[Ffmpeg-devel-irc] ffmpeg-devel.log.20130507

[burek]

[01:08] <ubitux> http://pastie.org/7810717
[01:08] <ubitux> ./record.sh bla.webm
[01:08] <ubitux> *clic*
[01:34] <cone-588> ffmpeg.git 03Matt Wolenetz 07master:004779cc59ae: avcodec/utils: Fix leak by dereferencing audio frame when side data causes whole frame skip
[03:07] <llogan> is there a chat log for the GSoC feedback?
[03:32] <Compn> llogan : ffmpeg had a private feedback
[03:33] <Compn> did you want that or the public channel log ?
[03:33] <Compn> (i dont have either)
[03:47] <llogan> the private of course. like the log from last year.
[03:47] <llogan> Compn: ^
[03:48] <llogan> i could not attend, unfortunately.
[03:48] <Compn> i forgot who handled it
[03:48] <Compn> saste maybe
[03:56] <llogan> why is the tag shown as n0.8 with "git describe"?
[04:10] <Compn> you are questioning the 'n' specifically then ?
[05:35] <cone-270> ffmpeg.git 03Reimar Döffinger 07master:762d4335aec2: matroska: set "done" only during resync fail.
[05:35] <cone-270> ffmpeg.git 03Reimar Döffinger 07master:399f6ef971eb: Add FATE test for matroska error recovery.
[06:39] <Zeranoe> I see vid.stab was added, and I'm getting a few people asking for it to be included. I tired to compile it from source, and everything worked, but no .pc pkgconfig file was generated. It keeps failing at configure with "check_pkg_config vidstab vid.stab/libvidstab.h vsMotionDetectInit" which is probably because I didn't include the path, plus no .pc file. So has anyone managed to compile vid.stab into a FFmpeg build?
[09:31] <durandal_1707> michaelni: so you did not review my gbrap patch? :(
[09:36] <highgod> Hi, I want to ask a question, where can I find any code to using rtsp for reference? Thanks
[10:02] <michaelni> durandal_1707, i wanted but it breaks fate
[10:03] <durandal_1707> it just missing new entries, should not break old ones
[10:05] <durandal_1707> but yes, i updated and running fate for 2nd time
[10:06] <michaelni> ok, thanks
[10:57] <cone-270> ffmpeg.git 03Janne Grunau 07master:74652af821d7: pixdesc: add function to return pixel format with swapped endiannes
[10:57] <cone-270> ffmpeg.git 03Michael Niedermayer 07master:60fd88059a62: Merge commit '74652af821d7e526a6541876a637fbf81261fa8e'
[11:10] <cone-270> ffmpeg.git 03Janne Grunau 07master:c1eb3e7fecdc: swscale: add support for endianness only conversion
[11:10] <cone-270> ffmpeg.git 03Michael Niedermayer 07master:2060d944090f: Merge commit 'c1eb3e7fecdc270e03a700d61ef941600a6af491'
[11:20] <cone-270> ffmpeg.git 03Janne Grunau 07master:9cfa21c8c2b8: swscale: support endianness conversion for AV_PIX_FMT_XYZ12
[11:20] <cone-270> ffmpeg.git 03Michael Niedermayer 07master:dce12f694864: Merge commit '9cfa21c8c2b81146aab574f55c93f27e5c2d435c'
[11:32] <cone-270> ffmpeg.git 03Janne Grunau 07master:5f87c277bd5c: vf_scale: add endianness conversion pixel format in query_formats
[11:32] <cone-270> ffmpeg.git 03Michael Niedermayer 07master:fe5b08959663: Merge commit '5f87c277bd5caa09cc4f9061d4ccdd58dc121110'
[11:41] <cone-270> ffmpeg.git 03Janne Grunau 07master:851465f9d9ed: fate-jpeg2000-dcinema: force output to xyz12le
[11:41] <cone-270> ffmpeg.git 03Michael Niedermayer 07master:9276467bcec5: Merge remote-tracking branch 'qatar/master'
[11:48] <michaelni> durandal_1707, fate-filter-pixfmts-scale fails for me
[11:48] <michaelni> -gbrap               fb463f07681f099871db2309dd438123
[11:48] <michaelni> +gbrap               7d4452269fab4e29b4a9458be112d03c
[11:50] <durandal_1707> only that one?
[11:52] <michaelni> apparently but let me doubel check
[11:55] <michaelni> yes fate passes if i chnage this one
[12:05] <durandal_1707> michaelni: perhaps the A thing you wrote in yuv2gbrp_full_X_c is wrong? (overflows)
[12:20] <durandal_1707> but that can't be as same is in packed rgb
[12:24] <durandal_1707> lol, i see problem
[12:24] <durandal_1707> alpha plane somehow get wrong
[13:10] <durandal_1707> michaelni: the alpha plane width get divided by two for some reason
[13:20] <durandal_1707> the commit 7b41c24c5fa221b55a9302efaf1a9 mentions planar RGB but set only >8 bit planar RGB
[13:21] <durandal_1707> same for 3c2ecfcc24761a
[13:25] <cone-270> ffmpeg.git 03Michael Niedermayer 07master:2efbbc464458: avutil/av_pix_fmt_swap_endianness: simplify and fix code
[13:32] <michaelni> durandal_1707, does adding 8bit there fix anything ?
[13:36] <durandal_1707> but why would it be needed, only alpha gets fcked up
[13:36] <durandal_1707> and only when scaling
[13:37] <durandal_1707> so its somewhere is scaling code
[13:42] <durandal_1707> michaelni: it does not help
[13:45] <durandal_1707> michaelni: http://oi41.tinypic.com/juh30n.jpg
[13:47] <cone-270> ffmpeg.git 03Gunther Piez 07master:acc4a7e5b350: Ticket #2540 Enable threading when using -c:v libx264rgb
[13:51] <michaelni> durandal_1707, dunno what is causing this assuminh that represents the alpha plane, have you tried the various flags ?
[13:52] <durandal_1707> i'm trying other scalers, gauss/bitexact/... does not help
[13:57] <michaelni> i was mainly thinking of the flags related to half resolution
[13:59] <durandal_1707> ffmpeg -v debug -i /tmp/1.nut -vf format=gbrap,scale=800x600:flags=+gauss-bilinear+full_chroma_inp+full_chroma_int,format=gbrap,extractplanes=a -vcodec png /tmp/x.nut
[13:59] <durandal_1707> same
[14:25] <Bor0> I'm trying to implement a "buffering" state to my player. so I added a new integer in the VideoState struct, and here's what I do: is->buffering = 1; av_read_frame(...); is->buffering = 0; but this always shows that is->buffering is 0, even when it's buffering. what am I doing wrong?
[14:27] <av500> its not buffering, its blocking
[14:27] <av500> er
[14:29] <av500> I mean, if av_read_frame(...) is fast enough, another thread wont see buffering == 1
[14:32] <cone-270> ffmpeg.git 03Carl Eugen Hoyos 07master:f381dc1d6186: Support decoding of Lagarith frame type FRAME_SOLID_COLOR.
[14:32] <cone-270> ffmpeg.git 03Michael Niedermayer 07master:144baf427a6b: Merge remote-tracking branch 'cehoyos/master'
[14:34] <Bor0> what's a proper way to implement this?
[14:34] <Bor0> I'm reading from http, so buffering can take some time
[14:39] <av500> first make sure you are reading your flag correctly
[14:40] <Bor0> I'm constantly printing it
[14:41] <av500> yes, from a different thread
[14:41] <av500> so the compiler might assume its constant
[14:41] <Bor0> hmm, why does that matter? my structure is global
[14:41] <durandal_1707> michaelni: actually bugs happen when doing gbrap->subsampled yuv conversion
[14:44] <durandal_1707> eg, chroma subsampling get leaked into alpha
[14:48] <durandal_1707> actually it happens with 444p too
[15:27] <cone-270> ffmpeg.git 03Clément BSsch 07master:c3dcd3a255a0: sws: factorize rgb 9+ planar funcs declarations.
[15:32] <durandal_1707> michaelni: i'm out of ideas where is error
[15:39] <cone-270> ffmpeg.git 03Clément BSsch 07master:5043d8d213f1: lavfi/alphaextract: copy width and not linesize.
[15:44] <cone-270> ffmpeg.git 03Clément BSsch 07master:d3f7b98f87ff: lavfi/overlay: reindent fix.
[16:16] <ubitux> michaelni: swscale.c:const int chrSrcSliceH           = -((-srcSliceH) >> c->chrSrcVSubSample);
[16:16] <ubitux> isn't this supposed to be HSub?
[16:16] <ubitux>     const int chrSrcSliceY           =     srcSliceY  >> c->chrSrcVSubSample;
[16:16] <ubitux>     const int chrSrcSliceH           = -((-srcSliceH) >> c->chrSrcVSubSample);
[16:18] <michaelni> SliceH = slice Height, VSubSample = Vertical sub sample, HSub would be Horizonbtal
[16:20] <ubitux> aah ok i see
[16:20] <ubitux> thx
[16:23] <durandal_1707> michaelni: its looks like alpha plane is not read when scaling, should i add readAlpPlane?
[16:28] <michaelni> durandal_1707, iam not sure what you want to add where, you seem to assume i know where the issue is and what you plan to do ...
[16:29] <durandal_1707> michaelni: see hyscale in swscale.c
[16:29] <durandal_1707> alpToYV12 is never set for gbrap
[16:35] <michaelni> maybe i miss something but why should it be set ? the data is already in 8bit planar format
[16:36] <durandal_1707> perhaps because it wants it in 16bit
[16:48] <durandal_1707> i added some code, and also shift alpha by 8 and get something sane, but with less detail
[16:52] <durandal_1707> actually shifting by 6 give correct output
[16:53] <michaelni> yes its a bit annoying, i had not realized the rgb 16bit issue
[17:03] <cone-270> ffmpeg.git 03Michael Niedermayer 07master:d23b8462b5a4: gifdec: reset previous Graphic Control Extension disposal type
[17:30] <durandal_1707> fflogger is duplicating stuf
[17:46] <durandal_1707> hmm, neigther id3info nor id3v2 list id3 tags created by ffmpeg
[17:53] <durandal_1707> michaelni: should i pass table when calling function, or just call it with NULL arg?
[17:54] <michaelni> doesnt matter currently
[17:54] <durandal_1707> yes, but what you prefer?
[18:01] <michaelni> null
[18:06] <xlinkz0> how do i transform the return value of av_seek_frame into a string to know why it failed?
[18:06] <xlinkz0> or more simply, how do i find out why it failed?
[18:07] <ubitux> michaelni: any name suggestion?
[18:09] <ubitux> ROUNDED_INF_RSHIFT is a bit long
[18:09] <durandal_1707> xlinkz0: some return values are defined in libavutil/error.h
[18:11] <xlinkz0> thanks 
[18:17] <michaelni> ubitux, no good ones atm
[18:26] <cehoyos> __gb__: Ping
[18:34] <durandal_1707> cehoyos: why you do not want to send patch from #2527 to ml?
[18:35] <cehoyos> (Why do you think that my answer changed in the last 24 hours?) I don't know if there is a bug or not. If you know that there is a bug, please (review it and) push my patch.
[19:18] <ubitux> cbsrobot: 'just submitted the true/sample peak thing
[19:19] <ubitux> no visual yet, but av_log'ed and metadata injected
[19:19] <ubitux> should be good enough
[20:06] <cone-605> ffmpeg.git 03Michael Niedermayer 07master:89d998f1c1ed: shorten: allocate space for padding
[20:24] <cbsrobot> ubitux: I tried it - but in what unit is it ?
[20:24] <ubitux> double :p
[20:24] <ubitux> [0;1.0]
[20:25] <durandal_1707> 0 too?
[20:25] <cbsrobot> would be nice to have it in dBTP
[20:36] <ubitux> mmh true
[20:37] <cbsrobot> and if you plan to print the values to the summary - which would be very usedful - I think there is also some channel weighting going for multichannel files
[20:41] <cone-605> ffmpeg.git 03Michael Niedermayer 07master:f96e0eb2387f: dxa: check vectors of 4x4 motion blocks
[20:41] <cone-605> ffmpeg.git 03Michael Niedermayer 07master:ead590c25619: dxa: check vectors of 2x2 motion blocks
[20:46] <durandal_1707> Compn: have any sbc samples?
[21:11] <cone-605> ffmpeg.git 03Michael Niedermayer 07master:ad002e1a13a8: avcodec/cdgraphics: check buffer size before use
[21:43] <Compn> durandal_1707 : are they in trac ?
[21:43] <Compn> i know aac sbr
[21:43] <Compn> what sbc ?
[21:55] <nevcairiel> the only sbc i know is a subband codec for bluetooth audio devices, not sure you can even put that into a file
[21:55] <gnafu> Anything is possible ;D.
[21:56] <Compn> http://wiki.multimedia.cx/index.php?title=Bluetooth_SBC
[21:56] <Compn> yeah i think theres a rfc for sbc 
[21:57] <Compn> durandal_1707 : so probably no samples 
[21:58] <durandal_1707> i just created some files ...
[21:59] <durandal_1707> i just got interested because it got mentioned on ha
[21:59] <Compn> if you put them in some http i can put them on samples repo
[22:00] <durandal_1707> no need, i just wanted to know is it popular enough to be included in lavc
[22:04] <cone-605> ffmpeg.git 03Michael Niedermayer 07master:9dd04f6d8cdd: sanm: Check dimensions before use
[22:12] <durandal11707> michaelni: what interger overflow 9dd04f6d8cdd fixes?
[22:32] <cone-605> ffmpeg.git 03Michael Niedermayer 07master:a527e692592b: avformat/vqf: check number of channels before use.
[22:41] <Sergey> Hello. About the bug #2425 ([regression] Smacker audio played with clicks). It was introduced by lib av, should I report it to their bugtracker instead?
[22:43] <durandal_1707> no, join #libav-devel, and explain them bug
[23:10] <Sergey> durandal_1707: Ok, thanks, trying...
[23:53] <cone-605> ffmpeg.git 03Nicolas George 07master:449704b47500: ffmpeg: move a local variable definition later.
[23:53] <cone-605> ffmpeg.git 03Nicolas George 07master:78d16b6a5109: lavfi: fix filter format negotiation loop.
[23:53] <cone-605> ffmpeg.git 03Nicolas George 07master:4582e4c086bc: lavfi: add comments to explain the negotiation loop.
[23:53] <cone-605> ffmpeg.git 03Michael Niedermayer 07master:bc0cbaca62b6: Merge remote-tracking branch 'cigaes/master'
[00:00] --- Wed May  8 2013