[Ffmpeg-devel-irc] ffmpeg-devel.log.20151029

burek burek021 at gmail.com
Fri Oct 30 02:05:02 CET 2015


[02:43:40 CET] <cone-093> ffmpeg 03Ganesh Ajjanagadde 07master:fab1562a5068: avutil/ripemd: make rol macro more robust by adding parentheses
[02:45:25 CET] <cone-093> ffmpeg 03Ganesh Ajjanagadde 07master:6dc0db6634f4: avcodec/motion_est_template: replace qsort with AV_QSORT
[03:15:59 CET] <cone-093> ffmpeg 03Ronald S. Bultje 07master:63fca9df9c6f: vp9_parser: don't overwrite cached timestamps with nopts.
[04:07:47 CET] <cone-093> ffmpeg 03Ganesh Ajjanagadde 07master:dd367495578d: avutil/audio_fifo: add av_warn_unused_result
[06:13:08 CET] <cone-093> ffmpeg 03James Almer 07master:fc460fe61847: avfilter/showcqt: fix compilation with libswscale disabled
[09:54:29 CET] <flux> I'm writing an extension to the iso media format to support timed metadata. it seems to fit in relatively well, but the one thing I'm stuck at is that I need to add a track reference.
[09:54:52 CET] <flux> currently it seems the only thing using track references is the RTP hinting and it works sort of automatically, so it doesn't have its own special interface it seems
[09:55:17 CET] <flux> but in this case the user is providing a separate metadata track and needs a way to tell that the track is referencing (say) a certain video track
[09:56:00 CET] <flux> I'm wondering: what would be the nice interface to pass this information? so, basically two integers :-). I would hope to not add a new interface for this, but I suppose if that's the only option then I do it that way..
[10:38:20 CET] <wm4> did ganesh really send thousands of pings
[10:38:40 CET] <nevcairiel> nah just 500, and 500 new oneline patches
[10:39:55 CET] <rcombs> this is why I have a filter setup to remove ffmpeg-devel messages that don't explicitly mention me from my inbox
[10:54:18 CET] <durandal_1707> :(
[11:37:22 CET] <paulrobur> hi
[11:39:11 CET] <durandal_1707> hi
[11:39:31 CET] <paulrobur> do you know mediacoder?
[11:40:22 CET] <durandal_1707> nope
[11:40:44 CET] <paulrobur> how can i convert a wmv into avi?
[11:41:32 CET] <kierank> durandal_1707: interested in fuzzing sliced threads h264?
[11:43:04 CET] <durandal_1707> why h264? Don't you have faster machine...
[11:43:21 CET] <kierank> I am fuzzing frame threads
[11:43:26 CET] <kierank> probably a week left or so
[11:43:57 CET] <kierank> well I think you'll agree h264 is the most important codec in ffmpeg
[11:47:09 CET] <BBB> kierank: did you find anything?
[11:47:25 CET] <BBB> kierank: I did the chrome codecs back when I worked at google, I was pretty sure I found all the bugs already
[11:47:37 CET] <BBB> (frame threads, not slice threads)
[11:47:48 CET] <kierank> I found a few frame threads crashes earlier in the year
[11:47:53 CET] <BBB> :(
[11:48:22 CET] <kierank> and a lot of sliced thread crashes (many of which affected me in the real world)
[11:48:33 CET] <BBB> can someone not-me step in and respond to ganesh comments in all: fix enum definition for large values (and michaels also), I feel like Im repeating myself and were going in circles
[11:48:42 CET] <BBB> I never fuzzed slice threads
[11:55:46 CET] <durandal_1707> google fuzzed ffmpeg with super computers already
[12:06:10 CET] <kierank> yeah, it's why I've found 4-5 crashes in the past week or two of fuzzing
[12:11:54 CET] <cone-868> ffmpeg 03Matthieu Bouron 07master:cbe2dfa4e51b: lavc/pngdec: honor skip_frame option
[12:11:54 CET] <cone-868> ffmpeg 03Matthieu Bouron 07master:ff0dfb5c361a: lavc/mjpegdec: honor skip_frame option
[12:16:03 CET] <michaelni> BBB, no need to reply to me, i dont really have an oppinon on this, i just saw 2 people disagreeing and wanted to throw some other random option in to see if it would be liked by both sides.
[12:28:27 CET] <BBB> ganesh doesnt like reviews, I feel& he takes them too personally
[12:29:23 CET] <kierank> I wonder if he'll move to libav
[12:31:54 CET] <BBB> dont give him ideas
[12:31:55 CET] <BBB> ...
[12:32:18 CET] <BBB> I think he has good energy and we should direct that at something positive
[12:32:30 CET] <BBB> Im not sure why hes working on compiler warnings, its kind of weird to spend so much time at
[12:37:37 CET] <nevcairiel> (especially generating new ones)
[12:40:19 CET] <wm4> BBB: yeah, agreed
[12:40:40 CET] <BBB> I dont actually mind the log10ish work he did
[12:40:50 CET] <BBB> although I thought there were portability concerns
[12:40:55 CET] <BBB> I cant remember for which platforms
[12:41:03 CET] <BBB> I thought it was msvc, but msdn claims it has all of them
[12:41:10 CET] <BBB> at least log10 and log
[12:41:12 CET] <BBB> maybe not log2
[12:41:17 CET] <nevcairiel> BBB: i'm slightly confused re: vp9, how would the parser change make the timebase change?
[12:41:32 CET] <BBB> nevcairiel: the second frame had no timestamp
[12:41:38 CET] <wm4> BBB: have fun checking every single patch, and doing exactly the same for the next 100 patches he's going to send
[12:41:43 CET] <BBB> so if theres only 2 frames and ffmpeg makes up the timestamp
[12:41:51 CET] <BBB> then it can make any timebase to fit the made-up timestamp
[12:41:56 CET] <nevcairiel> i see
[12:42:01 CET] <BBB> I dont know why it does that
[12:42:04 CET] <BBB> Im sure michaelni knows
[12:42:11 CET] <BBB> but the timebase was wrong and is now correct
[12:42:14 CET] <wm4> what timebase change?
[12:42:17 CET] <nevcairiel> the new timebase makes much more sense anyway, so lgtm
[12:42:32 CET] <BBB> log2 does not exist in msvc2010
[12:42:34 CET] <BBB> which we do support
[12:42:36 CET] <BBB> so theres that
[12:42:38 CET] <nevcairiel> we have emulation
[12:42:45 CET] <wm4> msvc2010 can be ignored anyway
[12:42:45 CET] <nevcairiel> i think
[12:42:46 CET] <kierank> could get him to write some simd
[12:42:56 CET] <wm4> i.e. let's drop the shit
[12:43:13 CET] <nevcairiel> yeah libavutil/libm.h defines log2/log2f
[12:43:18 CET] <nevcairiel> if absent
[12:49:59 CET] <michaelni> i do remember compat issues too but theres log2 and log10 in git already so these must have been solved 
[12:53:00 CET] <rcombs> who cares about non-latest msvc
[12:55:31 CET] <cone-868> ffmpeg 03Ronald S. Bultje 07master:c04830382316: vp9: update timestamps in ref files using multiple invisible frames.
[12:55:59 CET] <BBB> nevcairiel: yes we do, I just checked that also
[12:56:02 CET] <BBB> so its fine I guess
[12:56:32 CET] <BBB> wm4: I mean, he wrote it already, if he likes it, I dont mind that particular patchset& 
[12:57:25 CET] <wm4> feel free to override me
[12:57:33 CET] <nevcairiel> i find his "precision" examples funny though, double has a precision of about 15-17 digits, and his values start to differ at 17-18 digits
[12:57:42 CET] <nevcairiel> its sheer luck, really
[12:57:54 CET] <wm4> nevcairiel: he wrote "likely", so he didn't verify it
[12:59:06 CET] <BBB> double is 48 bits precision, so 2^48=~10^3*5, so 15 digits
[12:59:18 CET] <BBB> float is half of that
[12:59:23 CET] <BBB> or so
[12:59:44 CET] <BBB> brb
[12:59:49 CET] <nevcairiel> his "proof" posts are always somewhat half-assed like this
[13:00:23 CET] <nevcairiel> using some sort of statements combined with imprecise language so he can back out if someone calls him on it
[13:04:28 CET] <nevcairiel> like someone said the other day, is "sloppy"
[13:05:53 CET] <wm4> that's why I'm going to nack by default, so he can't push them due to timeout if nobody else reviews
[13:13:44 CET] <durandal_1707> kierank: what happened with overlay simd?
[13:14:13 CET] <kierank> durandal_1707: can't remember, had some minor crisis to deal with then forgot about it
[13:24:51 CET] <Compn> kierank : they probably have fixes they have not shared back with us :P
[13:25:21 CET] <Compn> where they = google
[13:33:45 CET] <durandal_1707> how you can type something like that?
[13:41:58 CET] <kierank> durandal_1707: ?
[13:42:54 CET] <durandal_1707> was directed to Compn
[13:49:33 CET] <kierank> the big G can do no wrong
[13:49:58 CET] <nevcairiel> who was working on improving the fate overview page some time ago?
[13:50:12 CET] <nevcairiel> i would love an option to hide builds from release branches from the main listing
[13:50:44 CET] <nevcairiel> (and while we're at it, black is an odd choice for builds that recently failed)
[13:54:04 CET] <wm4> maybe timothy?
[13:59:06 CET] <Daemon404> it was Timothy_Gu 
[13:59:16 CET] <Daemon404> he rewrote the fate server completely
[13:59:20 CET] <Daemon404> i dont know if it was deployed
[14:00:10 CET] <Daemon404> aside: the amount of replies to these ganesh patches is nuts
[14:00:15 CET] <Daemon404> it's like 50% of my ML traffic
[14:07:07 CET] <cone-868> ffmpeg 03Luca Barbato 07master:d584533cf381: avformat: Rework add_to_pktbuf
[14:07:08 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:54de179caa39: Merge commit 'd584533cf38141172e20bae5436629ee17c8ce50'
[14:17:49 CET] <cone-868> ffmpeg 03Luca Barbato 07master:9b56d5c11488: avpacket: Deprecate av_dup_packet
[14:17:50 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:87a6f532b46f: Merge commit '9b56d5c11488656254e9aed8d964ef2b7c2ff5e6'
[14:20:37 CET] <cone-868> ffmpeg 03Luca Barbato 07master:a9a60106370f: avpacket: Provide an alloc and a free function for the struct
[14:20:38 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:3ec049b85dac: Merge commit 'a9a60106370f862e191dea58e748626da6a8fe97'
[14:22:49 CET] <cone-868> ffmpeg 03Alexandra Hájková 07master:cd0e08813a04: avconv: support infinite loop for the loop option
[14:22:50 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:9f7de9951790: Merge commit 'cd0e08813a0484002b5defbf557c859f123953ae'
[14:25:09 CET] <cone-868> ffmpeg 03Luca Barbato 07master:f3aff31e5f66: avpicture: Deprecate the single fields
[14:25:10 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:7c99384f3c41: Merge commit 'f3aff31e5f66a4f1c4e34ce4944f4a402aca61ed'
[14:26:14 CET] <cone-868> ffmpeg 03Michael Niedermayer 07master:6dcd82290438: opus: Buffer the samples from the correct offset
[14:26:14 CET] <cone-868> ffmpeg 03Kieran Kunhya 07master:b8deb7c34f75: opus: Do not call vector_fmul_scalar on zero samples
[14:26:16 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:f2dc68e2e2dd: Merge commit '6dcd82290438eb810936de4febd0558b38a08ec5'
[14:26:16 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:313f53824ef4: Merge commit 'b8deb7c34f755d5e3eee0b5930c3a6ad2dda96bc'
[14:27:25 CET] <cone-868> ffmpeg 03Martin Storsjö 07master:9e14a9924093: libopenh264enc: Add an option for controlling the frame skipping
[14:27:26 CET] <cone-868> ffmpeg 03Martin Storsjö 07master:8edaf625f3c3: libopenh264enc: Count and warn about the number of skipped frames
[14:27:27 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:f79bb118f589: Merge commit '9e14a992409348630ea96521411127418a92e8a3'
[14:27:28 CET] <cone-868> ffmpeg 03Hendrik Leppkes 07master:7968b1f85339: Merge commit '8edaf625f3c38c695c33745822182e94e17d6e1b'
[17:00:52 CET] <BBB> j-b: pokey
[17:01:22 CET] <j-b> BBB: pikey
[17:01:48 CET] <Timothy_Gu> nevcairiel, Daemon404: yes it was deployed at fatebeta.ffmpeg.org
[17:02:14 CET] <Timothy_Gu> It's kinda buggy in some ways tho and I can't find much time to work on it :(
[17:03:19 CET] <Daemon404> [16:01] < j-b> BBB: pikey <-- dont say that in the UK ;)
[17:03:33 CET] <BBB> ?
[17:03:49 CET] <Daemon404> it's a derogatory term her
[17:03:52 CET] <Daemon404> here*
[17:04:00 CET] <j-b> lol
[17:04:03 CET] <j-b> what does it mean?
[17:04:12 CET] <Daemon404> irish trabellers / gypsies
[17:04:24 CET] <j-b> so it's racist against irish?
[17:04:25 CET] <Daemon404> go watch Snatch (the movie)
[17:05:10 CET] <Daemon404> j-b, no, the northern irish, specifically a subset of them who are this: https://en.wikipedia.org/wiki/Irish_Travellers
[17:05:41 CET] <Daemon404> looks like some irish too actually.
[17:06:09 CET] <nevcairiel> Timothy_Gu: thats much better than the original already :)
[17:06:13 CET] <j-b> Daemon404: ok, thanks for the info.
[17:06:14 CET] <nevcairiel> hope you can find some time to finish it
[17:06:24 CET] <j-b> Daemon404: can I be racist against irish? :D
[17:07:13 CET] <Daemon404> sure, doesnt affect me ;)
[17:07:36 CET] Action: Daemon404 runs
[17:07:38 CET] <j-b> :D
[17:08:32 CET] <Daemon404> (i do recomment watching Snatch regardless, btw)
[17:08:35 CET] <Daemon404> recommend, even
[17:08:43 CET] <Daemon404> it's one of Guy Ritchie's better movies
[17:16:41 CET] <Timothy_Gu> nevcairiel: :)
[17:18:39 CET] <Daemon404> i really wish i could hide the sidebar
[17:22:39 CET] <Timothy_Gu> Daemon404: would it be better if I just remove it
[17:23:25 CET] <Daemon404> old fate has it to
[17:23:28 CET] <Daemon404> i dislike the sidebar fullstop
[17:23:30 CET] <Daemon404> ;)
[17:24:12 CET] <Timothy_Gu> nothing I can do then :) also I might rewrite the entire website in the future using Ajax so sorry Compn 
[18:10:53 CET] <durandal_1707> Huh, Ajax, why not in flash?
[18:16:22 CET] <cone-868> ffmpeg 03Kyle Swanson 07master:e5451f25d3c2: avfilter/asrc_sine: fix options typos
[18:34:05 CET] <cone-868> ffmpeg 03Kyle Swanson 07master:3b1939bb6679: avfilter/tremolo: fix wavetable buffer size
[18:47:52 CET] <dropd2na> ajax?  flash?  silverlight or nothing!
[19:06:27 CET] <wm4> png: Invalid PNG signature 0xFFD8FFE000104A46.
[19:06:30 CET] <wm4> what could this be?
[19:06:48 CET] <wm4> (mp4 with attached picture, generated by itunes, don't have the actual file)
[19:08:57 CET] <durandal_1707> jpeg?
[19:29:09 CET] <Daemon404> yes a jpeg
[19:29:15 CET] <Daemon404> FFD8 is a jpeg marker
[19:40:00 CET] <wm4> storing jpegs as pngs sounds familiar
[19:40:06 CET] <wm4> I think we've seen this in id3v2 tags
[19:40:12 CET] <wm4> but not mp4 yet (apparently)
[19:40:32 CET] <wm4> see 622f1468c92cb399e824894aa456d5590da6cc56
[19:41:10 CET] <Daemon404> what marks it as png?
[19:41:23 CET] <wm4> itunes, apparently
[19:41:53 CET] <Daemon404> i mean is it a flag in the mp4 somewhere
[19:41:56 CET] <Daemon404> or is it a stream
[19:41:57 CET] <Daemon404> or what
[19:42:07 CET] <wm4> good question
[19:42:30 CET] <Daemon404> inb4 some old hack for a broken file from the past
[19:42:47 CET] <wm4> there's a type id part of the 'data' atom
[19:42:57 CET] <wm4> apparently.
[19:43:00 CET] <wm4> maybe this is wrong
[19:44:07 CET] <Daemon404> not covr?
[19:44:13 CET] <Daemon404> where are you seeing this
[19:44:41 CET] <wm4> the covr atom is contained by the data atom, right?
[19:44:57 CET] <Daemon404> i dont recall the exact nesting
[19:45:07 CET] <Daemon404> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavformat/mov.c#l175
[19:45:18 CET] <wm4> see the only call to mov_read_covr in mov.c
[19:45:18 CET] <Daemon404> is the only place it could be aside from being a real "png" stream
[19:45:29 CET] <Daemon404> right
[19:49:56 CET] <Daemon404> http://sourceforge.net/p/atomicparsley/code/HEAD/tree/trunk/atomicparsley/src/AP_MetadataListings.cpp#l259
[19:49:59 CET] <Daemon404> lol.
[19:50:25 CET] <JEEB> lol
[19:50:27 CET] <Daemon404> i cant actually find any docs on 'covr'
[19:50:27 CET] <wm4> yeah.
[19:51:01 CET] <Daemon404> ah there it is... maybe
[19:51:27 CET] <Daemon404> i see where the type bit is coming from
[19:52:16 CET] <Daemon404> https://developer.apple.com/library/mac/documentation/QuickTime/QTFF/Metadata/Metadata.html
[19:52:19 CET] <Daemon404> covr isnt covered at all here
[19:53:23 CET] <wm4> https://developer.apple.com/library/mac/documentation/QuickTime/QTFF/Metadata/Metadata.html#//apple_ref/doc/uid/TP40000939-CH1-SW34
[19:53:32 CET] <wm4> the data atom is, and these are the types it supports
[19:53:36 CET] <wm4> (apparently)
[19:53:49 CET] <Daemon404> yes
[20:01:49 CET] <llogan> Compn: can you add mailer-daemon at mplayerhq.hu email address to Privacy Options > Sender Filters > discard_these_nonmembers? for dvdnav-discuss ML
[20:39:40 CET] <llogan> i should read the twitter more than twice a week. missed some re-tweet fodder.
[21:30:12 CET] <Compn> llogan : i dont think i have the power to admin that list
[21:30:20 CET] <Compn> (i know i dont)
[21:30:33 CET] <Compn> noone ever told me password
[21:30:47 CET] <Compn> llogan : i hope you passed twitter password to some other devs :P
[21:47:54 CET] <durandal_1707> michaelni: care to fix ogv sample?
[21:48:48 CET] <michaelni> durandal_1707, which ogv sample ?
[21:52:33 CET] <llogan> Compn: who does have access to that list? re: twitter, reynaldo has access. michael too, IIRC.
[21:53:19 CET] <durandal_1707> michaelni: https://0x0.st/iJi.ogv
[21:57:01 CET] <cone-618> ffmpeg 03Paul B Mahol 07master:e99321a3b6e1: avformat: add 3dostr demuxer
[21:57:02 CET] <cone-618> ffmpeg 03Paul B Mahol 07master:af3f793f6048: avcodec: add ADPCM AICA decoder
[21:57:03 CET] <cone-618> ffmpeg 03Paul B Mahol 07master:48f0d3accc30: avformat: add DC STR demuxer
[21:57:04 CET] <cone-618> ffmpeg 03Paul B Mahol 07master:09ef1f5923af: doc/general.texi: move Screenpresso to right section
[21:57:05 CET] <cone-618> ffmpeg 03Paul B Mahol 07master:863c9189eeed: doc/general.texi: fix SDX2 description
[21:57:28 CET] <michaelni> durandal_1707, whats the problem with the sample ?
[22:01:22 CET] <durandal_1707> michaelni: it doesn't play correctly inside mpv
[22:03:24 CET] <michaelni> it seems to play fine in ffplay but then i dont know how its supposed to look so maybe iam missing something
[22:08:00 CET] <durandal_1707> the audio is not playing correctly, player reports missing/invalid video timestamps and drops audio
[22:08:55 CET] <durandal_1707> how it plays with mplayer?
[22:12:23 CET] <ubitux> isn't frame_rate option in avfilter/buffersrc suposed to be of type AV_OPT_TYPE_VIDEO_RATE>
[22:12:25 CET] <ubitux> ?
[22:14:07 CET] <michaelni> durandal_1707, mplayer shows errors too, is that a regression ?
[22:15:33 CET] <durandal_1707> dunno
[22:21:04 CET] <Compn> llogan : no idea, maybe Rathann ? but then videolan took over dvdnav stuff? do they use our ml? lol
[22:21:49 CET] <wm4> michaelni, durandal_1707: libavformat doesn't output video timestamps, and yes, it's a regression
[22:23:18 CET] <llogan> Compn: i believe videolan forked it, as opposed to taking it over. I don't know.
[22:23:38 CET] <Compn> llogan : but then mplayer dropped using internal and went with the videolan fork
[22:23:48 CET] <Compn> iirc they took over maintaining it
[22:23:57 CET] Action: llogan queues Yackety Sax
[22:24:23 CET] <Compn> ask j-b :P
[22:25:05 CET] <Compn> ogv sample ? hentai of course.
[22:25:54 CET] <Compn> yes its a regression, wm4 / michaelni durandal_1707 , it works in mplayer r37289 (2014)
[22:25:56 CET] <j-b> of course I forked it.
[22:26:21 CET] <j-b> and after being annoyed that my mails were moderated on the mailing list, I forked the ml.
[22:26:39 CET] <Compn>   i would have fixed the ml, but no one gave me the password
[22:26:50 CET] <j-b> Sorry, dead project, I took over.
[22:27:05 CET] <Compn> j-b : its fine, thanks for doing it. we are just trying to figure out the old ml now :P
[22:27:10 CET] <j-b> and distros have picked it off
[22:28:02 CET] <Compn> michaelni : libavformat version 56.7.101 (internal)  , in the working mplayer on that ogv sample, if that helps any :P
[22:36:09 CET] <jamrial> that's an xvid ogm file. wonder why the ogv extension
[23:46:42 CET] <cone-618> ffmpeg 03Ganesh Ajjanagadde 07master:4870d8518a92: avcodec/snowenc: use log2 instead of log() / log(2...)
[23:46:43 CET] <cone-618> ffmpeg 03Ganesh Ajjanagadde 07master:68ba63ebd5fa: avcodec/nellymoserenc: use log2 instead of log()/M_LN2
[23:46:44 CET] <cone-618> ffmpeg 03Ganesh Ajjanagadde 07master:603d62745710: avcodec/zmbvenc: use log2 instead of log()/M_LN2
[23:46:45 CET] <cone-618> ffmpeg 03Ganesh Ajjanagadde 07master:4d0d85c94ad9: avfilter/af_volume: use log10 instead of log()/M_LN10
[23:51:47 CET] <cone-618> ffmpeg 03Andreas Cadhalpun 07master:41455dca60c4: avformat: stop exporting ffurl_read_complete, ffurl_seek and ffurl_size
[23:51:48 CET] <cone-618> ffmpeg 03Andreas Cadhalpun 07master:8bfbc8c5e504: swscale: rename sws_context_class to ff_sws_context_class
[23:51:49 CET] <cone-618> ffmpeg 03Andreas Cadhalpun 07master:e8c371606409: swscale: rename sws_rgb2rgb_init to ff_sws_rgb2rgb_init
[23:53:49 CET] <cone-618> ffmpeg 03Ganesh Ajjanagadde 07master:59d37f5a4e70: avfilter/vf_psnr: use log10 instead of log()/log(10)
[23:56:22 CET] <cone-618> ffmpeg 03Ganesh Ajjanagadde 07master:ec66bcc0e75c: avfilter/avf_showspectrum: use log10 instead of log()/...
[00:00:00 CET] --- Fri Oct 30 2015


More information about the Ffmpeg-devel-irc mailing list