[Ffmpeg-devel-irc] ffmpeg-devel.log.20121231

[burek]

[00:09] <cone-681> ffmpeg.git 03Clément BSsch 07master:faa94061dd72: Add SubViewer v1 subtitles demuxer and decoder.
[00:17] <michaelni> <ubitux> btw, sub/MicroDVD_capability_tester.srt can be removed <--- was it never used ?
[00:17] <ubitux> it was badly named at first, i don't remember if it was part of the tests, maybe at some point, but that's a long time ago
[00:18] <michaelni> if it was used then it cannot be removed
[00:21] <ubitux> ah? :(
[00:21] <michaelni> consider bisect with fate ...
[00:22] <ubitux> it's really supported?
[00:23] <ubitux> i guess you would fate run a specific test for bisect
[00:23] <ubitux> of course it might be a pain for tracking microdvd regressions
[00:44] <cone-681> ffmpeg.git 03Clément BSsch 07master:1f265f52050e: microdvd: sanitize AVPackets.
[00:44] <ubitux> here we go
[00:44] <ubitux> DONE.
[00:46] <wm4> ok how do I use this shit
[00:46] <wm4> is converting to ass still done by "decoding"?
[00:47] <ubitux> yes
[00:47] <ubitux> again, you can look at vf subtitles
[00:48] <wm4> right
[00:48] <ubitux> wm4: also note that charset/encoding/codepage/etc is not done
[00:48] <ubitux> (fribidi, ..)
[00:48] <wm4> why would fribidi be needed here?
[00:49] <ubitux> no idea, i didn't look into this stuff yet
[00:49] <wm4> from what I understand, just passing the text to libass should be enough
[00:49] <wm4> though there may be weird issues with the default base direction...
[00:49] <ubitux> mmh yeah right, if libass is able to deal with it
[00:50] <ubitux> note that if the input is not utf-8, it will likely be broken
[00:50] <ubitux> a lot of subtitles are not in utf-8, and since we don't do any convert to utf-8, libass might not like the input
[00:51] <wm4> subreader.c in mplayer does conversion
[00:51] <ubitux> yes i know
[00:52] <wm4> the fribidi stuff there is only needed for the internal sub renderer, I think
[00:52] <wm4> while with -ass it's not needed and/or causes corruption (I don't quite remember)
[00:52] <ubitux> see http://ffmpeg.org/pipermail/ffmpeg-devel/2012-December/136506.html for the next plan
[00:53] <ubitux> ok, i'll focus on charset convert then
[00:53] <wm4> libass also has some charset stuff AFAIK, but you'll have to read the libass source, heh
[00:54] <ubitux> ah? that's interesting if so
[01:02] <ubitux> michaelni: so, do you have sub/MPSub_capability_tester.sub?
[01:12] <Compnn> is that a sample ?
[01:12] <Compnn> what links to it ?
[01:13] <ubitux> i sent him a link a while ago in private
[01:13] <ubitux> i think.
[01:13] <Compnn> its not on ffmpeg box
[01:13] <Compnn> at least not that i see
[01:14] <michaelni> ubitux, dunno, locate doesnt find it on my disk
[01:14] <ubitux> yes that's why i'm asking
[01:14] <ubitux> ok
[01:14] <Compnn> ubitux : i mean, who linked to it in the first place ?
[01:14] <ubitux> http://b.pkh.me/MPSub_capability_tester.sub
[01:14] <ubitux> if you guys could put this in fate-samples/sub/... :)
[01:15] <Compnn> that looks familiar
[01:15] <ubitux> Compnn: yes that's the default sample from mplayer
[01:15] <Compnn> http://www1.mplayerhq.hu/DOCS/tech/mpsub.sub
[01:15] <Compnn> bah
[01:15] <ubitux> i may have done slight changes, i don't remember
[01:16] <ubitux> Compnn: anyway, it's used for one test, but since it was forgotten i disabled the test
[01:16] <michaelni> ubitux, uploaded
[01:16] <ubitux> michaelni: thanks a lot :)
[01:16] <ubitux> i'll re-enable the test in about two days
[01:16] <michaelni> np
[01:16] <michaelni> ok
[01:41] <cone-681> ffmpeg.git 03Clément BSsch 07master:21b6991cd530: lavfi/Makefile: make FFLIBS list consistent.
[01:41] <cone-681> ffmpeg.git 03Clément BSsch 07master:8d06e83d16eb: lavfi/Makefile: add missing dependencies to FFLIBS.
[01:44] <saste> - av_basename and av_dirname
[01:45] <saste> since when we add new functions in the changelog?
[01:45] <ubitux> &
[01:45] <ubitux> i didn't even realize
[01:45] <ubitux> feel free to move it to APIChanges
[02:33] <cone-681> ffmpeg.git 03Clément BSsch 07master:fe5a2fc46c55: Changelog: remove av_{base,dir}_name addition.
[02:48] <ubitux> 'Enca' (Extremely Naive Charset Analyser)
[02:48] <ubitux> 'wasn't aware what that meant :)
[02:49] <ubitux> btw, too bad we don't have subtitles in lavfi
[02:49] <ubitux> we could use it to rescale the subtitles pts with a setpts filter
[03:09] <ubitux> what's the point of check_lib instead of check_lib2?
[03:09] <ubitux> it seems check_lib2() is testing the function using the include
[03:10] <ubitux> while check_lib is doing a first test with the include
[03:10] <ubitux> and a second one with the function
[08:14] <cone-668> ffmpeg.git 03Paul B Mahol 07master:9a742826443a: Sony Wave64 muxer
[09:37] <ubitux> the charset conversion was simpler than expected
[09:37] <ubitux> at least a naive one
[10:03] <durandal_1707> our news page should be in separate html
[10:04] <durandal_1707> currently i cant link to news from g+ because it will show description of project and not actual news
[10:10] <cone-668> ffmpeg.git 03Clément BSsch 07master:e3c25860dc6e: lavc/utils: reindent in avcodec_decode_subtitle2().
[10:11] <cone-668> ffmpeg.git 03Clément BSsch 07master:72ff8ee158c7: lavc: correctly set AVSubtitle format for text based subtitles.
[10:23] <ubitux> wm4: do you want the charset convertion to support enca as well when available?
[10:57] <ubitux> wm4: enca is going to be a bit complicated
[10:57] <ubitux> i suggest that you keep that at mplayer level
[10:58] <ubitux> i may add it to the subtitles support
[10:58] <ubitux> saste: do you mind if i rename vf_ass.c to vf_subtitles?
[10:59] <saste> ubitux, no i don't
[11:00] <ubitux> ok :)
[11:00] <durandal_1707> why are there two filters with same description?
[11:01] <saste> yes it should be fixed
[11:01] <durandal_1707> in that file
[11:02] <saste> adding "ASS subtitles" to ass should be enough
[11:41] <ubitux> arg utf16
[11:41] <ubitux> fuck that shit :(
[11:42] <nevcairiel> lol
[11:43] <nevcairiel> iconv doesnt do utf16?
[11:43] <durandal_1707> michaelni: apad pads silence at end only?
[11:44] <durandal_1707> it could be nice to pad start and/or in middle
[11:46] <saste> durandal_1707, you can use aresample for that
[11:49] <durandal_1707> how?
[11:55] <saste> durandal_1707, aresample=min_comp=0.001:min_hard_comp=0.100000
[11:56] <saste> min_comp enable compensation, and min_hard_comp enable HARD compensation
[12:01] <ubitux> nevcairiel: i guess it does, the problem is that i'm using strlen()
[12:01] <ubitux> (and assuming ascii compatible)
[12:01] <ubitux> and afaik utf-16 files are most of the time half composed of 0x00
[12:02] <ubitux> wtf these guys were thinking..
[12:03] <saste> ubitux, UTF-16 is not that broken
[12:03] <nevcairiel> if you have normal ascii chars in utf-16, there are indeed a lot of 0x00 in there
[12:03] <ubitux> just half of it
[12:04] <saste> it is an extended ASCII, with a simple encoding (well an identical encoding indeed), and who cares about text files size anyway?
[12:04] <ubitux> who cares about using strlen()
[12:04] <ubitux> :p
[12:04] <nevcairiel> i'm just used to dealing with utf16 because i deal with windows, which deals with wchars =p
[12:04] <durandal_1707> there is code that reads 16 utf and converts it to 8utf
[12:05] <saste> ah strlen()
[12:05] <saste> i assume there are specific functions to deal with that
[12:05] <ubitux> saste: and any other string functions&
[12:05] <saste> like wstrlen() or such
[12:05] <ubitux> yes
[12:05] <nevcairiel> wcslen i think
[12:05] <ubitux> special cases
[12:05] <ubitux> nothx
[12:05] <ubitux> i'd better demux directly that stuff into utf-8
[12:05] <saste> ubitux, can you convert to UTF-8 and always assume UTF-8 internally?
[12:06] <ubitux> i'm assuming !utf16 now
[12:06] <ubitux> saste: the patch adds charset convert at codec level
[12:06] <ubitux> the problem is that it works only with ascii compliant charset
[12:06] <saste> wait, i have to actually read the patch...
[12:07] <ubitux> now what i'd like to do is make the text demuxers actually do a first pass from utf-16 to utf-8 if necessary
[12:07] <ubitux> so the following charset convert and any normal string processing will work
[12:08] <durandal_1707> hmm, why you need strlen at all...
[12:08] <ubitux> rects[i]->ass
[12:09] <ubitux> decoded text subtitles
[12:09] <ubitux> and actually, we are dealing with text subtitles
[12:09] <ubitux> we are assuming ascii compliant string all over the code
[13:07] <cone-668> ffmpeg.git 03Clément BSsch 07master:87350d6b44b2: lavc/textdec: fix subviewer1 preproc check.
[13:17] <cone-668> ffmpeg.git 03Diego Biurrun 07master:015da965a68b: libavcodec/utils: Add braces to shut up gcc warnings
[13:17] <cone-668> ffmpeg.git 03Anton Khirnov 07master:185c2ef8495c: cavsdec: cosmetics, reformat
[13:17] <cone-668> ffmpeg.git 03Anton Khirnov 07master:64bb3fa6b7f5: cavs: cosmetics, reformat top part
[13:17] <cone-668> ffmpeg.git 03Michael Niedermayer 07master:6ddda83c3f93: Merge commit '64bb3fa6b7f5b2db22b7577585d5861be24eebd8'
[13:38] <cone-668> ffmpeg.git 03Anton Khirnov 07master:7d8482640058: cavs: deMpegEncContextize
[13:38] <cone-668> ffmpeg.git 03Michael Niedermayer 07master:ca95d1b32df1: Merge commit '7d8482640058d66867602a6f58446c9342d9d146'
[13:40] <durandal_1707> who owns darwin clang2.9 x86_32 fate machine?
[13:41] <durandal_1707> michaelni: that machines that where red all the time have been removed?
[13:44] <michaelni> durandal_1707, ask the owner of these boxes i dont know, i didnt remove any
[13:44] <michaelni> but it can be a bug in fate too, ive seen machies disappear previously with noone doing anything
[13:44] <durandal_1707> yes, but interface is so missing that i cant find such info
[13:46] <durandal_1707> saste: are you against porting noise and instead want noise source?
[13:46] <saste> durandal_1707, i'm not against porting noise
[13:46] <saste> why should I?
[13:47] <durandal_1707> well you did some work on some noise filter iirc
[13:47] <saste> durandal_1707, it was eons ago
[13:47] <saste> and my implementation was not efficient at all
[13:48] <saste> now it would be easier to start from scratch (even if i don't like mp=noise syntax very much)
[13:49] <saste> i had a file source reading from /dev/urandom at some point
[13:49] <saste> in case you need "perfect" noise
[13:55] <nevcairiel> hrm, sometimes i find the whole decode API rather limiting ... I'm trying to fix the dvd sub decoder to properly deal with "animated" subtitles (dvd subs allow fade-in/fade-out by changing their alpha), but it requires output of several subtitles from one input packet, but the API doesn't particularly like that...
[13:56] <nevcairiel> i suppose i could somehow try to put the fade info into some metadata...
[13:57] <durandal_1707> doesnt vf_lut do rgba map thing which is done by spec function?
[13:59] <cone-668> ffmpeg.git 03Anton Khirnov 07master:fb0cb11f09c4: cavsdec: export picture type in the output frame
[13:59] <cone-668> ffmpeg.git 03Anton Khirnov 07master:5e6ee38bd3ce: FATE: add cavs test
[13:59] <cone-668> ffmpeg.git 03Michael Niedermayer 07master:98275283d526: Merge commit '5e6ee38bd3cef0dd05f1dd7977c71f3479eb6d01'
[13:59] <saste> wbs: is hls_wrap option compliant with M3U8 spec? 
[14:00] <saste> wbs: from the spec:  The sequence number of a segment is equal to the sequence number of the segment that preceded it plus one.  The EXT-X-MEDIA-SEQUENCE tag indicates the sequence number of the first segment that appears in a Playlist file.
[14:10] <wbs> it only wraps the sequence number in the filenames, not the EXT-X-MEDIA-SEQUENCE numbers - they never wrap
[14:23] <saste> wbs: yes but the second part states "the EXT-X-MEDIA-SEQUENCE tag indicates the sequence number of the first segment that appears in a Playlist file"
[14:24] <cone-668> ffmpeg.git 03Diego Biurrun 07master:117f79de2d0c: truemotion2: Sanitize tm2_read_header()
[14:24] <cone-668> ffmpeg.git 03Diego Biurrun 07master:bcb8d9eb8ff6: Drop unnecessary 'l' length modifier when printfing double values.
[14:24] <cone-668> ffmpeg.git 03Hendrik Leppkes 07master:506409776c49: configure: suppress -fPIC in msvc builds
[14:24] <cone-668> ffmpeg.git 03Michael Niedermayer 07master:08196c785a1d: Merge remote-tracking branch 'qatar/master'
[14:24] <saste> so if the segment index wraps, the EXT-X-MEDIA-SEQUENCE should wrap as well... or not?
[14:24] <wbs> saste: for the hls spec part, the sequence number is only what is in EXT-X-MEDIA-SEQUENCE
[14:24] <wbs> the hls spec doesn't know that the file names by conincidence happen to have a number in them
[14:24] <wbs> they could just as well be named a.ts, b.ts, etc
[14:25] <saste> wbs: ahhh... ok
[14:25] <saste> thank you
[14:47] <durandal_1707> michaelni: rgba<->abgr is not lossless here. is unscaled variant disabled recently?
[14:48] <michaelni> i dont remember anything having been disabled
[14:48] <durandal_1707> playing with format filter changes md5
[14:55] <durandal_1707> hmm it appears to work with images and ffmpeg
[14:55] <durandal_1707> but with histogram filter i get different results while in fact i should not
[15:02] <durandal_1707> nvm, guess this is caused by using frong format in filter
[15:17] <durandal_1707> wtf, i should have black frame both times, but i get black frame with 2 different hashes
[15:22] <durandal_1707> perhaps i left something not initialized
[19:09] <cone-668> ffmpeg.git 03Nicolas George 07master:5f12d5ad78f7: lswr: improve rematrixing error message.
[19:09] <cone-668> ffmpeg.git 03Nicolas George 07master:11ab2c25b6b2: lavfi/avfiltergraph: print query_formats errors.
[19:22] <cone-668> ffmpeg.git 03Nicolas George 07master:9e28e0035a15: lavc/flacenc: improve error return.
[21:01] <cone-668> ffmpeg.git 03Michael Niedermayer 07master:343900a14777: mpeg12: print error for invalid skip cases
[21:01] <cone-668> ffmpeg.git 03Michael Niedermayer 07master:49ec0d8ed246: mpeg12: print invalid cbp value
[21:01] <cone-668> ffmpeg.git 03Michael Niedermayer 07master:6e8df250fd98: mpeg12: fix frame_pred_frame_dct handling in field pictures
[00:00] --- Tue Jan  1 2013