[Ffmpeg-devel-irc] ffmpeg-devel.log.20120902

[burek]

[01:24] <ubitux> philipl: duration = end looks weird :p
[01:24] <ubitux> hey btw, wtf http://b.pkh.me/wtf.idx ??
[01:25] <philipl> what's wrong with that output?
[01:25] <philipl> duration = end is weird, but the way the AVSubtitle fields are used is just insane. end_display_time is the duration, start_display_time must be zero, and pts is pts.
[01:25] <philipl> ...
[01:26] <ubitux> < philipl> what's wrong with that output? // well.. two streams at index=0 with the same id
[01:26] <ubitux> matching timestamps
[01:26] <ubitux> but non-matching offsets
[01:26] <philipl> With sequential file positions. neat.
[01:26] <philipl> You sure the original source is correct?
[01:27] <philipl> Not that I know anything about vobsub except that it's easy to make messed up dvds. :-)
[01:28] <ubitux> i think this one was dumped from a dvd i own at some point
[01:28] <ubitux> maybe i'll re-dump it again
[01:28] <ubitux> it's like the first chapter has be handled differently
[01:28] <ubitux> and messed up everything
[01:32] <philipl> Yeah.
[01:32] <philipl> Maybe some multi-angle craziness.
[01:32] <ubitux> :(
[03:51] <ubitux> michaelni: i have kind of a weird config myself concerning git
[03:51] <ubitux> my origin is my github
[03:51] <ubitux> and "upstream" is ffmpeg.org
[03:52] <ubitux> so why not grepping only the url?
[03:53] <ubitux> i guess it's simpler to hardcode origin, but i'm not sure it will handle all the cases
[03:53] <ubitux> well anyway, got to sleep
[04:18] <michaelni> ubitux, better solutions are welcome!
[05:07] <CIA-47> ffmpeg: 03Michael Niedermayer 07master * r4654bf44a7 10ffmpeg/libavformat/flvdec.c: 
[05:07] <CIA-47> ffmpeg: flvdec: fix creation of lots of phantom data streams
[05:07] <CIA-47> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[10:49] <CIA-52> ffmpeg: 03Stefano Sabatini 07master * r06fc74bda0 10ffmpeg/libavcodec/codec_desc.c: lavc/codec_desc: fix name of wavesynth codec
[10:49] <CIA-52> ffmpeg: 03Stefano Sabatini 07master * r34bf06a35c 10ffmpeg/libavfilter/avfilter.c: 
[10:49] <CIA-52> ffmpeg: lavfi: when configuring a link, set channel layout from previous link if unspecified
[10:49] <CIA-52> ffmpeg: In avfilter_config_links(), if the current channel layout is not
[10:49] <CIA-52> ffmpeg: specified, set the same value of the previous link in the filterchain.
[10:49] <CIA-52> ffmpeg: Consistent with the logic used to set the other link parameters.
[10:49] <CIA-52> ffmpeg: 03Stefano Sabatini 07master * rf7eec8956d 10ffmpeg/libavformat/segment.c: 
[10:49] <CIA-52> ffmpeg: lavf/segment: set EXT-X-VERSION to 3 rather than to 4
[10:49] <CIA-52> ffmpeg: There is nothing we use of the latest version, downgrade version number
[10:49] <CIA-52> ffmpeg: to increase decoding support.
[10:49] <CIA-52> ffmpeg: Based on this comment by vel2000:
[10:49] <CIA-52> ffmpeg: http://ffmpeg.org/trac/ffmpeg/ticket/1642#comment:17
[10:49] <CIA-52> ffmpeg: 03Stefano Sabatini 07master * rc8ef309379 10ffmpeg/ (.gitignore doc/examples/Makefile doc/examples/demuxing.c): examples: add demuxing.c example
[10:55] <saste> ubitux: ping on "lavf/segment: add CVS escaping for CSV list file filename field"
[10:56] <saste> ops i just noticed the patch subject contains a typo
[11:06] <ubitux> saste: will you send another patch or i can review it right now?
[11:06] <saste> i already fixed it locally
[11:33] <ubitux> mmh where are the frame fields free'd?
[11:33] <ubitux> (AVFrame)
[11:33] <ubitux> since the user just has to av_free the ptr
[12:16] <saste> ubitux: the frame is only a container where the codec puts the information
[12:16] <saste> the information/memory is usually allocated in the codec context, and the user needs to copy it in the application context if she wants to preserve them
[12:17] <saste> so there is no need to free data pointed by frame struct, since it is already done by the codec itself
[12:20] <ubitux> ah, i see
[12:20] <ubitux> thx ;)
[12:20] <ubitux> i wonder if it would make sense to move the rects to the codec context
[12:20] <ubitux> the rects from avsubtitles
[12:24] <ubitux> [~/src/ffmpeg]- time dash ./configure > /dev/null
[12:24] <ubitux> dash ./configure > /dev/null  3.14s user 1.82s system 61% cpu 8.083 total
[12:24] <ubitux> [~/src/ffmpeg]- time bash ./configure > /dev/null
[12:24] <ubitux> bash ./configure > /dev/null  13.08s user 2.43s system 72% cpu 21.322 total
[12:24] <ubitux> meh.
[13:05] <saste> ubitux: I like your prompt
[13:06] <ubitux> it's for utf-8 testing purpose obviously
[13:23] <saste> I didn't know unicode had that symbol
[13:25] <ubitux> with compose key can access it using C-C-C-P (case sensitive, no dashes)
[13:33] <ubitux> (saste: and even better it gets red when $? ` 0)
[13:34] <saste> useful
[13:34] <ubitux> :)
[13:34] <saste> it's much harder to set unicode chars with no compose key
[13:34] <saste> italian keyboards are lame
[13:35] <ubitux> i'm using a qwerty + compose key
[13:35] <ubitux> so when writing french i have to use <compose>'e to do a simple é but well
[13:36] <ubitux> it allows to write stuff like « & e - », which is way more important!
[14:25] <saste> ffplay ffmpeg-logo.png
[14:25] <saste> looks bad
[18:55] <ubitux> saste: will we be able at some point to have the lavfi commands stuff frame accurate?
[18:56] <saste> ubitux: what do you mean?
[18:56] <ubitux> instead of 0:05:00-0:15:00, {1234}-{1300}
[18:56] <saste> with the possibility to specify the *number* of frame to which to apply a command?
[18:56] <ubitux> yes
[18:56] <saste> that can't work in case of seeking
[18:57] <saste> why do you need that?
[18:57] <ubitux> then be it limited to a linear processing
[18:57] <ubitux> to have a precise filtering
[18:57] <ubitux> if you are editing a video, you most like want it frame accurate
[19:23] <CIA-52> ffmpeg: 03Peter Ross 07master * r083e65cfd2 10ffmpeg/tests/ (fate/video.mak ref/fate/ansi256): 
[19:23] <CIA-52> ffmpeg: fate: add ansi256 test case
[19:23] <CIA-52> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:06] <CIA-52> ffmpeg: 03Michael Niedermayer 07master * r50b0edea9f 10ffmpeg/: 
[21:06] <CIA-52> ffmpeg: Merge remote-tracking branch 'qatar/master'
[21:06] <CIA-52> ffmpeg: * qatar/master:
[21:06] <CIA-52> ffmpeg:  intrax8: move functions from dsputil to own context
[21:06] <CIA-52> ffmpeg: Merged-by: Michael Niedermayer <michaelni at gmx.at>
[21:06] <CIA-52> ffmpeg: 03Mans Rullgard 07master * r15616eb042 10ffmpeg/libavcodec/ (6 files): 
[21:06] <CIA-52> ffmpeg: intrax8: move functions from dsputil to own context
[21:06] <CIA-52> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[21:07] <maister> Looking at STORE_BGR24_MMX in swscale. Can anyone see if these two lines are redundant? https://gist.github.com/3603273
[21:08] <maister> mask24hh = 0xffffffff00000000ULL;
[21:08] <maister> err
[21:08] <maister> mark24hh = 0xffff000000000000ULL;
[21:08] <maister> mask24hhh = 0xffffffff00000000ULL;
[21:20] <michaelni> maister, looks redundant to me 
[21:21] <maister> should I test it, or is it too obvious?
[21:25] <michaelni> depends, if you have some testcase that just takes a minute or so for you to run it cant hurt
[21:25] <maister> can try
[21:26] <michaelni> maister, the 3rd pand looks redundant too
[21:26] <maister> :P
[21:27] <maister> trying to write something similar for sse2 atm, but it's ... quite dense. Working through it with pen and paper, interpreting each opcode :v
[21:29] <maister> hm, my connection is really slow, trying from a fresh clone to be sure.
[21:54] <maister> https://gist.github.com/3603955 ok, works here.
[22:18] <CIA-52> ffmpeg: 03Hans-Kristian Arntzen 07master * rf099fbf5f3 10ffmpeg/libswscale/x86/rgb2rgb_template.c: 
[22:18] <CIA-52> ffmpeg: Remove redundant masks in STORE_BGR24_MMX.
[22:18] <CIA-52> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:48] <nyuhu> in a filter I have implemented the start_frame and draw_slice callbacks; in draw_slice() inlink->dst->outputs[0]->out_buf is null, did I forget to do something in start_frame ? (if I let the default start_frame callback I have no problem)
[00:00] --- Mon Sep  3 2012