[Ffmpeg-devel-irc] ffmpeg-devel.log.20130509

[burek]

[00:00] <michaelni> FF_CEIL_RSHIFT() OTOH might be understood by a random reader as it refers just to iso C terminology and r close to "shift" likely would get correctly guesseed to right
[00:01] <ubitux> fine with me
[00:03] <cone-463> ffmpeg.git 03Michael Niedermayer 07master:df9f9caba58a: avfilter: Make enabled/disabled state available as a field of the AVFilterContext
[00:03] <cone-463> ffmpeg.git 03Michael Niedermayer 07master:1e1015f330e3: avfilter/af_apad: Support AVFILTER_FLAG_SUPPORT_TIMELINE
[00:03] <cone-463> ffmpeg.git 03Michael Niedermayer 07master:410905f7fd0d: ffmpeg: use av_freep() for forced_kf_pts
[00:03] <cone-463> ffmpeg.git 03Michael Niedermayer 07master:33ff943f70f0: ffmpeg: add -apad option which injects a apad filter only in the presence of video streams
[00:03] <cone-463> ffmpeg.git 03Michael Niedermayer 07master:e4723a82f708: avfilter: fix 2 warnings about discarding const
[00:06] <ubitux> ah, not a bad idea
[00:06] <ubitux> it could help simplifying overlay
[00:08] <ubitux> michaelni: maybe the passthrough callback could be removed then
[00:08] <ubitux> and replaced with another flag
[00:08] <ubitux> like "AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL"
[00:10] <ubitux> #define AVFILTER_FLAG_SUPPORT_TIMELINE_INT AVFILTER_FLAG_SUPPORT_TIMELINE | (1 << 17)
[00:10] <michaelni> maybe, i dont know which way its better
[01:26] <ubitux> michaelni: https://github.com/ubitux/FFmpeg/commit/61bd2b341ea579c55d85b823fb53aaa8139d7db9
[01:26] <ubitux> that's definitely sexier
[01:27] <ubitux> also allows: https://github.com/ubitux/FFmpeg/commit/c44bdcf1998a9193cb33aad4e0d8b1ff7ff2bb6a
[01:43] <llogan> looks like fflogger forgot his meds today
[02:06] <ubitux> "The alpha channel is encoded separately and the output is placed in AVPacket's side_data"
[02:06] <ubitux> huh??
[03:01] <highgod> Hi, all. I want to ask two questions, I have use ffmeg dshow to capture the desktop. First is, does ffmpeg has comand line to use dshow to capture a specific window? The second is, dose it can capture a specific area of desktop? Thanks
[07:26] <burek> hi, can anyone, familiar with the topic "HLS with encrypted stream", help to answer this: http://ffmpeg.gusari.org/viewtopic.php?f=11&t=897
[07:28] <burek> also, this one, about using vf inside preset files: http://ffmpeg.gusari.org/viewtopic.php?f=11&t=917
[11:20] <cone-764> ffmpeg.git 03Diego Biurrun 07master:87614667606b: x86: dsputil: Move ff_pd assembly constants to the only place they are used
[11:20] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:3469c24a1084: Merge commit '87614667606b42476f9017d79faf12b45a0bd77c'
[11:34] <cone-764> ffmpeg.git 03Diego Biurrun 07master:ed880050edf0: x86: dsputil: Group all assembly constants together in constants.c
[11:34] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:5747d835c775: Merge commit 'ed880050edf061b38d3e39e25657c59ad9108b27'
[11:42] <cone-764> ffmpeg.git 03Diego Biurrun 07master:71469f3b636f: x86: dsputil: Move constant declarations into separate header
[11:42] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:164899c6e858: Merge commit '71469f3b636fbe06b6aca5933f9fdebddd8d5f57'
[11:51] <cone-764> ffmpeg.git 03Diego Biurrun 07master:3d40c1ee742d: x86: dsputil: Move TRANSPOSE4 macro to the only place it is used
[11:51] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:bda5487d92fc: Merge commit '3d40c1ee742db5f13ebcf53c2d1fa4bf4f39bcd2'
[11:55] <cone-764> ffmpeg.git 03Diego Biurrun 07master:6e9f8d6a7d73: x86: vf_yadif: Remove stray dsputil_mmx #include
[11:55] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:696f5f98e282: Merge commit '6e9f8d6a7d7392a236df19fef6f4eba41f18167e'
[12:36] <cone-764> ffmpeg.git 03Diego Biurrun 07master:f243bf7aa27f: x86: dsputil: Remove unused argument from QPEL_OP macro
[12:36] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:769efe56b145: Merge remote-tracking branch 'qatar/master'
[12:43] <durandal_1707> michaelni: there is no way to workaround issue for le/be pixformats and pixfmts filter tests?
[12:43] <durandal_1707> because filter would for highger bitdepths pick native endian format
[12:49] <cone-764> ffmpeg.git 03Paul B Mahol 07master:d1cb559a6a35: fate: add test for lut filter
[12:54] <ubitux> durandal_1707: do you need review of anything?
[12:55] <durandal_1707> ubitux: chapter encoding for id3v2
[12:56] <durandal_1707> test with some program that does not use lavf but can read & show chapters
[12:56] <ubitux> anything in lavfi ?
[12:57] <durandal_1707> want to extend colorkeymask ?
[12:57] <ubitux> what's the usage of such filter?
[12:57] <ubitux> s/usage/common case/
[12:58] <durandal_1707> http://en.wikipedia.org/wiki/Chroma_key
[12:58] <ubitux> oh
[12:58] <ubitux> that will be pretty nice
[13:02] <durandal_1707> well, i did not compared but its current funcionality should be better or same as avisynth internal filter of same name
[13:02] <durandal_1707> i also added yuv support locally
[13:04] <ubitux> should i wait for the next version for a review?
[13:05] <durandal_1707> nope, you are free to extend it and/or write documentation
[13:05] <ubitux> i'm busy with my own filters :)
[13:05] <ubitux> i was just asking if you need a ok or anything for some lavfi-related work
[13:05] <ubitux> can't help you much with id3 chaptering
[13:06] <durandal_1707> nobody here have amarok/songbird installed?
[13:06] Action: ubitux uses mpd
[13:06] <durandal_1707> is vlc using lavf for id3 tags?
[13:07] <funman> no
[13:07] <funman> we use taglib
[13:11] <durandal_1707> ubitux: you can comment/try avectorscope (added zoom locally)
[13:12] <ubitux> is your remote up-to-date?
[13:16] <ubitux> oh it's pretty with colors
[13:25] <ubitux> durandal_1707: the zoom creates some kind of "maggots" effect, it's fun
[13:29] <durandal_1707> what ?
[13:32] <ubitux> my imagination certainly :)
[13:32] <ubitux> if you zoom, there is some kind of little arc circle
[13:33] <ubitux> not strong, but enough with the motion to give a swarm feeling
[13:33] <durandal_1707> you mean different color?
[13:34] <durandal_1707> what zoom ?
[13:39] <ubitux> durandal_1707: http://lucy.pkh.me/avectswarm.webm
[13:39] <ubitux> it's a very mild effect, and most likely in my head ;)
[13:41] <ubitux> (zoom=10)
[13:42] <durandal_1707> it happens with zoom 1 too, its because fade/contrast differs, creating such effect
[13:45] <ubitux> why +1 for x?
[13:48] <durandal_1707> because value on the left is in -1, +1 range
[13:54] <durandal_1707> funman: and vlc displays id3 chapters?
[13:54] <funman> never heard of id3 chapters so no clue :D
[13:55] <funman> i doubt it does
[13:56] <durandal_1707> http://forum.videolan.org/viewtopic.php?f=7&t=104274
[13:57] <durandal_1707> i downloaded songbird for windows (run under wine) and it looks to not show chapters too
[14:02] <funman> https://trac.videolan.org/vlc/ticket/7485 apparently someone wanted to support it
[14:05] <durandal_1707> the files linked does not have text null terminated, as described in spec
[14:09] <funman> can you add a good sample ?
[14:09] <durandal_1707> one created by ffmpeg?
[14:13] <durandal_1707> looks like only apps that can display it need mono or java
[14:14] <durandal_1707> but taglib should export chapters as CHAP frames
[14:14] <durandal_1707> so it should be trivial to add in vlc
[14:21] <funman> i'd guess especially since mkv/mp4 chapters are already supported
[14:22] <durandal_1707> https://github.com/taglib/taglib/issues/81
[14:27] <Compn> what?
[14:32] <Compn> ehe
[14:32] Action: Compn comes up with solution for michaelni and fate crc/md5
[14:47] <Compn> michaelni : does fate disable video (-vn) for audio tests and audio (-an) for video testing ?
[14:48] <ubitux> why does "MurMur" makes me think of "Moon Moon"?
[14:49] <Compn> course i didnt see much speed up with -f framecrc -an anyways...
[14:50] <michaelni> Compn, i think so but didnt check, also running both v& a can be a good thing as more is tested
[14:53] <Compn> michaelni : i'm just looking for ways to save time in fate runs :)
[14:53] <Compn> maybe have a fate-fast with these alternates
[14:54] <Compn> like fate-fast would only tests the first 10 frames in each sample, instead of the entire sample?
[14:55] <Compn> unless the entire ref is checked... maybe too difficult
[14:56] <ubitux> sounds like a maintainance burden
[14:57] <ubitux> i wonder if moving lavf-regressions.sh to .mak wouldn't help
[14:57] <ubitux> (all the successive if in that file for each test might be a bit too much)
[14:59] <ubitux> also, md5.sh might be slow to call
[14:59] <Compn> what about optimizing md5.c ?
[15:00] <Compn> i dont see any yasm, cpu opts, gpu offloading, etc
[15:00] <Compn> :)
[15:01] <ubitux> we could switch to sha and use git implementation
[15:02] <michaelni> we also could use md5 or sha1 with fewer rounds
[15:02] <michaelni> but i doubt that would win the speed record either
[15:05] <Compn> are you running complete fate tests when the code you changed only affects video decoding ? or is it compartmentalized enough already ?
[15:05] Action: Compn afk
[15:32] <michaelni> Compn, i tend to run full fate tests
[15:49] <funman> michaelni: hm swscale's init_hscaler_mmxext is problematic for Windows Store
[15:50] <funman> VirtualAlloc (mmap) is forbidden and i doubt there's another way to have self modifying code
[15:50] <funman> that's why mozilla complained Microsoft was doing a favor for IE by forbidding other browsers to use javascript JIT
[15:54] <michaelni> funman, only the fast bilinear scaler uses self modifying code, this shouldnt be hard to disable ...
[15:56] <funman> add some configure option ?
[15:56] <michaelni> possible
[15:56] <michaelni> if it cant be autodetected
[15:57] <michaelni> i dont know if theres anything special set when compiling for "Windows Store"
[15:59] <funman> hmm well yes and no
[16:00] <funman> in our case we don't set anything special
[16:00] <funman> there is some other code we need to remove also
[16:00] <funman> like GetProcessAffinityMask
[16:01] <funman> i think j-b is working on this
[16:23] <michaelni> ok, if someone needs help with making that optional in sws ... but i think it should be fairly straightforward
[16:24] <durandal_1707> what to make optional in sws?
[16:28] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:3d66ce17b7f8: tools: Add adler32 checksum tool
[16:28] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:a11593189b26: ffadler: add multifile support
[16:30] <michaelni> durandal_1707, self modifying code as windows store doesnt like it
[16:32] <durandal_1707> sws have that?
[16:34] <funman> durandal_1707: see call to mmap in init_hscaler_mmxext
[16:39] <durandal_1707> michaelni: am I correct that there is not way for user(via ffmpeg) to add chapters and chapter metadata?
[16:44] <JEEBsv> I think there are ways to add metadata?
[16:44] <JEEBsv> saw on the documentation pages
[16:45] <durandal_1707> but not chapter one...
[16:50] <JEEBsv> yes, true
[16:50] <JEEBsv> missed that chapter word in the second one
[16:52] <ubitux> michaelni: any comment on the timeline patchset?
[16:53] <cone-764> ffmpeg.git 03Paul B Mahol 07master:88e0e2054d91: lavfi: remove alphaextract special code
[16:54] <cone-764> ffmpeg.git 03Michael Niedermayer 07master:d9cb1e0e15e5: ffadler: fix duplicate \n on errors
[16:54] <ubitux> durandal_1707: and the dep?
[16:55] <durandal_1707> it does not need one
[16:58] <michaelni> ubitux, needing 2 flags is somehow comlpex otherwise i see nothing bad on the patch
[17:00] <cone-764> ffmpeg.git 03Clément BSsch 07master:570d63eef3f9: lavu: add FF_CEIL_RSHIFT and use it in various places.
[17:00] <cone-764> ffmpeg.git 03Clément BSsch 07master:f0250cc46366: lavfi/overlay: simpler up-rounded w/h computations.
[17:00] <cone-764> ffmpeg.git 03Clément BSsch 07master:4db84bac1353: lavfi/opencl: replace SHIFTUP with FF_CEIL_RSHIFT.
[17:00] <ubitux> michaelni: do you have a better suggestion?
[17:02] <ubitux> btw, it seems we do link->[wh] >> xsub in various filters
[17:02] <ubitux> michaelni: mmh i'll push a fix in 1.2
[17:03] <durandal_1707> ubitux: what filters?
[17:06] <ubitux> durandal_1707: git grep 'link->[wh].*sub'
[17:06] <ubitux> possibly variants with width/height and log2_chroma
[17:08] <cone-764> ffmpeg.git 03Clément BSsch 07release/1.2:3d6219b6dbf4: lavfi/subtitles: zero-init AVSubtitle.
[17:08] <ubitux> ok i think that's all i want to backport :)
[17:13] <michaelni> ubitux, no, not really, the 2 flags just dont look pretty ...
[17:16] <durandal_1707> ubitux: going to fix that hsub/vsub issues?
[17:17] <ubitux> durandal_1707: maybe a few ones
[17:17] <ubitux> i need a test case
[17:17] <durandal_1707> i have one - do not use ffplay - it pads
[17:22] <cone-764> ffmpeg.git 03Reimar Döffinger 07release/0.11:d12294f447f4: matroska: set "done" only during resync fail.
[17:22] <cone-764> ffmpeg.git 03Reimar Döffinger 07release/1.0:9fef5d55538c: matroska: set "done" only during resync fail.
[17:22] <cone-764> ffmpeg.git 03Reimar Döffinger 07release/1.1:2922ab7e6fec: matroska: set "done" only during resync fail.
[17:22] <cone-764> ffmpeg.git 03Reimar Döffinger 07release/1.2:a857811c755b: matroska: set "done" only during resync fail.
[17:22] <cone-764> ffmpeg.git 03Michael Niedermayer 07release/1.2:84df0dc40bd6: avcodec/mpegvideo: Fix block height for lowres 3 interlaced blocks
[17:22] <cone-764> ffmpeg.git 03Michael Niedermayer 07release/1.2:3ca6253beb12: avcodec/mpegvideo: Fix edge emu with lowres
[17:45] <cone-764> ffmpeg.git 03Carl Eugen Hoyos 07master:43b6482d0a8f: Add fate test for solid colour lagarith frames.
[18:00] <ubitux> durandal11707: http://b.pkh.me/0001-lavfi-use-ceil-right-shift-for-chroma-width-height.patch
[18:00] <ubitux> let's break everything \o
[18:00] <ubitux> (possibly fixing issues though)
[18:04] <funman> btw the external swscale repo has been abandoned?
[18:05] <durandal11707> there was such thing?
[18:05] <durandal11707> ubitux: missing vflip
[18:05] <kierank> funman: you not mean postproc?
[18:05] <ubitux> durandal11707: it's using a weird hack so i didn't touch
[18:05] <ubitux> same for pad, where there is likely something to fix btw
[18:06] <funman> kierank: hum right, i'm confused
[18:06] <ubitux> swscale repo was outside too a while ago
[18:06] <nevcairiel> a long while now
[18:06] <kierank> oh yeah ages ago
[18:06] <kierank> 3yrs or so ago
[18:06] <funman> yeah i meant postproc indeed
[18:07] <ubitux> postproc is maintained in ffmpeg
[18:07] <durandal11707> ubitux: whatever hack does its not working for "weird" dimensions for subsamples yuvs
[18:07] <ubitux> the external library is just an excuse for the fork to drop it afaik
[18:08] <ubitux> durandal11707: possibly in another commit then
[18:08] <ubitux> cause i think this one might break a few things already
[18:09] <durandal11707> what it can break?
[18:09] <durandal11707> it should fix weird green lines with odd dimensions and normal are unaffected
[18:10] <funman> i'll wait for j-b to be back and work on a patch to disable the self-modifying code
[18:10] <funman> ./configure --disable-compiler :)
[18:10] <ubitux> we should have more tests with odd dimensions
[18:12] <durandal11707> ubitux: push/send to ml/
[18:12] <ubitux> yeah, i'll send the patch
[18:18] <ubitux> michaelni: maybe a less ugly macro could be AVFILTER_FLAG_INTERNAL_TIMELINE?
[18:18] <ubitux> i could remove the internal timeline implying the first flag too
[18:28] <durandal11707> ubitux: adding more tests is makeing michaelni wait more
[18:29] <ubitux> i'd be pleased to have a faster fate
[18:30] <ubitux> and more tests ;)
[18:31] <durandal11707> ubitux: so what i should do to get you to push that (chroma w/h) patch?
[18:31] <ubitux> test if it fixes anything maybe
[18:31] <ubitux> or encourage me by fixing lut/vflip/pad
[18:32] <ubitux> or finding moar similar issues in lavc/lavu
[18:32] <durandal11707> ubitux: that lut code is also wrong, i get pink color 
[18:33] <durandal11707> ffmpeg -f lavfi -i smptehdbars=1600x903,negate -f sdl -
[18:35] <durandal11707> hmm it may be other way around ...
[18:36] <ubitux> nothing in particular here
[18:36] <durandal11707> but il/interlace/blend/extractplanes are ok
[18:38] <durandal11707> -f sdl shows black last line, mpv shows green, so i will use mpv to test this
[18:39] <ubitux> right, i noticed the black line
[18:43] <durandal11707> ubitux: it could be bug in source filter...
[19:18] <durandal11707> ubitux: interlace have same bug, as lines get / 2 later
[20:59] <michaelni> latest news: trac.ffmpeg.org finally works, no more /trac/ffmpeg needed
[21:11] <llogan> nice.
[21:11] <llogan> any plans to update trac to 1.0?
[21:15] <ubitux> btw, fate moved to the new machine?
[21:19] <michaelni> ubitux, hmm, judging from the ip id say not yet
[21:19] <michaelni> llogan, any features in that that we would want / need ?
[21:24] <llogan> michaelni: for me, mostly a nicer looking theme to make wiki look nicer.
[21:24] <llogan> "Usability improvements for the tickets, with a better support for conflict detection and resolution" might be interesting
[21:25] <llogan> carl might like the now integrated BatchModifyPlugin
[21:25] <llogan> i guess nothing too important since you have better things to do
[22:05] <durandal11707> oh noes
[22:08] <llogan> please indicate which ticket # it is a duplicate of when closing (although it is obvious here).
[22:19] <Compn> doesnt it leave a comment ?
[22:19] <Compn> bugzilla did...
[22:25] <llogan> Compn: not that i know of but it can be manually added in the comment box when closing. i guess that's the only good thing about bugzilla.
[22:30] <michaelni> can you elaborate what bugzilla does better than trac? I dont think i understand
[22:30] <Daemon404> bugzilla is way better at being ugly
[22:30] <Daemon404> than trac
[22:38] <cone-954> ffmpeg.git 03Paul B Mahol 07master:13cb6ed3c449: takdec: simplify use of scalarproduct_int16 in decode_subframe()
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:54f390239316: rmdec: flush audio packet on seeking
[22:52] <cone-954> ffmpeg.git 03ArnoB 07release/1.2:298e03d1023d: dpxenc: fix data offset
[22:52] <cone-954> ffmpeg.git 03Dale Curtis 07release/1.2:f8d5f3dff5d6: Fix heap-buffer-overflow in matroska_parse_block
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:4d0090f90a97: xbmdec: fix off by one error in scanf()
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:1ce3f736d2aa: src_movie: fix scanf string
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:840931e76608: mjpegdec: fix overlapping memcpy with upscale_v
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:7fa6db254564: mmvideo/mm_decode_inter: check horizontal coordinate too
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:039f6921c276: mmvideo/mm_decode_intra: check horizontal coordinate too
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:0baa0a5a02e1: vmdav: Try to fix unpack_rle()
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:cc0dd86580b3: avutil/log: Fix context pointer used for get_category()
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:ec89046fa110: avutil/intfloat_readwrite: avoid comparission with INFINITY, use isinf()
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:9f7baf139f72: avutil/intfloat_readwrite: include common.h for isinf()
[22:52] <cone-954> ffmpeg.git 03Clément BSsch 07release/1.2:91f04e741096: cmdutils: avtool -> fftool (cherry picked from commit 7d8ad6c1fa11ec548fc63427656989e0e7c6af8b)
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:524d0d2cfc7b: sanm: Check dimensions before use
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:7ef2dbd2392e: avcodec/cdgraphics: check buffer size before use
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:7ee5e97c46e3: gifdec: reset previous Graphic Control Extension disposal type
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:c5948b472b7d: update for 1.2.1
[22:52] <cone-954> ffmpeg.git 03Michael Niedermayer 07release/1.2:f166a02b67a2: exr: silence warning "libavcodec/exr.c:351:9: warning: variable ret set but not used"
[22:59] <saste> bugzilla is very good
[22:59] <saste> at confusing people
[22:59] <saste> it prevents bad user reports, since only masochistic users are able to  complete the bug registration process
[00:00] --- Fri May 10 2013