaves you to talk to the socket
[22:49:17 CEST] <nevcairiel> or is that a lower levle
[22:49:20 CEST] <nevcairiel> i cant tell anymore
[22:51:15 CEST] <rcombs> what I mean is that Secure Transport has its read and write functions, and it calls the consumer's (here, lavf's) lower-level read and write functions as needed
[22:51:32 CEST] <rcombs> it meshes super-nicely with lavf's interface
[22:59:46 CEST] <wm4> rcombs: why poll?
[23:00:36 CEST] <rcombs> hmm?
[23:01:11 CEST] <wm4> why does it call do_tls_poll()?
[23:03:26 CEST] <rcombs> because that's where all the error handling code is, for some reason
[23:04:17 CEST] <wm4> that's some prime ifdeffery then
[23:06:54 CEST] <rcombs> AFAIK tcp.c will never give tls.c an EAGAIN anyway
[23:07:13 CEST] <rcombs> because it seems we never open the TCP connection in non-blocking mode
[23:08:08 CEST] <wm4> yes, but avio stuff already handles this
[23:08:21 CEST] <wm4> demuxers or e.g. the http protocol don't get EAGAIN either
[23:08:44 CEST] <rcombs> my point being that it's pointless to do work to handle that case
[23:19:44 CEST] <michaelni> rcombs, the tls patch breaks build on mingw64
[23:19:59 CEST] <michaelni> libavformat/tls.c:59:31: fatal error: Security/Security.h: No such file or directory
[23:22:28 CEST] <rcombs> michaelni: oops, typo, sec
[23:24:20 CEST] <michaelni> rcombs, also [[ ]] is not POSIX according to http://mywiki.wooledge.org/Bashism
[23:24:35 CEST] <Daemon404> correct
[23:30:36 CEST] <rcombs> michaelni: corrected
[23:37:00 CEST] <rcombs> in which some rando opens a track ticket to say "please merge rcombs's patch"
[23:37:23 CEST] <rcombs> that's a new one
[23:38:16 CEST] <wm4> lol
[23:39:46 CEST] <michaelni> rcombs,  ./configure --enable-openssl && make -j12  ---> libavformat/tls.c:204:50: error: unknown type name CFArrayRef
[23:40:14 CEST] <michaelni> thats on linux x86 ubuntu
[23:40:37 CEST] <michaelni> x86-64 to be precisse
[23:42:56 CEST] <wm4> lol ifdeffery
[23:43:02 CEST] <rcombs> michaelni: did CONFIG_SECURETRANSPORT get defined as 1 in config.h?
[23:43:24 CEST] <michaelni> config.h:#define CONFIG_SECURETRANSPORT 1
[23:43:38 CEST] <rcombs> wm4: less ifdeffery and more rcombs can't into bash
[23:44:19 CEST] <rcombs> or I'm missing some subtlety of ffmpeg's configure
[23:44:41 CEST] <michaelni> same failure with --enable-gnutls
[23:45:46 CEST] <rcombs> (and technically none of these are ifdefs but that does roll off the keyboard nicely)
[23:47:21 CEST] <rcombs> something something bash operator precedence
[23:49:25 CEST] <rcombs> michaelni: https://gist.github.com/23892ee2beb94ac2e556 <-- want to try this before I squash and send
[23:49:52 CEST] <rcombs> tested in more detail; looks good here on OS X, at least
[23:54:23 CEST] <rcombs> "and now the configure autodetection works properly" <-- unless you're not on OS X, lol
[23:58:49 CEST] <michaelni> rcombs, compile succeeds here
[23:58:57 CEST] <rcombs> whooo
[00:00:00 CEST] --- Thu May 21 2015