[Ffmpeg-devel-irc] ffmpeg-devel.log.20130712

[burek]

[00:43] <michaelni> durandal_1707, it sounds usefull
[02:27] <cone-947> ffmpeg.git 03Carl Eugen Hoyos 07master:0fff7f039c14: Supply a User-Agent header when opening rtsp streams.
[02:58] <cone-947> ffmpeg.git 03Thierry Foucu 07master:38ce775eead7: libavformat/mov.c: Skipped duplicated MOOV atom
[03:24] <michaelni> aballier, if you want to maintain 1.0, please send a patch for MAINTAINERs and a link to the git repo where you (will) maintain 1.0
[04:27] <cone-947> ffmpeg.git 03Piotr Bandurski 07master:161047f010b1: rmdec: fix crash in case of oom
[09:16] <xlinkz0> I'm reading from a rtsp stream and want to save to a different file every 60 seconds so every 60 seconds I create a new output context and save offsets for pts and dts so the new file starts from pts and dts 0
[09:17] <xlinkz0> what are some other corrections i need to do to the packets that come from the input context?
[09:17] <xlinkz0> is packet->pos important?
[09:17] <av500> no
[09:18] <av500> pos would refer to the byte offset in the rtsp stream
[09:18] <xlinkz0> where is ffprobe getting coded_picture_number ?
[09:20] <xlinkz0> also, ffprobe complains about the file that is created after the first file, it says : Missing reference picture, default is 0
[09:20] <xlinkz0> does it identify the ref picture by dts, pts = 0 ?
[09:20] <av500> well, if you cut a stream at a random point, of course there will be erros
[09:20] <av500> errors
[09:21] <xlinkz0> why?
[09:21] <av500> because frames are not independent
[09:21] <xlinkz0> ah i should cut only on iframes
[09:22] <xlinkz0> but how do i get that information without decoding?
[09:23] <xlinkz0> ffprobe decodes every frame?
[09:29] <av500> no
[09:29] <av500> there is a key frame flag
[09:29] <xlinkz0> in AVFrame
[09:29] <av500> in avpakchet
[09:29] <av500> avpacket
[09:29] <xlinkz0> http://ffmpeg.org/doxygen/trunk/structAVPacket.html
[09:29] <xlinkz0> can't see it :\
[09:29] <xlinkz0> ooooh 
[09:30] <xlinkz0> flag
[09:30] <xlinkz0> thanks, will try it
[09:53] <funman> hello
[10:00] <xlinkz0> av500: so i need to cut on packets that have the AV_PKT_FLAG_KEY flag right?
[10:01] <xlinkz0> do i need to include that packet in both the previous and the next file?
[10:03] <av500> in the next file
[10:03] <av500> but in case of mpeg2 that is not enough
[10:03] <xlinkz0> h264 mp4
[10:03] <xlinkz0> thanks
[10:03] <av500> you need a "closed gop"
[10:27] <wm4> who really uses the 0.5/0.7 branches?
[12:13] <cone-138> ffmpeg.git 03Martin Storsjö 07master:ef5b70affc63: configure: Check for support for labels in the inline assembly
[12:13] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:c6953589f9c4: Merge commit 'ef5b70affc6376bfeadd1ff649b79bad9a124fa8'
[12:41] <cone-138> ffmpeg.git 03Luca Barbato 07master:170fb593c617: pcx: K&R formatting cosmetics
[12:41] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:54bbb9056801: Merge commit '170fb593c617cd797d00cf05bfb8d8c21d612893'
[12:55] <Guest80280> I'm not sure if there's anyone responsible for the homebrew formula here, but does anyone know if there's an updated version for ffmpeg 2.0 in the works?
[12:56] <av500> 2.1?
[13:01] <michaelni> Guest80280, whats "homebrew formula" ?
[13:01] <michaelni> and 2.0.1 is planed in 1-2 weeks
[13:03] <relaxed> michaelni: I think it's like FreeBSD's ports for OS X.
[13:03] <Guest80280> Mac users install ffmpeg via homebrew, but the formula "ffmpeg" links to is still on 1.21 https://github.com/mxcl/homebrew/commits/master/Library/Formula/ffmpeg.rb
[13:04] <michaelni> ahh, understand
[13:07] <Guest80280> yeah, see here http://ffmpeg.org/trac/ffmpeg/wiki/MacOSXCompilationGuide Homebrew just makes it easy to install, but it hasn't been updated yet
[13:08] <cone-138> ffmpeg.git 03Luca Barbato 07master:3abde1a3b49c: pcx: Do not overread source buffer in pcx_rle_decode
[13:08] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:50617fce073c: Merge commit '3abde1a3b49cf299f2aae4eaae6b6cb5270bdc22'
[13:17] <cone-138> ffmpeg.git 03Luca Barbato 07master:2b379a925162: mlpdsp: x86: Respect cpuflags
[13:17] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:9d47333e3eee: Merge commit '2b379a925162b6783bd9a81dc03e647e8b65494c'
[13:36] <cone-138> ffmpeg.git 03Luca Barbato 07master:e9d394f3fad7: mlpdec: Do not set invalid context in read_restart_header
[13:36] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:e9348e107baf: Merge commit 'e9d394f3fad7e8fd8fc80e3b33cb045bbaceb446'
[13:36] <durandal_1707> a warning should be printed if user give multiple -vf/-af
[13:48] <cone-138> ffmpeg.git 03Luca Barbato 07master:c0d973c41b45: vdpau: use the correct namespace for the union
[13:48] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:0da0caa9a9ea: Merge commit 'c0d973c41b4568d5bad1295879e35cfa611bdcf2'
[14:13] <cone-138> ffmpeg.git 03Luca Barbato 07master:c82da343e635: pcm: always use codec->id instead of codec_id
[14:13] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:73f940e4d94c: Merge commit 'c82da343e635663605bd81c59d872bee3182da73'
[14:32] <cone-138> ffmpeg.git 03Luca Barbato 07master:f261e508459e: dca: Error out on missing DSYNC
[14:32] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:107a56c1ed8c: Merge commit 'f261e508459e28beca59868a878e1519a44bb678'
[14:50] <cone-138> ffmpeg.git 03Luca Barbato 07master:3802833bc1f7: dca: Respect the current limits in the downmixing capabilities
[14:50] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:cce254373c41: Merge commit '3802833bc1f79775a1547c5e427fed6e92b77e53'
[14:58] <cone-138> ffmpeg.git 03Sean McGovern 07master:52ea29867070: fate: use current syntax for the scale filter
[14:58] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:bc8240518830: Merge commit '52ea29867070243385220a223d7512ea7d4ae2d2'
[15:03] <aballier> michaelni: (re 1.0 maintainers) thanks for the offer but before claiming to be the maintainer id like to have started to do somthing about it, which I have not yet
[15:24] <michaelni> aballier, ok
[15:26] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:bec509db4364: vf_interlace: better handling of odd video size
[15:26] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:86d1a5d80f16: Merge remote-tracking branch 'qatar/master'
[16:16] <BBB> I don't think mac users necessarily use homebrew, a regular git checkout + xcode + yasm works
[17:20] <mateo`> hello, anyone to review the "lavf/movenc: support iTunes cover art" patch ?
[17:28] <b_jonas> I have verified, and it seems that https://trac.ffmpeg.org/ticket/2684 is indeed fixed, fix works for me. Thanks for all the developers.
[17:52] <cone-138> ffmpeg.git 03Paul B Mahol 07master:937cfebd72d3: lavfi/aevalsrc: remove hard limit on number of channels
[17:53] <cone-138> ffmpeg.git 03Michael Niedermayer 07master:8aea97a49d19: avcodec/jpeglsdec: Implement the 4th (lossy) transform
[17:59] <michaelni> mateo`, dont ask a group for help, ask individuals. If you ask "anyone help me iam bleeding to death" people will think "why me, there are other who could help, and i am in a hurry and have to catch that bus"
[18:05] <mateo`> michaelni: sure :)
[18:07] <mateo`> michaelni: is there an official maintainer of movenc ?
[18:07] <cone-138> ffmpeg.git 03Paul B Mahol 07master:b3405b1bdacf: lavfi/trim: use AV_OPT_TYPE_DURATION
[18:10] <michaelni> i think no active one, are you interrested ?
[18:15] <mateo`> michaelni: right now, i'm not
[18:17] <durandal_1707> this channel layout thing nonsense know nothing about >8 channels
[18:17] <durandal_1707> saste: how should i resolve this?
[18:17] <saste> a channel layout can specify more than 8 channels
[18:18] <saste> but i don't know which problem you're tackling
[18:18] <durandal_1707> how?
[18:19] <durandal_1707> try to create 9 channel audio with aevalsrc
[18:19] <wm4> there are no standard layouts for > 8 channels
[18:19] <saste> also there is a flag to tell "unknown" layout
[18:19] <wm4> but you should be able to create a layout by using a list of speakers
[18:20] <saste> see the work on nicholas, for specifying a generic channel count
[18:20] <durandal_1707> is it in master?
[18:23] <mateo`> michaelni: would you be interested in reviewing my patch then ?
[18:24] <saste> durandal_1707, sure
[18:28] <michaelni> mateo`, if noone else volunteers then sure (ping me if i forget, i probably will forget)
[18:29] <durandal_1707> saste: thing is if one does not specify layout, but just enters 9 expr it get error
[18:30] <mateo`> durandal_1707, ubitux: would you review the lavf/movenc: support iTunes metadata patch ?
[18:31] <durandal_1707> mateo`: only if you send me gift
[18:34] <durandal_1707> mateo`: you mean cover art?
[18:37] <mateo`> durandal_1707: yeay :)
[18:37] <mateo`> durandal_1707: another "evil" patch ^^
[18:46] <durandal_1707> saste: doing something lavfi related?
[18:50] <durandal_1707> if i remove that error check. than swr complains with silly error
[18:50] <durandal_1707> [SWR @ 0x2a04e000] Input channel count and layout are unset
[18:50] <durandal_1707> which is lie
[18:50] <durandal_1707> as channel count is set
[18:50] <durandal_1707> and layout if not set should not be problem
[18:50] <durandal_1707> libswresample is just useless
[18:51] <durandal_1707> perhaps devs only have 8 fingers
[19:07] <durandal11707> huh, why ticket get owned by michael? I never set it.
[19:13] <durandal11707> this pullup filter is one big mess
[19:17] <wm4> vf_pullup?
[19:20] <durandal11707> yes
[19:24] <saste> durandal11707, same limitation in libavresample?
[19:25] <durandal11707> don't know
[19:35] <durandal11707> this perspective filter is very interesting, gonna port it in 5 minutes
[19:38] <durandal11707> michaelni: ok to move it to LGPL?
[20:19] <michaelni> mateo`, do you want to review "0711 22:02 Matthew Heaney  (6.4K) 
[20:22] <durandal11707> ok, silence means it will stay GPL 
[20:32] <wm4> so, how do I actually know whether av_probe_input_format3 did a match by file extension?
[20:33] <wm4> is there any guarantee that the exact value of AVPROBE_SCORE_EXTENSION is used in this case, and never used for content based matching?
[20:33] <michaelni> wm4 isnt guranteed
[20:34] <wm4> so better do it manually by looping over all formats and do the matching yourself?
[20:34] <michaelni> what are you trying to do ?
[20:34] <wm4> doing matching by file extension as last fallback
[20:35] <michaelni> the probe code should already favor other means than file extension when possible
[20:36] <wm4> the probe function knows only the libavformat formats though
[20:36] <wm4> I might have my own code for certain things
[20:37] <michaelni> you could simply not tell the probe code the extension then it cant use it
[20:38] <wm4> that might work
[20:38] <michaelni> but that might end up with you doing probing twice first without extension then with
[20:38] <wm4> I first looked at the function av_guess_format() which sounds ideal, because it can handle mime types too
[20:39] <wm4> but it's for output formats only
[20:39] <wm4> (detection by mime types is missing from the probing API btw.)
[20:39] <wm4> well, probably not important at all for most people
[20:39] <michaelni> or just do your detection thing first and only when it fails use lavf probe
[20:40] <wm4> well, lavf content based probing should always be preferred
[20:40] <michaelni> or you could just "integrate" your detection by also picking scores and comparng againt the score fro lavf
[20:46] <michaelni> durandal11707, you broke trim
[20:46] <michaelni> ./ffmpeg -i matrixbench_mpeg2.mpg -ss 5 -t 1 out.mp4
[20:48] <durandal11707> how could i broke it?
[20:48] <michaelni> before it was 1 second now its 6
[20:48] <michaelni> out.mp4 that is
[20:49] <durandal11707> but why ffmpeg use trim?
[20:49] <durandal11707> trim stuff from fate did not complain at all
[20:50] <michaelni> ffmpeg uses trim because of: a83c0da539fb07260310bc3b34056239d2b138b2
[20:53] <durandal11707> if thats true fate would all be red
[20:55] <durandal11707> so that is not covered at all ...
[20:55] <durandal11707> thing is it code that insert trim filter need updating....
[20:55] <durandal11707> it still sets double instead of ints
[20:56] <durandal11707> michaelni: could you change it and see if it fixes it?
[20:56] <michaelni> it would be nice if ffmpeg didnt need to be chnaged because it could break other applications
[20:57] <michaelni> too
[20:57] <durandal11707> this thang is introduced recenly
[20:57] <durandal11707> and ffmpeg does not change
[20:58] <durandal11707> actually the double thing that changed in 2.0 is one that is "problematic"
[21:01] <durandal11707> i fixed it locally
[21:02] <michaelni> if you change ffmpeg to match trim it will break --enable-incompatible-fork-abi
[21:02] <durandal11707> fuck this
[21:04] <wm4> is anyone actually using this fork-abi thing
[21:06] <michaelni> if noone uses it we could drop it but still filters arguments should be compatible where possible
[21:07] Action: michaelni unhappy that durandal left
[21:07] <michaelni> nothing particularely hard to solve here 
[21:07] <michaelni> just a bit entangled
[21:12] <llogan> who wants some ffmpeg stickers? email me. if you're in europe then thilo will send them to you (but i need to send him a bunch of stickers first).
[21:12] <llogan> lou at lrcd.com
[21:13] <llogan> michaelni: ^ that should make up for the unhappiness
[21:24] <michaelni> llogan, yes i want some stickers 
[21:25] <llogan> i'll try to send them to thilo today.
[21:28] <cone-492> ffmpeg.git 03Carl Eugen Hoyos 07master:6476967f21dc: Fix compilation with --disable-everything --enable-demuxer=aac.
[21:28] <cone-492> ffmpeg.git 03Michael Niedermayer 07master:9219ec93b145: avfilter/trim: add compatibility layer to not break ABI used by ffmpeg
[21:28] <cone-492> ffmpeg.git 03Andrey Utkin 07master:1e85b5e077e7: ff_network_wait_fd_timeout(): check for interrupt before operation
[21:28] <cone-492> ffmpeg.git 03Andrey Utkin 07master:34fd21120d86: retry_transfer_wrapper(): check for interrupt before operation
[21:54] <cone-492> ffmpeg.git 03Paul B Mahol 07master:7f4a1fddd39e: lavf: add missing build dependency for tta demuxer
[21:55] <mateo`> michaelni: i might be able to review it tomorow
[21:55] <michaelni> mateo`, ok thanks
[23:52] <durandal_1707> michaelni: your 9-16 jpeg2000 patch fixes rgb48 pixfmt or not?
[23:54] <michaelni> durandal_1707, i still have to test
[23:56] <cone-492> ffmpeg.git 03Nicolas Bertrand 07master:dd1382ac9534: jpeg2000: Initialize only once mqc arrays
[23:57] <durandal_1707> michaelni: sending untested patches?
[00:00] --- Sat Jul 13 2013