[Ffmpeg-devel-irc] ffmpeg-devel.log.20130210

[burek]

[00:44] <Zeranoe> Slightly off topic here but anyone know why patches are created with -up or "--git" instead of just running a standard diff command with no options? Don't the patches work the same? Is it just for readability purposes? 
[00:46] <beastd> Zeranoe: --git is what you will get from git naturally
[00:48] <Zeranoe> beastd: Do the patches include the context around the modified area just for readability sake?
[00:49] <beastd> Zeranoe: not only
[00:50] <beastd> Zeranoe: it is also for finding the right spot to apply the patch (probably with "fuzz")
[01:02] <J_Darnley> Surely patches are made that way because people ask git to make them rather than running diff on the files themselves...
[01:50] <Compn> zera
[01:50] <Compn> sfakgj
[01:50] <Compn> J_Darnley : patches are made with git send email because you can apply these patches from your email client in one click / keystroke
[01:50] <Compn> one or two anyways
[02:27] <cone-273> ffmpeg.git 03Michael Niedermayer 07master:13ae02d03fbb: imgconvert: add get_pix_fmt_score()
[02:27] <cone-273> ffmpeg.git 03Michael Niedermayer 07master:cf8d9b74f306: imgconvert: use get_pix_fmt_score in avcodec_find_best_pix_fmt_of_2()
[05:34] <Darkarnium> Afternoon
[07:05] <Darkarnium> Bit of a strange query regarding AVOptions. I patched one of the demuxers and added a few AVOptions that can be passed as CLI arguments. This works without issue under ffplay, all options are passed and the relevant calls therein are performed with the options specified via the flags. However, if I instead use the ffmpeg binary, the flags seem to be ignored and the demuxer never uses the CLI arguments.
[07:05] <Darkarnium> Is there any major difference between the handling of options between these two applications?
[07:06] <Darkarnium> I'm sure it must just be something I've missed. However, I've also taken the newly compiled binary and executed it directly - in case I was having issues with ffplay and ffmpeg being different versions, etc.
[07:08] <Darkarnium> Any suggestions, documentation or mailing list threads would be appreciated - if I'm missing something here, which is most probable.
[07:18] <Skyler_> what's your ffmpeg commandline?
[07:57] <Compn> Darkarnium : wheres your patch ?
[07:58] <Compn> Darkarnium : did you specify the options in the right order? it matters in ffmpeg for some reason
[07:58] <Compn> eg ffmpeg -demuxeroptions filename is different than ffmpeg filename -options
[08:02] <Compn> anyone rememeber what i was supposed to add to ffmpeg tracker ?
[08:04] <Compn> Darkarnium : saste found some strange bugs with -user-agent and other options in ffmpeg. maybe there is bug...
[08:04] <Darkarnium> I'll have a look at the order then, see how far I get :)
[08:04] <Darkarnium> The patch isn;t something I'd be submitting to be honest, as it's for an internal application :)
[08:05] <Darkarnium> I can't see there being another user who is doing something this broken to be honest :P
[08:06] <Compn> someone remind me if bayer patch didnt get applied ...
[08:06] <Compn> oh i think it still crashes
[08:07] <Compn> Darkarnium : i meant we can check what you did
[08:07] <Compn> to see if you did it wrong
[08:07] <Compn> but you might want to ask saste
[08:08] <Compn> a lot of option code got moved around, apparently
[08:10] Action: Compn gone, night
[08:11] Action: Compn pings pross about bayer decoder
[08:11] <Compn> now night
[10:53] <cone-23> ffmpeg.git 03Carl Eugen Hoyos 07master:f02f231e05e0: Do not write invalid matroska files.
[11:12] <saste> may we remove the rfps spam from the log?
[12:02] <durandal_1707> why ffplay outputs only yuv420p ?
[12:02] <nevcairiel> because of SDL, more or less
[12:06] <durandal_1707> and it still cant output 444 or rgb?
[12:08] <durandal_1707> sdl 2 version i mean
[12:18] <cone-23> ffmpeg.git 03Diego Biurrun 07master:8658e1c68253: mss4, ra288: Remove unused DSPContext local codec context members
[12:18] <cone-23> ffmpeg.git 03Vladimir Pantelic 07master:36fab50e90d1: asfdec: silence a warning
[12:18] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:0af4506e15a6: Merge commit '36fab50e90d15352e403e4cc210890810f2fb4e2'
[12:18] <saste> durandal_1707, should we add a test for il?
[12:19] <saste> durandal_1707, in fact SDL1.2 supports a few more formats, check lavd/sdl
[12:21] <durandal_1707> saste: that are just packed variants
[12:21] <durandal_1707> saste: feel free to add test for il,lut,negate, ...
[12:38] <cone-23> ffmpeg.git 03Vladimir Pantelic 07master:61f9ad2dfcb3: asfdec: read the full Metadata Object, not just aspect ratio information
[12:38] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:3600049dae9e: Merge commit '61f9ad2dfcb3f98b7ac5777d19d0e7b61d0be01e'
[12:45] <saste> durandal_1707, currently il segfaults on rgb48
[12:45] <cone-23> ffmpeg.git 03Vladimir Pantelic 07master:84b721db366c: asfdec: also read Metadata Library Object
[12:45] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:30085bd43e35: Merge commit '84b721db366c0734fdfd23c8daaa7da7da21f761'
[12:45] <saste> so yes, we should add a test for it
[12:49] <durandal_1707> saste: not when i tested rgb48
[12:49] <durandal_1707> saste: don't you have patch for it?
[12:50] <saste> durandal_1707, it doesn't segfault with my patch
[12:52] <durandal_1707> saste: it does not segfault here
[12:52] <cone-23> ffmpeg.git 03Vladimir Pantelic 07master:db0a943266be: avplay: apply the stream sample_aspect_ratio to decoded video frames
[12:52] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:d80820655362: Merge commit 'db0a943266be29ff0596872ebb418dfed75d00de'
[12:53] <saste> durandal_1707, -vf format=bgr48be,il=l=d:c=d,showinfo,format=bgr48be
[12:53] <saste> anyway i don't mind, I'm going to visually test with my patch on
[13:17] <cone-23> ffmpeg.git 03Vladimir Pantelic 07master:f5fac6f77752: asfdec: support reading ID3v2 tags in ASF files
[13:17] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:9a714d871ac5: Merge commit 'f5fac6f77752931347ab302563802dcaa49c2419'
[13:21] <saste> durandal_1707, can you explain this: ffplay -f lavfi color -vf format=monow,il=l=d:c=d,format=monow
[13:21] <saste> check the first frame
[13:21] <durandal_1707> saste: about your noise source, how you expected to use it with other videos?
[13:22] <saste> with my patch it looks ok
[13:22] <cone-23> ffmpeg.git 03Anton Khirnov 07master:1ef0e8a6bf89: asfdec: do not assume every AVStream has a corresponding ASFStream
[13:22] <cone-23> ffmpeg.git 03Anton Khirnov 07master:81726a4f0b8a: FATE: add tests for additional flavors of asf cover art
[13:22] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:e88f392733fe: Merge commit '81726a4f0b8a43e19898e2a36fdde80583bafff0'
[13:22] <durandal_1707> saste: if it does not make sense, filter out bitstream formats in query_..
[13:22] <durandal_1707> saste: i said your patch is ok
[13:23] <saste> durandal_1707, yes but i can't understand what's wrong
[13:23] <durandal_1707> i think i just made it bug to bug compatible with mp=il
[13:25] <saste> durandal_1707, noise source? which one?
[13:25] <saste> the old and slow patch?
[13:26] <durandal_1707> there was only one noise source afaik
[13:28] <durandal_1707> saste: ffplay have strange buffer handling
[13:28] <durandal_1707> use ffmpeg, ffplay hide several bugs i had with il
[13:53] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:d6d5ef5534d5: sws: GBRP9, GBRP10, and GBRP16 output support
[13:53] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:42431a557e2d: Merge commit 'd6d5ef5534d582f9dbaf18ac2605cf5bb72cd821'
[13:55] <saste> durandal_1707, i don't think it is a good idea to have separate modes for il
[13:55] <saste> it was a (bad) idea ported from mplayer
[13:55] <saste> for example what's about alpha mode and packed?
[13:55] <saste> why should I want to set a different interleaving mode for the various components?
[13:56] <saste> why do i need to set it up to make it work with a given format (and i have to account for the pixel format planar/packed layout)?
[13:59] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:e58013dd8f00: sws: use planarRgbToRgbWrapper only for 8bit per component
[13:59] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:d1d9b93e9a21: Merge commit 'e58013dd8f00e17dd98ba64e41c72f1d152f6608'
[14:03] <cone-23> ffmpeg.git 03Stefano Sabatini 07master:63a99622876f: lavfi/il: simplify/generalize linesize computation
[14:03] <cone-23> ffmpeg.git 03Stefano Sabatini 07master:0b1896543446: lavfi/il: move query_formats() definition up
[14:16] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:7b41c24c5fa2: sws: dont enable chrSrcHSubSample for planar RGB
[14:16] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:f2b89f21cc3d: Merge commit '7b41c24c5fa221b55a9302efaf1a9eeb38b12551'
[14:27] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:813784a8bf90: sws: disable yuv2rgb warning for planar rgb.
[14:27] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:02b3ac87f81a: Merge remote-tracking branch 'qatar/master'
[14:34] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:2c5e4ae29180: Revert "Merge commit 'db0a943266be29ff0596872ebb418dfed75d00de'"
[14:45] <cone-23> ffmpeg.git 03Carl Eugen Hoyos 07master:034a125c8c51: Support more atrac3-in-mkv samples.
[15:18] <durandal_1707> so what is status of vf_psnr?
[16:05] <durandal_1707> one cant generate pink/brown/.. noise with aevalsrc?
[16:41] <durandal_1707> burek: why ubuntu packages are that old?
[16:50] <creep> h
[16:55] <wm4> durandal_1707: maybe because ffmpeg APIs change so often, and they can't get all packages to work with newer ffmpeg... that's generally the reason why ffmpeg packages are old
[16:55] <burek> durandal_1707 ?
[16:56] <nevcairiel> ubuntu doesnt have ffmpeg packages, it only has "fake" ones that actually are libav
[16:56] <wm4> it probably has old libav packages then :)
[16:58] <nevcairiel> someone should come up with a popular linux tool that only works with ffmpeg and not libav, and see what distributions do 0p
[16:58] <durandal_1707> burek: i talk about one in link provided from ffloger
[16:59] <durandal_1707> burek: also there could be static build for linux
[17:04] <burek> nevcairiel, ffmpeg is already a popular tool ^^
[17:04] <burek> durandal_1707, can you provide more references?
[17:04] <burek> so i could know what are you talking about
[17:06] <burek> btw, searching for enhancements on trac and trying to implement the most easy/quick ones would make ffmpeg have more popular functionality than libav (which it already has though), but it would allow users to choose ffmpeg, because "it can do something that libav cant" or something
[17:08] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:6462268e74fa: pcx: fix rounding in bytes_per_scanline check
[17:13] <cone-23> ffmpeg.git 03Stefano Sabatini 07master:e005697af685: tests: add fate-lavfi-il test
[17:13] <cone-23> ffmpeg.git 03Stefano Sabatini 07master:60950adc18fe: lavd/v4l2: implement consistent error handling
[17:13] <durandal_1707> burek: read #ffmpeg
[17:15] <durandal_1707> saste, michaelni, ubitux what is status of lavfi scritpting and dynamic/conditional filtergraph?
[17:15] <burek> oh you mean Jon Severinsson's FFmpeg PPA
[17:15] <burek> well, ask Jon Severinsson...
[17:15] <burek> I just linked his ppa
[17:15] <burek> because we don't have anything "ours"
[17:15] <saste> durandal_1707, still
[17:15] <burek> I mean, probably the guy is also here, so just ask him, i guess
[17:16] <saste> durandal_1707, what features do you exactly need?
[17:17] <saste> a lua filter is just part of it, in general we may need to add more events to the already implemented filters
[17:17] <saste> for example to change position to an overlay (or disable it)
[17:18] <durandal_1707> saste: crossfading for example, one apply fade out and mix/blend it with fade in of another input
[17:18] <durandal_1707> no caching of any kind is needed
[17:20] <saste> durandal_1707, concat+fade?
[17:20] <saste> now the problem is that you  don't know when a filter stream is going to end
[17:21] <durandal_1707> concat is useless here ....
[17:21] <saste> yes *cross*-fading i see
[17:22] <durandal_1707> it already can be done if several steps - which is PITA
[17:22] <durandal_1707> *in
[17:33] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:f41329c8aec0: pcx: use init_get_bits8()
[17:33] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:d24de4596c3f: pcx: Add missing padding to scanline buffer
[17:35] <michaelni> saste, libavdevice/v4l2.c:688:20: warning: ret may be used uninitialized in this function [-Wuninitialized]
[17:39] <saste> Assertion bytes >= needed_bytes failed at libavformat/matroskaenc.c:166
[17:39] <saste> now i wonder why this happens...
[17:50] <saste> ^^ anyone has ideas?
[17:51] <durandal_1707> saste: how it happens
[17:54] <saste> durandal_1707, ffmpeg -i IN.flv -an -codec:v copy -y OUT.mkv
[17:54] <saste> this is happening at the end, when closing the file
[17:56] <saste> apparently in end_ebml_master()
[17:59] <durandal_1707> i ported noise filter
[18:00] <saste> durandal_1707, do you plan to port all mp filters?
[18:01] <durandal_1707> saste: not just mp useful filters but all useful filters
[18:01] <durandal_1707> ilpack for examples is hardly useful
[18:01] <wm4> but, 2002!
[18:02] <nevcairiel> 2002 called, they want their filters back
[18:02] <wm4> I think vf_pullup and vf_unsharp are relatively useful (though still a league below modern filters)
[18:03] <wm4> vf_dsize doesn't make sense, you can delete it
[18:03] <durandal_1707> so every new year I need to rewrite each decoder/encoder/demuxer/muxer?
[18:04] <nevcairiel> every 10 years is enough
[18:05] <durandal_1707> the idea is to have addgrain like filter
[18:06] <saste> unsharp is already in, but it has some differences with the original unsharp
[18:07] <durandal_1707> like what?
[18:07] <wm4> uh, so you're keeping the mp one?
[18:08] <saste> durandal_1707, like I wrote a patch long time ago, i sent it and it was approved, but then I was not satisfied with the patch and is still there
[18:08] <saste> i wanted to read the papers and understand more about the algo, but never found the time/motivation to do that
[18:10] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:151067bbc220: dpx: Fix rounding in 10bit total_size calculation
[18:10] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:cb85779d459c: dpx: include offset in the  total_size calculation
[18:14] <durandal_1707> hmm, why i dont get rgba in filter_frame if original video is rgba?
[18:23] <durandal_1707> another double ones are kerndeint and tinterlace
[18:23] <durandal_1707> mp=tinterlace should be removed
[18:23] <saste> durandal_1707, if you can convince michaelni to drop it
[18:24] <saste> kerndeint can be optimized, ubitux was working on that
[18:24] <saste> it should be easy, but i leave it to someone else
[18:24] <saste> tinterlace is slower than mp=tinterlace
[18:25] <saste> although the absolute difference should be meaningless
[18:25] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:ea1d8465e6ec: Fix division by 0 due to audio frame size
[18:25] Action: wm4 wonders why you would optimize these filters
[18:27] <wm4> porting vivtc instead would be _actually_ useful: http://code.google.com/p/vapoursynth/source/browse/#svn/trunk/src/filters/vivtc
[18:27] <saste> wm4, patches are... you know what
[18:28] <wm4> now I don't want to be too offensive, "patches welcome" has to be the most stupid excuse ever
[18:30] <saste> wm4, why?
[18:30] <saste> it's not an excuse, it the truth
[18:30] <saste> i won't work on that, but i won't prevent you to do so
[18:31] <saste> that's how floss works
[18:32] <wm4> I was just suggesting that porting vivtc would be a better use of your time rather than _optimizing_ barely useful filters (which probably not even mencoder users ever use)
[18:33] <J_Darnley> Don't port it!  Write a filter that passes video through Vapoursynth.
[18:33] <J_Darnley> This is just the kind of stuff that thread on the ML was talking about
[18:33] <wm4> J_Darnley: that certainly would be the best, though connecting these two APIs could be cumbersome
[18:35] <saste> J_Darnley, getting rid of mp is more fun
[18:35] <wm4> mp shouldn't have been added in the first place
[18:36] <saste> about usefulness, it's not like i *personally* need il or any other interlacing filters
[18:36] <saste> wm4, tell it to the users which use it
[18:36] <saste> also it was useful for testing/porting stuff
[18:36] <saste> that was the main reason for inclusion imo
[18:45] <saste> that said, you should add ticket to trac/upvote, so we developers know what people really want
[19:04] <Compn> durandal_1707 : going to port libvo from mplayer to ffplay ?
[19:04] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:478fc7f57bac: pictordec: fix cga palette index limit
[19:07] <Compn> wm4 / J_Darnley : why not rally vapoursynth and avisynth together to tie into ffmpeg ?
[19:07] <Compn> surely theres plenty of *synth devs to make a measley wrapper
[19:09] <Compn> michael did mp wrapper in less than a day ...
[19:10] <Compn> of course
[19:10] <Compn> you could also bribe michael into doing it :)
[19:12] <durandal_1707> wrappers are bad
[19:12] <Compn> i dont mean wrapper
[19:12] <Compn> maybe i do
[19:12] <Compn> external wrapper 
[19:12] <durandal_1707> same
[19:12] <Compn> whats better? native api ?
[19:12] <Compn> or porting everything to ffmpeg ?
[19:13] <Compn> (or are you talking about libvo ? )
[19:13] <durandal_1707> Compn: there is libavdevice which is actually libavformat
[19:16] <Compn> what
[19:17] <Compn> oh libavdevice handles hardware and such ?
[19:17] <Compn> so what are you saying ?
[19:17] <Compn> sorry, i'm kind of dumb sometimes :)
[19:31] <durandal_1707> Compn: answer to question about porting libvo
[19:31] <durandal_1707> it is bad idea to port it to ffplay
[19:34] <saste> who can help with matroskaenc?
[19:43] <michaelni> saste, what exactly ?
[19:43] <michaelni> about what 
[19:49] <durandal_1707> saste: having hwaccel pix fmts in supported inputs/outputs of filter is nonsense
[19:58] <durandal_1707> saste: why has_alpha have double !! in vf_il
[19:58] <durandal_1707> it is not needed at all
[19:58] <saste> durandal_1707, flags & ALFA is a mask
[19:58] <saste> I need to convert to 1/0
[20:00] <saste> michaelni, what are this cuepoints in matroska?
[20:01] <saste> i have a (broken) file duplicated timestamps
[20:01] <saste> i suspect this is the problem
[20:01] <saste> it is trying to put too many entries into the same cuepoint, and asserting
[20:14] <nevcairiel> cuepoints are a seeking index
[20:14] <nevcairiel> more or less
[20:22] <saste> in mkv_write_cues() is is setting an expectedsize in start_ebml_master
[20:22] <saste> which is not honored, and which leads to an assert failure
[20:29] <saste> michaelni: in other words, the MAX_CUEPOINT_SIZE is not assuming duplicated timestamps in the same stream
[20:30] <saste> next question is, how to deal with that?
[20:31] <saste> I could assume unexpected size, and increase filesize
[20:32] <saste> now i'm not even sure how matroska copes with duplicated timestamps
[20:36] <michaelni> saste, assuming duplicate timestamps arent disallowed how would they be usefull in the cue list ?
[20:36] <michaelni> if you have 2 timestamps of 123 and you want to seek to 123 ...
[20:37] <saste> michaelni, so what do you suggest to do?
[20:37] <saste> the best thing would be to avoid to write duplicated cue points (for the same stream)
[20:37] <saste> so we don't hit the bytesize limitation
[20:38] <saste> do you agree?
[20:41] <saste> michaelni, ^^
[20:45] <saste> durandal_1707, i can exclude the hwaccell from the il list, even if this is not going to do any difference
[20:45] <michaelni> saste, yes, i agree dont wrote duplicate cue points
[20:47] <durandal_1707> saste: doing it at higher level would be even better, even if lavfi will never use such formats
[20:48] <durandal_1707> the noise filter makes it possible to select noise generation parameters per plane
[20:48] <saste> durandal_1707, allright, direct commit or do you want to see the patch?
[20:48] <durandal_1707> adding opts for all planes is pita
[20:48] <saste> durandal_1707, in case of noise, i think it is useful for testing purposes
[20:49] <saste> otoh the user should be able to specify with a single option the value for all the planes
[20:49] <saste> and btw same for il
[20:49] <durandal_1707> saste: it is usefull not only for testing but is there better way to handle this?
[20:50] <durandal_1707> i start to hate opts tables. they clearly needs some improvement, like shortname or something....
[20:50] <durandal_1707> instead of duplicating each and every line
[20:51] <durandal_1707> so it is 6*2*4 = 48 entries
[20:51] <saste> durandal_1707, you can use macros
[20:52] <durandal_1707> macros just hide issue
[20:52] <saste> it's not like you need duplicated-similar options in most components
[20:52] <saste> use them with caution
[20:52] <durandal_1707> there should be type of list item
[20:53] <durandal_1707> you set number of entries in list and size of each entry
[20:54] <saste> for the constants?
[20:54] <durandal_1707> no, for 0/1 and int
[20:55] <durandal_1707> i could use AV_OPT_TYPE_BINARY
[20:56] <durandal_1707> or actually not, but something similar
[20:57] <durandal_1707> actually 0/1 items are just flags
[20:58] <durandal_1707> and avopt don't care for this
[20:58] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:f28043d0a34a: tiff: check for failure in search_keyval()
[20:58] <cone-23> ffmpeg.git 03Michael Niedermayer 07master:0e3dacb11eac: tiff: dont leave geotag_count in an invalid state on errors.
[20:58] <durandal_1707> or actually does .....
[21:00] <durandal_1707> oh so i get now 4*(1+1) = 8, i can handle this
[21:04] <cone-23> ffmpeg.git 03Stefano Sabatini 07master:f0703b6cba65: lavd/v4l2: fix error logic when enumerating standards
[21:04] <cone-23> ffmpeg.git 03Stefano Sabatini 07master:6b171f046c6f: lavf/matroskaenc: provide more feedback in case of unsupported codec
[21:04] <cone-23> ffmpeg.git 03Stefano Sabatini 07master:3acaea26e5b6: lavfi/il: filter out hwaccels formats from the list of supported formats
[21:41] <michaelni> an3k / Guest9410, why do you constantly change nick back and forth ?
[21:42] <saste> he can't decide
[22:16] <Compn> michaelni : hes got nick protect on, and since hes not here to identify, chanserv changes his nickl
[23:19] <durandal_1707> Compn: buy how do you explain periodic changes?
[23:27] <Compn> durandal_1707 : irc client script auto changes nick or maybe bnc script etc
[23:28] <Compn> didnt mean to do that, oops
[23:28] <Compn> ehe
[23:40] <durandal_1707> michaelni: fine to relicense noise to lgpl?
[00:00] --- Mon Feb 11 2013