[Ffmpeg-devel-irc] ffmpeg-devel.log.20130522

[burek]

[00:11] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:c649ecb3b232: avcodec/jpeg2000dec: get_qcx, fix stack and heap overwrites
[01:22] <llogan> is DV actually big endian?
[01:23] <llogan> "FIXME: may be we have to admit that DV is a big endian PCM"
[01:23] <llogan> http://code.google.com/p/ffmbc/issues/detail?id=72
[01:23] Action: llogan has no specs
[01:35] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:0b42631641d9: avcodec/jpeg2000dec: merge bytestream2 reading from j2kdec
[01:35] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:8d9fa738f5cd: j2kdec: cosmetics from jpeg2000dec
[02:21] <Compn> llogan : the specs = final cut pro
[02:21] <Compn> what final cut pro says is lawwwwwwwwwwww
[02:44] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:4a9a6a1f358c: vcodec/jpeg2000dec: fix av_log() type
[02:44] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:f399b33c2bb7: jpeg2000dec & j2kdec: Fix return type of get_sot()
[02:44] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:fec8783883f7: jpeg2000dec: read and check curtileno
[02:58] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:73ffab41959a: jpeg2000dec: check TPsot
[03:24] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:53f04424646d: avcodec/jpeg2000dec: Port tile size check from j2kdec
[03:24] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:17e66c9ff477: j2k & jpeg2000dec: reset numX/Ytiles if tiles failed allocation
[04:21] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:914ab4cd1c59: j2kdec & jpeg2000dec: check lengthinc against array size
[04:40] <cone-957> ffmpeg.git 03Carl Eugen Hoyos 07master:898268a23b3c: Add a fate test for ProRes Transparency.
[04:40] <cone-957> ffmpeg.git 03Michael Niedermayer 07master:32f550ee01bb: Merge remote-tracking branch 'cehoyos/master'
[04:59] <highgod> Hi, I want to ask a question, how can I check the video stream packets is all getted but the audio stream is not finished
[11:24] <cone-921> ffmpeg.git 03Martin Storsjö 07master:5b2b0b912dbc: configure: Enable the strtod workaround when building for bionic
[11:24] <cone-921> ffmpeg.git 03Michael Niedermayer 07master:c88e52e79782: Merge remote-tracking branch 'qatar/master'
[13:52] <burek> why is this ticket marked as invalid? https://ffmpeg.org/trac/ffmpeg/ticket/2587
[14:32] <michaelni> burek, i suspect that carl marked it as invalid as he could not reproduce and theres really not enough information to reproduce
[14:33] <michaelni> ill add some comments in the ticket
[14:44] <durandal_1707> well one could add defaults for libfaac to help novice users
[14:46] <durandal_1707> i dunno what bit rates libfaac supports
[14:57] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:cc0db8cf3042: avienc: Disallow the first frame to be skiped
[15:06] <michaelni> durandal_1707, ill fix the libfaac issue
[15:13] <michaelni> durandal_1707, burek patch to fix #2587 on the ML
[15:13] <michaelni> burek, please test the patch!
[15:36] <cone-978> ffmpeg.git 03Ash Hughes 07master:14ec9d259016: Support playing SMV files.
[16:38] <ubitux> we have a loop utils to check for the presence of a value in a int c-array, no?
[16:38] <ubitux> i somehow remember nicolas adding it somewhere
[16:39] <ubitux> ah, av_int_list_length_for_size!
[16:39] <ubitux> maybe.
[16:55] <ubitux> yay, hald clut done as well
[16:56] <ubitux> so we'll get at least 2 filters this week
[17:00] <durandal_1707> michaelni: is s16p to s16 path simd optimized?
[17:13] <durandal_1707> ubitux: +2 if you count me
[17:20] <ubitux> :)
[17:21] <michaelni> durandal_1707, its quicker if you look at the file than to ask me to look
[17:22] <michaelni> it should be SIMD optimized but id have to look to be 100% sure
[17:32] <Compn> :)
[17:33] <Compn> should ffmpeg report if its using cpuopts / c / asm version of each component (decoder, resampler, filter)
[17:34] <Compn> ... mencoder does it for some codecs and colorspaces
[17:36] <durandal_1707> michaelni: why is there no ssse3 variant of scalarproduct?
[17:38] <nevcairiel> because noone wrote one! :)
[17:39] <Compn> patch welcome ?
[17:39] <Compn> ehe
[17:49] <Skyler_> what exactly would ssse3 be used for in that function?
[17:50] <Skyler_> I don't see how you can get faster than movaps, mlups, addps without maybe FMA
[17:59] <durandal_1707> i know where little about asm right now....
[18:22] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:ad22767cb61c: shorten: fix "off by padding" bug
[18:28] <cone-978> ffmpeg.git 03Clément BSsch 07master:12f51c1f9e9a: lavfi: add lut3d filter.
[18:30] <durandal_1707> stop adding new stuff, i'm recompiling whole ffmpeg all the time
[18:32] <ubitux> haha
[18:33] <durandal_1707> all filters/decoder/encoders/muxers/demuxers should be split into small plugins
[18:34] <microchip_> ubitux: next task, port the deblock filters from mplayer :P
[18:34] <durandal_1707> adding new plugin(s) would not need recompiling of whole framework
[18:34] <ubitux> microchip_: -vf pp ?
[18:34] <microchip_> ubitux: pp7, spp, etc
[18:34] <durandal_1707> nobody uses that
[18:34] <microchip_> i do
[18:34] <microchip_> :p
[18:35] <ubitux> microchip_: pp is ported, and the only other relevant *pp* filter to port is spp
[18:35] <ubitux> which you can already use through -vf mp=spp
[18:35] <microchip_> yeah
[18:35] <ubitux> i've started porting it, but it's not trivial, and... i have various other stuff to do
[18:35] <ubitux> the port & add filter rate is pretty high already
[18:35] <microchip_> ic... no hurry, tho
[18:36] <ubitux> http://ubitux.fr/pub/pics/_ff-changes-filters.png
[18:37] <ubitux> filters everywhere ^
[18:37] <microchip_> oh, ow denoise is ported? Nice
[18:37] <microchip_> it's a very slow filter, tho
[18:37] <gnafu> Does it hurt?
[18:37] <gnafu> Painfully slow, I guess.
[18:38] <gnafu> ;D
[18:38] <ubitux> :)
[18:38] <ubitux> not as slow as dctdnoiz ;)
[18:38] <durandal_1707> i gonna remove whole libmpcodecs _now_
[18:38] <ubitux> durandal_1707: no :(
[18:38] <ubitux> durandal_1707: port filters first ;)
[18:40] <iive> durandal_1707: you are starting to sound a lot like libav.
[18:43] <saste> durandal_1707, on which filters are you working?
[18:43] <saste> mp filters i mean
[18:44] <saste> so we avoid to duplicate work
[18:45] <durandal_1707> saste: luckily 0, after libswscale that is next thing i don't want to look at next 100 years
[18:45] <ubitux> ETOOMUCHHATE
[18:45] <ubitux> saste: so, how is linuxtag? :)
[18:46] <saste> nobody doing eq?
[18:46] <saste> ubitux, not yet there
[18:46] <saste> i'll be there tomorrow, if i don't get lost
[18:46] <ubitux> the only filter i'm potentially working on is spp
[18:46] <ubitux> having eq* and the optims would be great.
[18:46] <ubitux> you can drop 2 mp filters at once when you have it
[18:47] <saste> eq seems the only useful (to me)
[18:47] <ubitux> eq2 is eq++ afaik
[18:47] <saste> but i started mcdeint as well
[18:47] <iive> there were talks in mplayer to drop eq as eq2 does everything and have fast-path for the eq cases.
[18:47] <ubitux> there is also pullup; Carl is interested in it
[18:48] <ubitux> but i don't want to do that port effort
[18:48] <ubitux> iive: mplayer is not really active anymore, and don't seem to plan to drop anything so far
[18:48] <ubitux> :(
[18:48] <microchip_> any plans for adding soft-telecine filter? to flag progressive content as soft-telecined for dvd encodes
[18:49] <iive> ubitux: that was some years ago..
[18:49] <ubitux> iive: my point exactly ;)
[18:50] <durandal_1707> microchip_: you have such (soft telecined) files?
[18:50] <iive> no, he wants to make such files
[18:50] <microchip_> durandal_1707: what iive said :)
[18:51] <microchip_> durandal_1707: mencoder can do it when using -of mpeg ...
[18:52] <durandal_1707> microchip_: no such files - no way to know you are doing nonsense
[18:52] <durandal_1707> no such files - softpulldown filter not gonna be ported
[18:52] <durandal_1707> no such files - softtelecine filter not gonna be done
[18:53] <microchip_> oh well
[18:53] <durandal_1707> both of those could be added to existing telecine filter
[18:54] <microchip_> durandal_1707: it's useful for when encoding to dvd
[18:54] <microchip_> to flag 23.976 fps as soft-telecined
[18:54] <durandal_1707> its useful to ... - does not make it happen
[18:55] <microchip_> i know, just asking
[19:04] <ubitux> microchip_: can you open a ticket?
[19:04] <ubitux> or better, just write the filter
[19:04] <microchip_> ubitux: i can't do C :(
[19:04] <ubitux> no
[19:05] <ubitux> you don't want to
[19:05] <microchip_> haha
[19:08] <ubitux> durandal_1707: i'm going to push something else, do you want me to wait?
[19:45] <durandal_1707> ubitux: what you gonna push?
[19:46] <durandal_1707> microchip_: so you do not have softelecine files?
[19:47] <ubitux> durandal_1707: v4l2 stuff
[19:47] <ubitux> i guess that doesn't affect you
[19:47] <ubitux> since you're on a bsd
[19:47] <microchip_> durandal_1707: i can create one with mencoder :)
[19:51] <durandal_1707> ubitux: something that needs to run configure again?
[19:51] <cone-978> ffmpeg.git 03Clément BSsch 07master:165bc9caa242: v4l2: make possible to disable libv4l2 at runtime.
[19:51] <ubitux> durandal_1707: nope
[19:51] <microchip_> durandal_1707: ftp://teambelgium.net:8082/pub/ter5.mpg <- soft-telecined with mencoder
[19:51] <ubitux> :)
[19:51] <durandal_1707> ubitux: goood
[19:58] <durandal_1707> ubitux: what is currently buggy in v4l2?
[19:59] <ubitux> playback is broken here
[19:59] <ubitux> some kind of framedrop
[19:59] <ubitux> or going back in time
[19:59] <ubitux> that's pretty weird
[19:59] <durandal_1707> regression?
[19:59] <durandal_1707> microchip_: i cant fetch that
[19:59] <ubitux> durandal_1707: if so, not within ffmpeg
[20:00] <durandal_1707> sure?
[20:00] <ubitux> yes i tried
[20:00] <ubitux> possibly with libv4l2, or my kernel
[20:00] <ubitux> and i'm not really willing to bisect any of those
[20:01] <durandal_1707> microchip_: strange works with firefox 
[20:02] <microchip_> hmm
[20:02] <durandal_1707> broken/misconfigured ftp server?
[20:02] <microchip_> could be :D
[20:04] <durandal_1707> looks like ffprobe can get it (so our ftp support is good)
[20:11] <durandal_1707> microchip_: and that file plays fine with mplayer -vf softpulldown ?
[20:12] <microchip_> durandal_1707: i haven't tested with -vf softpulldown. I just play it and mplayer detects the soft-telecine flag and switches FPS to 24000/1001
[20:14] <microchip_> (which is correct)
[20:16] <microchip_> it's basically a progressive MPEG2 with soft-telecine flag set, which is used for NTSC DVDs
[20:18] <durandal_1707> so it does not duplicate frames but just sets flags
[20:19] <microchip_> yes, correct
[20:19] <microchip_> the flag tells the DVD player to play it at 30000/1001
[20:19] <durandal_1707> ffprobe tells this files is 30.25 fps
[20:20] <microchip_> that's all soft-telecine is, as far as i know
[20:20] <microchip_> i made it with mencoder. MediaInfo says its 29.970
[20:21] <microchip_> I also encoded it with ffmpeg, and it outputted 29.970
[20:21] <microchip_> but, duh. As i set -r 30000/1001 :p
[20:22] <microchip_> anyways :)
[20:22] Action: microchip_ bbl
[20:23] <durandal_1707> ubitux: btw, v4l2 works on bsd too
[20:24] <ubitux> okay
[20:25] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:df3ee7b9a956: smvjpegdec: use av_pix_fmt_desc_get() avoid direct table access
[20:25] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:30df9789a974: proresdec: Fix end condition
[20:26] <iive> durandal_1707: soft telecine steams indicate higher fps, but encode less frames. the standard specifies that fields are repeated in according to the flags set.
[20:28] <iive> this is how a lot of soft-telecine turns out into hard telecine when transcoded.
[20:29] <durandal_1707> so what duplicates frames? ffmpeg?
[20:30] <t355u5> some .srt files start with 2 instead of 1 and ffmpeg gives the following message:
[20:30] <t355u5> subtitle.srt: Invalid data found when processing input
[20:31] <ubitux> use -f srt
[20:31] <ubitux> or fix your file
[20:32] <ubitux> or try to make srt_probe() more tolerant to this
[20:32] <t355u5> hmm, ok. how exactly? ffmpeg -i input.mp4 -i sub.srt -f srt ?
[20:32] <ubitux> -f srt -i sub.srt
[20:32] <ubitux> also, you should ask #ffmpeg for such question, unless you're willing to make srtdec.c:srt_probe() more tolerant
[20:33] <t355u5> ok. thx. I could also write a patch, but if -f exists, then people probably don't wan't a patch that ignores a minor error in the check. it is theoretically not spec, if it starts with 2...
[20:34] <ubitux> yes it's a broken file, so it's not recognized as a srt
[20:34] <ubitux> -f srt forces it
[20:34] <ubitux> we have a workaround to tolerate starting at 0, but that's all
[20:35] <t355u5> I asked here, because I was thinking about a patch, but you are right, maybe it would be wrong to do so
[20:36] <t355u5> what I usuallt do is replace the 2 with a 1, so 2 is actually missing, but neither ffmpeg, nor any player complains
[20:36] <t355u5> thanks for the info
[20:37] <ubitux> removing the counter check might make the probing less reliable, but it will likely make it slower as well
[20:37] <ubitux> (because a sscanf will always be run)
[20:38] <ubitux> there might be some room to improve the probing though
[20:38] <ubitux> feel free to send a patch
[20:38] <ubitux> (see libavformat/srtdec.c:srt_probe())
[20:39] <t355u5> I'll have a look. 
[20:40] <durandal_1707> ubitux: so what i need to use lut3d ?
[20:40] <ubitux> durandal_1707: a 3d lut
[20:41] <ubitux> durandal_1707: i'll send a patch for hald clut soon
[20:41] <ubitux> so it's easier to test
[20:41] <durandal_1707> what that does?
[20:41] <ubitux> it uses picture for the LUT
[20:41] <ubitux> and you can generate some using ImageMagick
[20:41] <durandal_1707> got picture? how?
[20:41] <ubitux> then process that LUT image with whatever filters, and then use that new picture as a LUT (rox!)
[20:42] <ubitux> durandal_1707: http://www.quelsolaar.com/technology/Hald_CLUT_Identity.png
[20:42] <ubitux> anyway, give me maybe an hour or two
[20:42] <ubitux> and i'll send it
[20:49] <ubitux> durandal_1707: you can try -vf lut3d=foobar.m3d with the pandora sample from the ticket
[20:49] <ubitux> or just look for random lut on the web
[20:52] <ubitux> durandal_1707: http://www.brianmeanley.com/resources/misc/sRGB_LUT.cube
[20:53] <ubitux> try this one, it's fun
[20:53] <ubitux> the effect with big buck bunny is nice
[20:55] <ubitux> it's possibly just a blue/red switch or something, but the effect is interesting
[20:55] <ubitux> anyway, there are a few lying around
[21:01] <durandal_1707> i can't understand what pict_type > 1 means
[21:01] <durandal_1707> some parsers do this
[21:39] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:657875b145c7: avcodec/aic: Fix vlc value checks
[22:01] <durandal_1707> -win4
[22:05] <durandal_1707> microchip_: the softtelecine filter can't be done yet, as it would be useless, no muxer/encoder make use of such flags
[22:13] <cone-978> ffmpeg.git 03Clément BSsch 07master:91b461459a5d: lavfi/lut3d: avoid overread.
[22:13] <microchip_> durandal_1707: sure it does, mpeg2video should support setting such flag when encoding to NTSC DVD format
[22:14] <durandal_1707> microchip_: i don't see ffmpeg support that
[22:14] <microchip_> ok, i give up
[22:24] <durandal_1707> what MP_IMGFIELD_ORDERED means?
[22:30] <cone-978> ffmpeg.git 03Paul B Mahol 07master:1f02927d288c: lavfi/mp: try to pass interlaced & tff flags
[22:43] <ubitux> i need to probe if a file is an image format
[22:43] <ubitux> but not print an error if it's not the case
[22:43] <ubitux> i only have the filename
[22:43] <durandal_1707> ?
[22:43] <ubitux> what should i use?
[22:48] <durandal_1707> one can't probe just with filename, looking into file is needed to. even attempting decoding maybe required
[22:56] <durandal_1707> i was wondering about writing psnr filter, which than could be used with drawtext(drawbar)
[22:56] <ubitux> yes, some analysis filters would be welcome
[22:56] <durandal_1707> but psnr just outputs some numbers, no?
[22:57] <ubitux> check tiny_{psnr,ssim}
[22:59] <durandal_1707> and for audio?
[23:02] <ubitux> no idea
[23:03] <ubitux> hald clut working within vf lut3d, yay.
[23:04] <durandal_1707> how you know its correct?
[23:04] <ubitux> i trust my instinct
[23:04] <ubitux> it's not very sharp
[23:04] <ubitux> but sometimes it works
[23:05] <durandal_1707> huh?
[23:05] <ubitux> anyway, i still have a 16-bit issue
[23:05] <ubitux> which sucks a bit
[23:06] <durandal_1707> well, first make 8bit working
[23:06] <ubitux> only native rgb16 is supported ’ owned if png 16 :(
[23:06] <ubitux> it works
[23:06] <ubitux> want to try it?
[23:07] <durandal_1707> what happens with png 16bit?
[23:07] <durandal_1707> ubitux: i can run ./configure once every week
[23:08] <ubitux> no need to run configure
[23:08] <ubitux> i've added it to lut3d
[23:10] <ubitux> durandal_1707: github/ubitux/lut3d ’ wget 'http://b.pkh.me/hald_sepia8.png' ’ -vf lut3d=hald_sepia8.png
[23:10] <durandal_1707> so whats problem with png?
[23:10] <ubitux> if it's a rgb16 png, it's likely be
[23:11] <ubitux> and only native rgb16 is supported in the filter
[23:11] <llogan> durandal_1707: is your computer old, or is that a BSD limitation/feature?
[23:12] <durandal_1707> ubitux: but swscale can do be/le conversion, use rgb48/rgba64
[23:12] <ubitux> (http://imgur.com/dhmFS3i  possinly faster than my bp)
[23:12] <ubitux> durandal_1707: yeah, i should use sws, but that complexifies the code slightly
[23:13] <durandal_1707> no, i mean filter should not call swscale by own .....
[23:14] <ubitux> i could use an input video stream yes, that's a possible improvement
[23:15] <ubitux> do you think i should do it that way directly?
[23:15] <durandal_1707> i'm confused
[23:17] <ubitux> basically, instead of relying on avformat/avcodec/sws, i could make lut3d work like this: -vf 'movie=hald_sepia8.png, lut3d=instream=1'
[23:18] <ubitux> and that also means we could replace the png with a video, to update the lut over the time
[23:18] <ubitux> but that will require some nasty magic like in overlay
[23:18] <durandal_1707> and what you currently do?
[23:18] <ubitux> lut3d=hald_sepia8.png
[23:19] <ubitux> then just like removelogo or movie etc, i open the file locally in the filter
[23:19] <durandal_1707> ugly, i find dynamic one more useful
[23:19] <ubitux> https://github.com/ubitux/FFmpeg/commit/19d8df66692659a987b63b650f859ceda2898dd6
[23:19] <ubitux> yes, dynamic is better in many ways
[23:20] <durandal_1707> and you do not need own sws/codec...
[23:20] <ubitux> except that it will make the code much more complex
[23:20] <durandal_1707> what complex?
[23:20] <ubitux> request frame thing, timing etc
[23:20] <ubitux> check vf_overlay..
[23:22] <durandal_1707> i disagree. it complex as you made it
[23:24] <ubitux> i'm going to do the stream approach
[23:25] <durandal_1707> llogan: when was last time you compiled ffmpeg yourself?
[23:26] <llogan> yesterday
[23:26] <llogan> but only because i wanted to see what was going on with recent x264
[23:27] <llogan> (users having trouble with ffmpeg not finding libx264 after latest x264 push)
[23:27] <durandal_1707> llogan: now imagine doing it every 5 minutes
[23:32] <llogan> sounds better than my current task...
[00:00] --- Thu May 23 2013