[Ffmpeg-devel-irc] ffmpeg-devel.log.20120726

burek burek021 at gmail.com
Fri Jul 27 02:05:02 CEST 2012


[00:06] <Freakshow> does out->size; include packaging overhead?
[00:37] <michaelni> crtmpserver, i dont think you are on the right track with debuging this
[00:38] <michaelni> can you try to remove probesize = 0 before av_find_stream_info ?
[00:39] <michaelni> i mean in hls.c
[00:45] <CIA-41> ffmpeg: 03Nicolas George 07master * r6bf87785e8 10ffmpeg/ffmpeg.c: ffmpeg: warn about new streams after startup.
[00:45] <CIA-41> ffmpeg: 03Nicolas George 07master * rd0d9272580 10ffmpeg/doc/faq.texi: faq: explain about new streams, probesize and analyzeduration.
[00:47] <michaelni> crtmpserver, also if removing probesize=0 works, please try "v->ctx->ctx_flags &= ~AVFMTCTX_NOHEADER;"  instead of probesize =0
[01:03] <Mista_D> A quick question - why MP4 atom moov/trak(X)/tkhd/flag = 15? Shouldn't it be 1 or 7 (ISO/IEC 14496-12:2004).
[01:09] <MP4_maniac> Mista_D: dunno. flag 0x8 indicates that the track is used in the movie's poster but it is defined in Apple's MP4 or MOV. ISO doesn't define it
[01:10] <Mista_D> Any way to specify Apple/ISO format during encoding?
[01:10] <cptspiff> -f applemustdie
[01:11] <MP4_maniac> -f ipod
[01:11] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r9054f6b66b 10ffmpeg/libavformat/utils.c: 
[01:11] <CIA-41> ffmpeg: probe_codec: fix memory corruption
[01:11] <CIA-41> ffmpeg: Found-by: Tanami Ohad
[01:11] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:12] <Mista_D> MP4_maniac: Thank you.
[01:14] <Mista_D> MP4_maniac: still 0F, I think ipod uses default MP4 which is "Apple's" format, not ISO.
[01:27] <crtmpserver> michaelni, I'm all over it :)
[01:40] <Mista_D> MP4_maniac: x0F is a hard coded value in libavformat/movenc.c  line 1540.
[01:41] <MP4_maniac> yup
[02:08] <crtmpserver> michaelni, removing probesize = 0 fixed it!
[02:08] <crtmpserver> amazing!
[02:08] <crtmpserver> :)
[02:09] <michaelni> crtmpserver, does it still work with v->ctx->ctx_flags &= ~AVFMTCTX_NOHEADER; instead ?
[02:10] <crtmpserver> let me check
[02:10] <michaelni> this would reduce the amount of delay / latency
[02:10] <crtmpserver> so, should I leave probesize in there?
[02:10] <crtmpserver> and add v->ctx->ctx_flags &= ~AVFMTCTX_NOHEADER 
[02:10] <crtmpserver> ?
[02:12] <crtmpserver> having both is not working
[02:12] <crtmpserver> I guess that is not an issue, because HLS is already high-latency streaming
[02:12] <crtmpserver> talking about 10-15 seconds
[02:13] <crtmpserver> because of the nature of it
[02:14] <crtmpserver> removing probesize=0 and adding v->ctx->ctx_flags &= ~AVFMTCTX_NOHEADER also wroks
[02:14] <crtmpserver> I think probesize=0 was the faulty piece
[02:50] <michaelni> crtmpserver, thanks alot for the testing ill commit a fix soon
[02:52] <crtmpserver> I thank you for it. You saved my ass :)
[02:53] <crtmpserver> now I can do -vcodec copy in a transrating environment
[02:54] <crtmpserver> before, I was forced to re-encode the original
[02:54] <crtmpserver> wasting valuable CPU horsepower
[03:10] <CIA-41> ffmpeg: 03Piotr Bandurski 07master * rc5f23df2dd 10ffmpeg/libavformat/movenc.c: 
[03:10] <CIA-41> ffmpeg: movenc: fix remuxing of yuv2
[03:10] <CIA-41> ffmpeg: Attached patch fixes remuxing of yuv2 in mov (and does not break remuxing of yuvs)
[03:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r4e7f9ebdc2 10ffmpeg/libavformat/wav.c: 
[03:10] <CIA-41> ffmpeg: wav: switch to AVSTREAM_PARSE_FULL_RAW
[03:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r6f6182372c 10ffmpeg/libavformat/hls.c: 
[03:10] <CIA-41> ffmpeg: hls: replace probsize=0 by noheader flag removial.
[03:10] <CIA-41> ffmpeg: This fixes h264 timestamps in hls
[03:10] <CIA-41> ffmpeg: Fixes ticket1572
[03:10] <CIA-41> ffmpeg: Tested-by: crtmpserver
[03:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r12d1ee6a5e 10ffmpeg/libavformat/utils.c: 
[03:10] <CIA-41> ffmpeg: ff_add_index_entry: reject AV_NOPTS_VALUE
[03:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r5ea4ee7bcb 10ffmpeg/libavformat/flacdec.c: 
[03:10] <CIA-41> ffmpeg: flacdec: mark as AVSTREAM_PARSE_FULL_RAW
[03:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r2880c31660 10ffmpeg/libavformat/utils.c: 
[03:10] <CIA-41> ffmpeg: parse_packet: use pkt.pos always instead of the avparser offset.
[03:10] <CIA-41> ffmpeg: a later point.
[03:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rfdc371114f 10ffmpeg/libavformat/rawdec.c: 
[03:10] <CIA-41> ffmpeg: ff_raw_audio_read_header: mark as AVSTREAM_PARSE_FULL_RAW
[03:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rea8c12c90f 10ffmpeg/libavformat/mp3dec.c: 
[03:10] <CIA-41> ffmpeg: mp3dec: mark as AVSTREAM_PARSE_FULL_RAW
[03:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r2cd491a47c 10ffmpeg/ (libavformat/utils.c tests/ref/seek/mp2_mp2): (log message trimmed)
[03:10] <CIA-41> ffmpeg: lavf: move generic index generation code to a later point
[03:10] <CIA-41> ffmpeg: By moving it to a later point relative and unknown timestamps
[03:11] <CIA-41> ffmpeg: are more likely to have been corrected
[03:11] <CIA-41> ffmpeg: similar patch reviewed-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[03:11] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:11] <CIA-41> ffmpeg: Conflicts:
[03:34] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r9c83f3e659 10ffmpeg/libavformat/loasdec.c: 
[03:34] <CIA-41> ffmpeg: loasdec: switch to AVSTREAM_PARSE_FULL_RAW
[03:34] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:34] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * re083a62e9d 10ffmpeg/libavformat/aacdec.c: 
[03:34] <CIA-41> ffmpeg: aacdec: switch to AVSTREAM_PARSE_FULL_RAW
[03:34] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:34] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r97c0d5357d 10ffmpeg/libavformat/omadec.c: 
[03:34] <CIA-41> ffmpeg: omadec: switch to AVSTREAM_PARSE_FULL_RAW
[03:34] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[08:18] <rohankasat> hi , can anybody tell me . how ffmpeg calls a decoder ( from the different registered decoders) , so that it can process further ? any particular function call or wat ?
[10:14] <CIA-41> ffmpeg: 03William Yu 07master * rd5d5e3dabe 10ffmpeg/libavfilter/af_amerge.c: lavfi/af_amerge: avfilter_copy_buffer_ref_props will overwrite output pts
[13:16] <dougle> does anyone know of a filter to reduce 4:4:4 chroma down to baseline h264 compatible?
[13:17] <dougle> have not used filters before and am a little lost
[13:33] <michaelni> dougle, you probably search for the -pix_fmt option
[15:08] <dougle> michaelni i salute you sir! very helpful thanks
[16:36] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r7a4e30f3b6 10ffmpeg/libavcodec/h264_cabac.c: 
[16:36] <CIA-41> ffmpeg: h264_cabac: switch to av_assert
[16:36] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:37] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r506ffa793f 10ffmpeg/libavcodec/h264_cavlc.c: 
[16:37] <CIA-41> ffmpeg: h264_cabac: switch to av_assert
[16:37] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:08] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r55fa883b30 10ffmpeg/ffmpeg.c: 
[18:08] <CIA-41> ffmpeg: ffmpeg: fix units of next_?ts in debug printing routine.
[18:08] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:08] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r5938b4d398 10ffmpeg/libavcodec/mpc7.c: 
[18:08] <CIA-41> ffmpeg: mpc7: Initialize AVFrame properly
[18:08] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:08] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * re15e5328a7 10ffmpeg/libavcodec/mpc8.c: 
[18:08] <CIA-41> ffmpeg: mpc8: Initialize AVFrame properly
[18:08] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:08] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rb616405c16 10ffmpeg/ffmpeg.c: 
[18:08] <CIA-41> ffmpeg: ffmpeg: reset decoded_frame->pts after rescaling.
[18:08] <CIA-41> ffmpeg: This ensures its not used afterwards.
[18:08] <CIA-41> ffmpeg: This commit should not change anything as it should not have been used afterwards.
[18:08] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:08] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r9b1fdb1c5a 10ffmpeg/ffmpeg.c: 
[18:08] <CIA-41> ffmpeg: ffmpeg: fix assumed decoded_frame->pts timebase
[18:08] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:08] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r1a7af514de 10ffmpeg/ffmpeg.c: 
[18:08] <CIA-41> ffmpeg: ffmpeg: move decoded_frame->pts computation together
[18:08] <CIA-41> ffmpeg: This clarifies the code.
[18:08] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:08] <CIA-41> ffmpeg: ffmpeg: Fix rounding errors in decoded_frame->pts computation
[18:08] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:42] <maker> Writing here, as burek suggested.      Hello, I was interested in the Summer of Code in Space program, in particoular to the libavfilter project. Is saste, or any other backup mentor going to join before the deadline?
[18:50] <maker> bbl
[19:06] <durandal_1707> was?
[19:27] <ubitux> maker: saste might come back tonight here
[19:28] <ubitux> so you might want to wait for him until them, or send him a mail
[19:29] <maker> ubitux: I will probably try to join tonight, thanks. I saw saste sending mails at 2AM, I suppose he did not change his habits :D
[19:30] <ubitux> well maybe you're not in the same timezone :)
[19:31] <maker> ubitux: I am :p
[19:31] <iive> deadline? I thought that deadlines for sign up are somewhere between February and April. The deadlines that are coming is for finished projects.
[19:31] <iive> are...
[19:32] <maker> iive: for socis?
[19:32] <maker> http://sophia.estec.esa.int/socis2012/?q=timeline
[19:32] <iive> oh, sorry, i missed the "in space" bit.
[20:00] <michaelni> maker, if you want to apply for SOCIS make sure you submit whatever is needed (application i guess) before the deadline
[20:02] <michaelni> maker, but also wait and talk with saste, he is mentor of the libavfilter stuff IIRC and you will have to do some kind of qualification task
[20:05] <philipl> ubitux: Any chance you can look at the timed text encoder?
[20:05] <ubitux> yep i was purging my mails
[20:06] <ubitux> i didn't have time for me lately sorry
[20:06] <ubitux> i'll have a look tonight
[20:06] <ubitux> sorry for the delay :)
[20:07] <orbisvicis> is there a list of api changes, ie av_close_input_file -> avformat_close_input
[20:07] <orbisvicis> etc
[20:08] <ubitux> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=doc/APIchanges;hb=HEAD
[20:08] <orbisvicis> thanks
[20:20] <ubitux> it seems diego wants to drop gopher support because protocols don't belong to the project
[20:20] <ubitux> it makes me willing to support samba just like mplayer or vlc
[20:24] <philipl> ubitux: no probs. thanks.
[20:33] <durandal_1707> michaelni: you devastated Changelog in one of your merges
[20:33] Action: durandal_1707 took look at Changelog and got WTF moment
[20:34] <durandal_1707> git show 2cb4d516549526b5e17e941f6d2375a2c501ade6
[20:34] <nevcairiel> HEAD changelog looks ok
[20:35] <nevcairiel> oh some lines went missing there
[20:37] <ubitux> isn't it missing a few things?
[20:38] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r307a20cca2 10ffmpeg/Changelog: 
[20:38] <CIA-41> ffmpeg: Changelog: put entries lost in 2cb4d516549526b5e17e941f6d2375a2c501ade6 back
[20:38] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:38] <ubitux> thx :)
[20:55] <ubitux> philipl: is your mov_text branch in sync with your patchset?
[20:56] <philipl> Sorry, no. I forgot to push to that, and I can't push it right now.
[20:56] <philipl> I can do it tonight.
[20:56] <ubitux> ok no worry, i'll deal with the patches
[21:28] <orbisvicis> what is the version test for lavu 51.8.0 ? LIBAV..._VERSION_INT
[21:33] <orbisvicis> oh libavutil
[21:37] <orbisvicis> is this a typo:
[21:37] <orbisvicis> Add av_get_bits_per_sample_fmt() to libavcore/samplefmt.h
[21:37] <orbisvicis> shouldn't it be libavutil/samplefmt.h ?
[21:38] <orbisvicis> http://ffmpeg.org/doxygen/trunk/samplefmt_8c.html#ed9c391585b4d4d7c49a0ef88ce139ef
[21:38] <orbisvicis> ie version test against libavu vs libavc
[21:38] <nevcairiel> libavcore has been removed a long time ago
[21:47] <orbisvicis> ah ok
[21:48] <orbisvicis> how long ago? is it safe to completely replace av_get_bits_per_sample_format?
[21:50] <nevcairiel> all its functions are still present, just moved to libavutil
[21:55] <CIA-41> ffmpeg: 03Samuel Pitoiset 07master * r9b498148ca 10ffmpeg/libavformat/rtmpproto.c: 
[21:55] <CIA-41> ffmpeg: rtmp: Factorize the code by adding handle_server_bw
[21:55] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[21:55] <CIA-41> ffmpeg: 03Samuel Pitoiset 07master * r912ecc9a19 10ffmpeg/libavformat/rtmpproto.c: 
[21:55] <CIA-41> ffmpeg: rtmp: Factorize the code by adding handle_client_bw
[21:55] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[21:55] <CIA-41> ffmpeg: 03Samuel Pitoiset 07master * r7be2a7d8ff 10ffmpeg/libavformat/rtmpproto.c: 
[21:55] <CIA-41> ffmpeg: rtmp: Factorize the code by adding handle_chunk_size
[21:55] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[21:55] <CIA-41> ffmpeg: 03Samuel Pitoiset 07master * r6d1c9945dd 10ffmpeg/libavformat/rtmpproto.c: 
[21:55] <CIA-41> ffmpeg: rtmp: Factorize the code by adding handle_invoke
[21:55] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[21:55] <CIA-41> ffmpeg: 03Yang Wang 07master * r845e92fd6a 10ffmpeg/libavcodec/x86/dsputil_mmx.c: (log message trimmed)
[21:55] <CIA-41> ffmpeg: dsputil_mmx: fix incorrect assembly code
[21:55] <CIA-41> ffmpeg: In ff_put_pixels_clamped_mmx(), there are two assembly code blocks.
[21:55] <CIA-41> ffmpeg: In the first block (in the unrolled loop), the instructions
[21:55] <CIA-41> ffmpeg: "movq 8%3, %%mm1 \n\t", and so forth, have problems.
[21:55] <CIA-41> ffmpeg: From above instruction, it is clear what the programmer wants: a load from
[21:55] <CIA-41> ffmpeg: p + 8. But this assembly code doesnt guarantee that. It only works if the
[21:55] <CIA-41> ffmpeg: 03Ronald S. Bultje 07master * r79195ce565 10ffmpeg/libavcodec/ (8 files in 2 dirs): 
[21:55] <CIA-41> ffmpeg: x86/dsputil: put inline asm under HAVE_INLINE_ASM.
[21:55] <CIA-41> ffmpeg: This allows compiling with compilers that don't support gcc-style
[21:55] <CIA-41> ffmpeg: inline assembly.
[21:55] <CIA-41> ffmpeg: Signed-off-by: Derek Buitenhuis <derek.buitenhuis at gmail.com>
[21:55] <CIA-41> ffmpeg: 03Ronald S. Bultje 07master * r48098788c2 10ffmpeg/libavcodec/vp8.c: 
[21:55] <CIA-41> ffmpeg: vp8: Replace x*155/100 by x*101581>>16.
[21:55] <CIA-41> ffmpeg: Idea stolen from webp (by Pascal Massimino) - because it's Cool.
[21:55] <CIA-41> ffmpeg: Signed-off-by: Derek Buitenhuis <derek.buitenhuis at gmail.com>
[21:55] <CIA-41> ffmpeg: 03Ronald S. Bultje 07master * ra1878a88a1 10ffmpeg/libavcodec/x86/vp3dsp.asm: 
[21:55] <CIA-41> ffmpeg: vp3: don't use calls to inline asm in yasm code.
[21:55] <CIA-41> ffmpeg: Mixing yasm and inline asm is a bad idea, since if either yasm or inline
[21:55] <CIA-41> (44 lines omitted)
[21:55] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r7333798c85 10ffmpeg/: (log message trimmed)
[21:55] <CIA-41> ffmpeg: Merge remote-tracking branch 'qatar/master'
[21:55] <CIA-41> ffmpeg: * qatar/master:
[21:56] <CIA-41> ffmpeg:  libopenjpeg: support YUV and deep RGB pixel formats
[21:56] <CIA-41> ffmpeg:  Fix typo in v410 decoder.
[21:56] <CIA-41> ffmpeg:  vf_yadif: unset cur_buf on the input link.
[21:56] <CIA-41> ffmpeg:  vf_overlay: ensure the overlay frame does not get leaked.
[21:56] <CIA-41> ffmpeg: 03Samuel Pitoiset 07master * r0ffd5161c4 10ffmpeg/libavformat/rtmpproto.c: 
[21:56] <CIA-41> ffmpeg: rtmp: Factorize the code by adding handle_ping
[21:56] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[21:59] <orbisvicis> nevcairiel: so which version of avutil is the last to have av_get_bits_per_sample_format?
[22:06] <orbisvicis> i see -> digging in the git history
[22:08] <ubitux> hey saste, a socis student was looking for you
[22:08] <ubitux> 18:42:19 < maker> Writing here, as burek suggested.      Hello, I was interested in the Summer of Code in Space program, in particoular to the libavfilter project. Is saste, or any other backup mentor going to join before the deadline?
[22:08] <ubitux> philipl: i'll review the other patch when i'll be able to test it, i'd like to make some tests
[22:08] <saste> ubitux: yes, I see he applied
[22:08] <saste> i'll contact him later
[22:09] <saste> application deadline is tomorrow, for the selection there is time until July 31
[22:09] <ubitux> philipl: feel free to nudge me as soon as you update your branch
[22:09] <ubitux> saste: ok; he might get back on irc
[22:09] <saste> i'm going to post a reminder to the ffmpeg-* socis thread
[22:10] <ubitux> ok :)
[22:18] <philipl> ubitux: will do.
[22:19] <philipl> ubitux: btw: movtext.h is not used. You can just remove it to test.
[22:19] <ubitux> ah? great ok
[22:20] <philipl> That's from my initial styling work.
[22:20] <ubitux> then maybe i'll end the review tonight 
[22:20] <ubitux> ok
[22:21] <ubitux> philipl: btw, that's great to be able to mux subtitles in mp4
[22:21] <ubitux> thank you for working on this :)
[22:24] <ubitux> fate is broken :(
[22:25] <CIA-41> ffmpeg: 03Clément BSsch 07master * r2755abff3e 10ffmpeg/libavcodec/realtextdec.c: lavc/rt: small wording fix in a comment.
[22:25] <CIA-41> ffmpeg: 03Clément BSsch 07master * r316f8db2c2 10ffmpeg/ (13 files in 6 dirs): SubViewer demuxer and decoder.
[22:45] <Freakshow> burek... or anyone else for that matter. Is av_log designed to output data in 500mS samples?
[22:46] <philipl> ubitux: You're welcome
[22:46] <philipl> I broke fate?
[22:47] <Freakshow> nevermind, I see it in print_report
[22:47] <Freakshow> < 5000000
[22:47] <Freakshow> sorry... 500000
[22:49] <ubitux> philipl: no no, server is down
[22:50] <ubitux> philipl: btw, is there any player actually showing these subtitles?
[22:50] <ubitux> seems like mplayer & vlc don't
[22:50] <philipl> ubitux: mplayer and vlc both can handle them.
[22:50] <philipl> I test with mplayer and ipad.
[22:51] <ubitux> i did it wrong then ok
[22:51] <philipl> mplayer has very basic support but it works for unstyled subs.
[22:51] <ubitux> btw, two things:
[22:52] <ubitux> (which might not be related to your patch)
[22:52] <ubitux> 1) can't we avoid the -c:s mov_text? (auto select codec)
[22:52] <ubitux> 2) seems like subrip demuxer has a problem: ./ffmpeg -f lavfi -i testsrc=d=60 -i ~/fate-samples/sub/SubRip_capability_tester.srt -map 0:0 -map 1:0 -c:s mov_text -y out.mp4
[22:53] <philipl> 2) Yes. That's the duration vs convergence_duration problem
[22:53] <philipl> 1) I'm sure we can but there's no logic to do auto-matching for subs today. We'd have to add it.
[22:53] <ubitux> ok
[22:55] <ubitux> srt doesn't seem to have anything with convergence_duration
[22:57] <philipl> Sorry. It's actually that there's *no* duration
[22:57] <philipl> You need this: https://github.com/philipl/FFmpeg/commit/bd701c6dab3e57381d6c8bb1c9e651f755ea4f0b
[22:58] <ubitux> ah, makes sense.
[22:58] <ubitux> did you submit this?
[22:58] <philipl> No. But I guess I should :-)
[22:59] <ubitux> yup
[22:59] <philipl> On a related note: What's our official story on convergence_duration then? Matroska doesn't write duration. Which piece should be responsible for making things line up.
[22:59] <ubitux> no idea, i didn't follow the convergence_duration story
[23:00] <ubitux> it's all about having a 64-bit ts vs 32?
[23:00] <philipl> Yeah. After we established that there was value in having 64bits, the discussion died off.
[23:04] <ubitux> well does it make any sense?
[23:05] <ubitux> a subtitle event hardly last for more than 1 minute
[23:05] <ubitux> even with an insane timebase i don't see how that could really be a problem
[23:05] <ubitux> or maybe i'm missing the point?
[23:07] <philipl> I don't think it makes any sense myself.
[23:08] <philipl> The one thing I did think about was if the container is like mp4, then there's no stored duration, and the duration gets calculated to the next sample at demux time. In that case, an empty sample might have a very long duration.
[23:08] <philipl> But the decoder discards those samples, so their duration doesn't effect anything.
[23:08] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r236ecc3502 10ffmpeg/tests/Makefile: (log message trimmed)
[23:08] <CIA-41> ffmpeg: fate: change fate-suite rsync server.
[23:08] <CIA-41> ffmpeg: there are some technical problems with fate.ffmpeg.org
[23:08] <CIA-41> ffmpeg: thus split the subdomain between fate-suite and fate
[23:08] <CIA-41> ffmpeg: fate-suite is now (temporary) provided by our main server
[23:08] <CIA-41> ffmpeg: until fate-suite.ffmpeg.org is setup to point somewhere
[23:08] <CIA-41> ffmpeg: we use fate-suite.avcodec.org
[23:09] <ubitux> philipl: don't we use duration = -1 for these case anyway?
[23:09] <philipl> ubitux: No. I'm not sure what magic goes on inside the mov demuxer, but it manages to correctly recover subtitle durations.
[23:14] <philipl> It looks like there's some kind of sample index that is provided so it can calculate durations ahead of time.
[23:19] <ubitux> i have some timing issue here, but i need to investigate
[23:19] <ubitux> i'll take some time this week end
[00:00] --- Fri Jul 27 2012


More information about the Ffmpeg-devel-irc mailing list