[Ffmpeg-devel-irc] ffmpeg-devel.log.20130514

Wed May 15 02:05:02 CEST 2013

[00:10] <michaelni> j-b, that w32threads code originates from / is based on x264, its author (Steven Walters  kemuri9 at ... ) might be best qualified to review
[00:11] <Skyler_> bugmaster might be vaguely aware as well because he knows windows code
[00:11] <j-b> both don't seem to be on IRC...
[00:11] <j-b> I'll go through the ml then.
[00:11] <Daemon404> kem doesnt show up much anymore
[00:11] <Daemon404> i see him sometimes on rizon
[00:11] <j-b> michaelni: thx.
[00:11] <j-b> gn
[01:10] <burek> hmm, i tried to checkout git://ffmpeg.org/ffmpeg-web but i dont see any familiar src files that produce the current pages on ffmpeg.org.. did something change recently regarding that?
[01:11] <ubitux> make && firefox htdocs/index.html
[01:11] <ubitux> (or something like that)
[01:13] <burek> i mean, i tried to grep for some keywords that should be there (like names of some filters) and i ended up with nothing.. now, looking at Makefile, i see it only (con)cats some stuff, but nothing too much, so i doubt it will automagically create all those filter documentation, but i'll try anyway
[01:17] <ubitux> ah, these are generated from the source repository
[01:17] <burek> oh, i see
[01:17] <burek> thanks, ill try with that :)
[01:18] <llogan> burek: why is "fmpeg.org/
[01:18] <ubitux> i don't know if there is any helper to copy the generated doc from the ffmpeg repository to the the web dirs
[01:18] Action: llogan learns to paste
[01:18] <burek> :)
[01:19] <ubitux> but make doc && cp ffmpeg/doc/*.html ffmpeg-web/htdocs or something should do the trick
[01:19] <llogan> ..."Creating x264 video for html5 from pictures" a sticky topic in your forum?
[01:19] <burek> yeah, it makes sense to keep the docs in one place, ubitux
[01:19] <burek> llogan, umm.. i dunno.. i dont even recall of creating that topic
[01:20] <llogan> it is always listed first despite its age
[01:20] <ubitux> burek: we can't really move the doc from the ffmpeg source to the ffmpeg-web repo
[01:20] <ubitux> if that's what you meant
[01:20] <ubitux> (because packagers use the ffmpeg source repo to generate ffmpeg and its man pages)
[01:21] <burek> ubitux, it was separated before, i think, so i just commented it makes sense to keep it in one repo, not to duplicate stuff
[01:21] <burek> (man and www)
[01:22] <burek> llogan fixed :)
[01:22] <burek> i didnt even notice that
[01:22] <burek> btw, do you have an admin there?
[01:23] <llogan> i don't think so
[01:24] <burek> you have now :)
[01:24] <burek> also, if anyone else is interested in helping on the forum, please let me know
[01:25] <llogan> thanks, and good job at keeping down the spam. i haven't seen one yet.
[01:25] <burek> so i can assign the proper access rights to make things easier
[01:25] <burek> llogan :beer: :) well, i'm just happy it ended with just the dummy auto-search-and-spam bots
[01:26] <burek> on the other hosting server, i've got a kid so persistent and annoyin
[01:26] <burek> i had to ban half of the chinese subnet to keep his botnet away..
[01:55] <cone-303> ffmpeg.git 03Dale Curtis 07release/1.0:e489d5c54d3f: Fix heap-buffer-overflow in matroska_parse_block
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:f072b69f39a3: xbmdec: fix off by one error in scanf()
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:6f2f754d97d9: src_movie: fix scanf string
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:e32acc455db2: avcodec/lcldec: Check that dimensions are a multiple of the subsample factors
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:d8c4ee6feaef: avcodec/mpegvideo: Fix block height for lowres 3 interlaced blocks
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:4a442ac117c4: avcodec/mpegvideo: Fix edge emu with lowres
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:ee57bfc47a98: mjpegdec: fix overlapping memcpy with upscale_v
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:cd24fdad4707: mmvideo/mm_decode_inter: check horizontal coordinate too
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:1d285db8a3a8: mmvideo/mm_decode_intra: check horizontal coordinate too
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:d814b9b51ddd: vmdav: Try to fix unpack_rle()
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:52fa7a860318: avutil/log: Fix context pointer used for get_category()
[01:55] <cone-303> ffmpeg.git 03Michael Niedermayer 07release/1.0:c446bc31f501: avutil/intfloat_readwrite: avoid comparission with INFINITY, use isinf()
[02:01] <ubitux> burek: huh??
[02:01] <ubitux> burek: -vf fps=25 should work
[02:01] <ubitux> if it doesn't, then there is a bug
[02:02] <burek> ubitux http://ffmpeg.gusari.org/viewtopic.php?f=11&t=915
[02:02] <burek> search for Error initializing filter 'fps' with args '25'
[02:06] <durandal11707> is there way to open http stream wiht (a)movie?
[02:07] <ubitux> burek: honestly, it works
[02:07] <ubitux> i've just tried it
[02:07] <burek> ubitux, ok, then just discard the patch :)
[02:07] <durandal11707> what patch?
[02:07] <burek> that was the sole reason i made it :)
[02:08] <ubitux> adding examples is fine
[02:08] <ubitux> but you can definitely omit the "fps="
[02:08] <burek> durandal11707, maybe something like curl | ffmpeg -vf amovie=stdin
[02:08] <ubitux> burek: try it; ./ffplay -f lavfi testsrc -vf "yadif,scale=iw/2:-1,fps=25"
[02:08] <burek> ubitux, ok, i believe you
[02:08] <ubitux> i'm just using the guy command
[02:09] <durandal11707> burek: ugly....., it should work normal way, if not its bug
[02:09] <burek> probably 0.11.1 was too old for that syntax or something
[02:09] <ubitux> ah, that is true
[02:09] <ubitux> 0.11.1 hadn't shorthand defined for this filter (it's a libav filter, they didn't have shorthands)
[02:10] <burek> mistery resolved :)
[02:10] <ubitux> yup
[02:10] <ubitux> good job
[02:10] <ubitux> :)
[02:10] <burek> :)
[02:22] <durandal11707> burek: awesome, we certainly need more examples
[02:22] <durandal11707> but for next case see other examples, you can use @itemize, @item ...
[02:35] <cone-303> ffmpeg.git 03burek 07master:56ba3312574f: doc/filters: add some examples for the fps filter
[03:39] <llogan> no more system install
[03:40] <llogan> feel free to edit or ping me if anyone sees anything particularly retarded
[03:41] <llogan> though i haven't decided how to handle man pages yet
[03:47] <ubitux> it's a bit ugly to clutter the home directory with all kind of source repo but well..
[03:55] <llogan> ubitux: where would you prefer it?
[03:56] <llogan> i wanted to avoid sudo as much as possible
[03:58] <llogan> i guess i can put sources in ~/ffmpeg_build as well.
[04:10] <ubitux> i meant all the source dirs in a separate place in the home
[04:10] <ubitux> i would have them in a common dir
[04:10] <ubitux> src/ build/ ff/ whatever/
[04:14] <llogan> ok. good point. i'll update it tomorrow.
[04:15] Action: llogan has reached Ubuntu tolerance level for the day
[04:24] <cone-303> ffmpeg.git 03burek 07fatal: ambiguous argument 'refs/tags/n1.0.7': unknown revision or path not in the working tree.
[04:24] <cone-303> Use '--' to separate paths from revisions
[04:24] <cone-303> refs/tags/n1.0.7:HEAD: doc/filters: add some examples for the fps filter
[10:32] <cone-108> ffmpeg.git 03Diego Biurrun 07master:46bb456853b1: x86: dsputil: Refactor pixels16 wrapper functions with a macro
[10:32] <cone-108> ffmpeg.git 03Michael Niedermayer 07master:eda9d97b7a1d: Merge commit '46bb456853b197f4562de7acf5d42abf11ded9be'
[11:07] <nevcairiel> colorchannelmixer test is still broken on various systems
[11:08] <nevcairiel> odly enough, not on msvc :P
[11:12] <nevcairiel> or its the perms=random causnig the breakage in some cases
[11:28] <cone-108> ffmpeg.git 03Diego Biurrun 07master:245b76a10858: x86: dsputil: Split inline assembly from init code
[11:28] <cone-108> ffmpeg.git 03Michael Niedermayer 07master:b2da63db507b: Merge commit '245b76a108585b6fb52eebc2626c472d6fa530dc'
[12:21] <cone-108> ffmpeg.git 03Diego Biurrun 07master:1399931d07f0: x86: dsputil: Rename dsputil_mmx.h --> dsputil_x86.h
[12:21] <cone-108> ffmpeg.git 03Michael Niedermayer 07master:a887372109a2: Merge commit '1399931d07f0f37ef4526eb8d39d33c64e09618a'
[12:25] <durandal_1707> michaelni: gmake: *** No rule to make target ibavcodec/x86/hpeldsp_init.o'.  Stop.
[12:26] <cone-108> ffmpeg.git 03Martin Storsjö 07master:05165c2f7a24: configure: Disable PIC for cygwin just as for win32/mingw
[12:26] <cone-108> ffmpeg.git 03Michael Niedermayer 07master:509cc1e8837e: Merge remote-tracking branch 'qatar/master'
[12:26] <durandal_1707> looks like removing *.d files fixed it
[12:59] <cone-108> ffmpeg.git 03Michael Niedermayer 07master:f268727a8fa2: fate: change colorchannelmixers coefficients to be less unstable
[12:59] <cone-108> ffmpeg.git 03Michael Niedermayer 07master:6b13f54262b6: vf_colorchannelmixer: round lut entries to nearest
[13:58] <burek> is ffmpeg able to "play" sdp files, like: ffmpeg -i bla.sdp ...
[13:58] <burek> s/ffmpeg/ffplay
[14:08] <durandal_1707> burek: "ffmpeg -formats" lists sdp
[14:12] <burek> oh cool, so it does support it :) thanks :)
[14:12] <burek> i didn't know it's a format, i thought sdp is just a descriptor file
[16:39] <cone-719> ffmpeg.git 03Michael Niedermayer 07master:dc6f1a8dda00: avformat/srtdec: Fix pointer corruption
[17:23] Action: ubitux still wonders how worse the default libvpx wrapper settings could be
[17:36] <durandal_1707> lol, lets play ping pong game with cehoys once more
[17:40] <nevcairiel> the ticket is about converting XYZ with RGB, which works just fine now, anything else should go in a new ticket, is all
[17:41] <saste> ubitux: ping on zmq?
[17:42] <ubitux> saste: yup, in a few minutes
[17:44] <cone-719> ffmpeg.git 03Paul B Mahol 07master:811b17fbad07: lavfi: cosmetics: fix vertical alignment for pads in some filters
[18:29] <BuxiNess> XYZ to RGB,  is not really a colorspace transformation. Its actuallly a stdr 16 bit tiff img, with R=X, G=Y, and B=Z. so its just a matrix and gamma correction who changes the value of the 3 channels
[18:46] <cone-719> ffmpeg.git 03Stefano Sabatini 07master:e65c4a233047: doc/outdevs: fix typos (devices -> device)
[18:58] <cone-719> ffmpeg.git 03Michael Niedermayer 07master:e751481cd82d: swscale/output: silence compiler warnings about uninitialized variables
[19:43] <cone-719> ffmpeg.git 03Paul B Mahol 07master:b24530019f28: lavfi/blend: timeline support
[19:44] <cone-719> ffmpeg.git 03James Almer 07master:dc7656279ae1: lavu/hash: Fix name of SHA224
[20:12] <cone-719> ffmpeg.git 03Stefano Sabatini 07master:7ddb0ef9af50: lavfi: add zmq filters
[20:12] <cone-719> ffmpeg.git 03Stefano Sabatini 07master:2a1b7dee6d03: tools: add zmqsend tool, useful to test the zmq filters
[22:01] <durandal_1707> new filter every 12h - that is too much
[22:02] <llogan> blame durandal_1707
[22:02] <llogan> no example for colorlevels?
[22:03] <cone-719> ffmpeg.git 03Michael Niedermayer 07master:640e524ff6e1: libavfilter/Makefile: unbreak make alltools
[22:03] <durandal_1707> you need to inspect histogram with -vf histogram and adjust values according to this
[22:03] <durandal_1707> there is also 'auto' when you set values to < 0, but that is not perfect ...
[22:04] <llogan> an example would still be useful. especially if histogram and "auto" is mentioned
[22:05] <durandal_1707> its funcionality is subset of curves filter (its just more user friendly than curves)
[22:06] <durandal_1707> (without 'auto' mode ...)
[22:06] <durandal_1707> its somehow related to histeq filter
[22:06] <durandal_1707> its just dumb normalization
[22:21] <saste> durandal_1707, i lost track of the filters to review
[22:21] <saste> and there are also ubitux filters (ow, lut3d)
[22:22] <ubitux> dctdnoiz too
[22:22] <ubitux> which i'll push soon
[22:24] <saste> ubitux, what's the status of ow
[22:25] <saste> dctdnoiz => dctdenoise?
[22:25] <ubitux> <@saste> ubitux, what's the status of ow // i need to come up with a better name
[22:25] <saste> owdenoise?
[22:26] <ubitux> <@saste> dctdnoiz => dctdenoise? // i don't want to have the exclusiveness of the dct denoising, so i'm using a random name
[22:26] <saste> and i agree, we can forget the old name and no one will ever complain
[22:26] <ubitux> sure
[22:26] <saste> dctubidenoise
[22:26] <ubitux> i'm fine with owdenoise
[22:26] <ubitux> i like dctdnoiz! :(
[22:27] <saste> look i'm not a fun of random contractions, but name you son as you like
[22:28] <ubitux> there are all kind of ways to do dct denoising, so i don't want to use a generic name
[22:28] <saste> ubitux, so it's just a naming issues?
[22:28] <ubitux> you mean for owdenoise?
[22:28] <saste> yes
[22:28] <ubitux> well i guess that's all yes
[22:29] <ubitux> i was just waiting a little for reviews
[22:29] <saste> i'm for owdenoise, short and clear
[22:29] <ubitux> sure ok fine
[22:29] <ubitux> i'll drop mp=ow at the same time
[22:29] <saste> i'll do a quick review, but i'm not going to say nothing intelligent about it
[22:35] <saste> ubitux: know something about Texas Instruments Cinecanvas?
[22:35] <ubitux> nope
[22:35] <ubitux> never heard of it
[22:35] <ubitux> why?
[22:35] <saste> it's another XML based subtitles format, with support for effects (fade in/out)
[22:36] <ubitux> oh no.
[22:37] <saste> grep for texas instruments and cinecanvas and you should find the spec
[22:37] <saste> i got a consulting request about it
[22:37] <saste> i don't even know if it is possible to hack it into ffmpeg
[22:38] <saste> what's sure is that it would require a serious XML parser
[22:38] <ubitux> ass has fade capabilities, so it should be possible
[22:39] <ubitux> there is also some smil helpers to parse shit like sami, rt etc
[22:39] <ubitux> it might work for you as well
[22:39] <ubitux> feel free to ask me for help, but please don't tell the guy to contact me
[22:39] <ubitux> i'm *not* willing to parse xml
[22:40] <saste> i wonder if we could create a smil demuxer, by creating a filtergraph and injecting commands into it
[22:40] <saste> ubitux, did you already considered to use some xml parsing library?
[22:41] <ubitux> yes
[22:41] <ubitux> i tried with libexpat
[22:41] <ubitux> but it didn't work, obviously
[22:41] <ubitux> because sami & ci. were mainly randomly broken html
[22:41] <ubitux> and libexpat is kinda strict
[22:41] <ubitux> if your xml is well parsed it's fine, otherwise...
[22:42] <ubitux> feel free to write a ffxml and use it
[22:43] <ubitux> and actually most of the time with subtitles it's not xml
[22:43] <ubitux> it's just... html-like tags
[22:44] <saste> cinecanvas should be sane, at least from that point of view
[22:44] <ubitux> are you sure ff_smil_* functions won't be enough for you?
[22:45] <ubitux> eventually adjusting the function to be slightly more reliable
[22:46] <saste> ubitux, did you ask michaelni about ow relicensing?
[22:47] <ubitux> nope
[22:47] <ubitux> i don't care much
[22:48] <durandal_1707> i asked about noise and iirc never got any reply
[22:49] <ubitux> also since diego did all kind of cosmetics to original mplayer, i will likely have to contact him
[22:49] <ubitux> last times i got ignored so no thanks
[22:49] <saste> ubitux, so keep original licensing (GPL)
[22:49] <ubitux> yes
[22:49] <saste> and good luck if someone needs an LGPL version
[22:50] <ubitux> i'll mark in the commit desc that i don't care about a relicensing
[22:50] <saste> i wonder what happened of yadif relicensing
[22:50] <ubitux> etoomuchignore likely
[22:50] <durandal_1707> ubitux: asking licensing for cosmetic changes is nonsense
[22:51] <ubitux> http://pastie.org/7908579
[22:51] <saste> ubitux, that's what happens when you're in a condition when a single one amongst many can block things
[22:51] <ubitux> anyway, i don't want to ask for a relicensing for something i didn't really contribute to
[22:51] <durandal_1707> i think i could relicense stereo3d (let RMS sue me) as its trivial, and not trivial code was added by me
[22:52] <cone-719> ffmpeg.git 03Michael Niedermayer 07master:5d22ac488b4a: libx264: zero x264_picture before use.
[22:52] <cone-719> ffmpeg.git 03Vignesh Venkatasubramanian 07master:48de04f4ec08: avcodec/avpacket: Refactoring copy_side_data into a separate function
[22:52] <saste> funny how we will have tinterlace silly code licensed as gpl forever
[22:53] <durandal_1707> ubitux: also you probably changed code cosmetics again....
[22:53] <durandal_1707> and you are against lgpl joinfields
[22:53] <ubitux> wut?
[22:54] <durandal_1707> see related thread, and irc logs
[22:55] <ubitux> i have no idea what you are talking about
[22:56] <durandal_1707> http://lists.ffmpeg.org/pipermail/ffmpeg-devel/2013-April/142567.html
[22:59] <ubitux> where did i say i was against joinfields or lgpl ?
[22:59] <ubitux> also, what cosmetics are you talking about?
[22:59] Action: ubitux has absolutely no clue tonight
[22:59] <durandal_1707> ubitux: you where talking about being contacting don diego
[23:02] <saste> fps=25 is not valid syntax?
[23:03] <ubitux> not in 0.11
[23:03] <ubitux> there was no short hands
[00:00] --- Wed May 15 2013