[Ffmpeg-devel-irc] ffmpeg-devel.log.20120405

burek burek021 at gmail.com
Fri Apr 6 02:05:04 CEST 2012


[01:48] <ramiro> michaelni: the thing is that the granularity is huge. it's 1s. it can be made smaller, but then the table grows proportionally. the index table doesn't allow setting the real timestamp inside that 1s.
[01:48] <ramiro> for example in the regression tests the real timestamp is 960, but it gets indexed as 1900
[01:49] <ramiro> and the table is designed to work with backward seeks. so forward seeks shouldn't use it.
[01:50] <ramiro> imo the index table should be used as an initial estimate. find the packet from the table, then look for the next and the previous keyframe and determine their real pts. from that, look further if necessary or just return the frame if it's found
[01:52] <ramiro> or just drop the table altogether in asfdec and use normal binary search
[02:16] <burek> is there a way to select certain/specific alsa audio output device with ffplay
[03:57] <burek> is it too hard to make a feature, just like in vlc, that will allow ffmpeg to retrieve youtube videos directly?
[04:10] <Compn> burek : no, either using a wrapper to youtube-dl.py or just copying what vlc does
[04:10] <Compn> should only take probably 20 mins or less by any smart devel
[04:11] <burek> http://www.vbforums.com/showpost.php?s=d7c80ec67591eccfca871a0e1c0a8d1e&p=3033978&postcount=2
[04:11] <burek> :)
[04:13] <burek> yes, I noticed that a lot of people have this solved
[04:13] <burek> like JW Player (web video flash player)
[04:15] <burek> I see that youtube-dl has got support not only for yt but for a lot of content delivery websites too
[04:15] <burek> cool thing :)
[05:35] <CIA-17> ffmpeg: 03multiple authors 07master * r13cc3645b0 10ffmpeg/ (6 files in 4 dirs): 
[05:35] <CIA-17> ffmpeg: libavfilter: add colormatrix filter
[05:35] <CIA-17> ffmpeg: Ported by: Baptiste Coudurier
[05:35] <CIA-17> ffmpeg: cleanup+fate by ubitux
[05:35] <CIA-17> ffmpeg: For detailed authorship of the original code please see avisynth
[06:22] <CIA-17> ffmpeg: 03Ramiro Polla 07master * r97d36a1898 10ffmpeg/libavformat/asfenc.c: 
[06:22] <CIA-17> ffmpeg: asfenc: realloc index_ptr fewer times
[06:22] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:22] <CIA-17> ffmpeg: 03Ramiro Polla 07master * r1ceff0859d 10ffmpeg/libavformat/asfenc.c: 
[06:22] <CIA-17> ffmpeg: asfenc: rename some variables
[06:22] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:22] <CIA-17> ffmpeg: 03Ramiro Polla 07master * rae315a6a38 10ffmpeg/ (libavformat/asfenc.c tests/ref/lavf/asf): 
[06:22] <CIA-17> ffmpeg: asfenc: start at object 1 instead of 0
[06:22] <CIA-17> ffmpeg: This is how it is done in the official muxer.
[06:22] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:22] <CIA-17> ffmpeg: 03Ramiro Polla 07master * rbd603494f9 10ffmpeg/ (3 files in 3 dirs): 
[06:22] <CIA-17> ffmpeg: asfenc: properly write index information
[06:22] <CIA-17> ffmpeg: The index must take into account the pre-roll time and must seek backwards,
[06:23] <CIA-17> ffmpeg: not forwards.
[06:23] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:23] <CIA-17> ffmpeg: 03Ramiro Polla 07master * rf2fad251b8 10ffmpeg/libavformat/asfenc.c: 
[06:23] <CIA-17> ffmpeg: asfenc: reduce code duplication with new variable
[07:02] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * rca2b450cee 10ffmpeg/libavfilter/vf_idet.c: 
[07:02] <CIA-17> ffmpeg: vf_idet: use enum to represent the type.
[07:02] <CIA-17> ffmpeg: This will simplify future code.
[07:02] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[07:02] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r0477254dd2 10ffmpeg/libavfilter/vf_idet.c: 
[07:02] <CIA-17> ffmpeg: vf_idet: Print both the single frame based detection and the multi frame one.
[07:02] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[07:02] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r1a5c08eef6 10ffmpeg/libavfilter/vf_idet.c: 
[07:02] <CIA-17> ffmpeg: vf_idet: Add last_type variable to make code more readable.
[07:02] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[07:33] <CIA-17> ffmpeg: 03Clément BSsch 07master * rbfd30fa8a3 10ffmpeg/tests/ref/lavfi/ (colormatrix1 colormatrix2): lavfi/colormatrix: add forgotten tests.
[09:15] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r4d4df310f2 10ffmpeg/libavfilter/vf_idet.c: 
[09:15] <CIA-17> ffmpeg: vf_idet: print accumulated statistics for single and multi frame detection
[09:15] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[09:15] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * rc59e73d2c8 10ffmpeg/libavfilter/vf_idet.c: 
[09:15] <CIA-17> ffmpeg: vf_idet: change individual per frame statistics to debug log level.
[09:15] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[09:15] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * rfa1f92a4f1 10ffmpeg/libavfilter/vf_idet.c: 
[09:15] <CIA-17> ffmpeg: vf_idet: clean detected frame types based on a 4 frame history.
[09:15] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[09:50] <ubitux> we don't have a placeholder for "ffmpeg regressions", right?
[09:56] <michaelni> placeholder ?
[09:56] <ubitux> i mean just like lavfi-regressions.sh
[09:57] <ubitux> we don't have a "ffmpeg-regressions.sh"
[09:57] <ubitux> (for cmd line parsing or generic options etc)
[09:57] <michaelni> no, we dont have such a thing yet
[09:57] <ubitux> ok
[10:26] <spaam> ubitux: cant you use fate for it ? 
[10:28] <ubitux> you mean the fate/*.mak?
[10:28] <ubitux> *-regression.sh scripts are part of fate
[12:38] <CIA-17> ffmpeg: 03Thierry Foucu 07master * rc99bdd51b2 10ffmpeg/libavformat/utils.c: (log message trimmed)
[12:38] <CIA-17> ffmpeg: lavf: Fix the last PTS to be generated using -fflags +genpts
[12:38] <CIA-17> ffmpeg: To reproduce the problem, using ffprobe:
[12:38] <CIA-17> ffmpeg: ./ffprobe -show_packets -print_format compact -fflags +genpts -i
[12:38] <CIA-17> ffmpeg: fate_samples/mxf/C0023S01.mxf
[12:38] <CIA-17> ffmpeg: You will notice that the last video frame does not have it's PTS being
[12:38] <CIA-17> ffmpeg: set, even with using genpts.
[15:56] <burek> is it possible to have ffprobe -show_streams with ffmpeg?
[15:57] <Compn> its the same commands basically
[15:57] <Compn> so should be
[15:57] <Compn> ffmpeg -i file
[15:58] <burek> i tried ffmpeg -show_streams but no
[15:58] <burek> so I thought maybe it's ffprobe only
[17:03] <burek> hm, im puzzled with this paste: http://pastebin.com/WJ7qaxA2
[17:03] <burek> is it possible that format and all the streams can contain metadatas each
[17:04] <burek> metadata for format + meta for stream #0, meta for stream #1, etc.. ?
[17:05] <nevcairiel> yes, there is stream based metadata
[17:09] <burek> thanks :)
[17:24] <spaam> ubitux: yes. if the output differ then something is wrong..
[17:25] <ubitux> heh?
[17:33] <spaam> ubitux: yes. its like that 
[17:33] <CIA-17> ffmpeg: 03Clément BSsch 07master * r614716f583 10ffmpeg/libavformat/mov.c: (log message trimmed)
[17:33] <CIA-17> ffmpeg: mov: do not print string len in handler_name.
[17:33] <CIA-17> ffmpeg: The handler name is stored as a pascal string in the QT specs (first
[17:33] <CIA-17> ffmpeg: byte is the length of the string), thus leading to an invalid metadata
[17:33] <CIA-17> ffmpeg: string export.
[17:33] <CIA-17> ffmpeg: Also add a second length check based on the first character to avoid
[17:33] <CIA-17> ffmpeg: overwriting an already specified handler_name (it happens with Youtube
[17:33] <ubitux> spaam: i don't understand what you are talking about :p
[17:34] <spaam> ubitux: neither do i
[17:35] <ubitux> :)
[20:51] <bcoudu> how do we debug format timestamps in the latest version ?
[20:52] <bcoudu> -fdebug ts doesn't work anymore
[21:42] <michaelni> bcoudurier, -fdebug ts seems to need -v 99 
[22:02] <bcoudurier> right
[22:13] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r3e4b5e68c1 10ffmpeg/cmdutils_common_opts.h: 
[22:13] <CIA-17> ffmpeg: cmdutils: fix -fdebug without -v xy
[22:13] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:44] <CIA-17> ffmpeg: 03Anton Mitrofanov 07master * r14af74e9a0 10ffmpeg/libavcodec/h264.c: 
[22:44] <CIA-17> ffmpeg: h264: fix 4:2:2 PCM-macroblocks decoding
[22:44] <CIA-17> ffmpeg: Fixes bug 239.
[22:44] <CIA-17> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[22:45] <CIA-17> ffmpeg: 03Diego Biurrun 07master * r1f068e4450 10ffmpeg/libavfilter/vf_unsharp.c: 
[22:45] <CIA-17> ffmpeg: vf_unsharp: Mark readonly variable as const.
[22:45] <CIA-17> ffmpeg: This fixes the following warning:
[22:45] <CIA-17> ffmpeg: libavfilter/vf_unsharp.c:106: warning: initialization discards qualifiers from pointer target type
[22:45] <CIA-17> ffmpeg: 03Christophe GISQUET 07master * r6b81da2fd0 10ffmpeg/libavcodec/x86/ (dsputil_yasm.asm sbrdsp.asm): 
[22:45] <CIA-17> ffmpeg: dsputil x86: use SSE float instruction instead of SSE2 integer equivalent
[22:45] <CIA-17> ffmpeg: All the more required since the users are pure SSE functions.
[22:45] <CIA-17> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[22:45] <CIA-17> ffmpeg: 03Christophe GISQUET 07master * rf9888520cc 10ffmpeg/libavcodec/x86/vp8dsp.asm: 
[22:45] <CIA-17> ffmpeg: vp8dsp x86: perform rounding shift with a single instruction
[22:45] <CIA-17> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[22:45] <CIA-17> ffmpeg: 03Anton Khirnov 07master * rcfec77aaf8 10ffmpeg/configure: configure: add dl to frei0r extralibs.
[22:45] <CIA-17> ffmpeg: 03Christophe GISQUET 07master * rcd88105f6f 10ffmpeg/libavcodec/x86/dsputil_mmx.c: 
[22:45] <CIA-17> ffmpeg: dsputil x86: remove deprecated parameter from scalarproduct_int16 prototype
[22:45] <CIA-17> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[22:45] <CIA-17> ffmpeg: 03Anton Khirnov 07master * ra03f8ea0a8 10ffmpeg/libavfilter/vsrc_buffer.c: vsrc_buffer: when no frame is available, return an error instead of segfaulting.
[22:45] <CIA-17> ffmpeg: 03Diego Biurrun 07master * r0becb07842 10ffmpeg/libavcodec/ (h264.c h264.h h264_cabac.c h264_cavlc.c): h264: Factorize declaration of mb_sizes array.
[22:45] <CIA-17> ffmpeg: 03Diego Biurrun 07master * re484265c97 10ffmpeg/libavcodec/aacenc.c: 
[22:45] <CIA-17> ffmpeg: aacenc: Mark deinterleave_input_samples argument as const.
[22:45] <CIA-17> ffmpeg: This fixes the warning:
[22:45] <CIA-17> ffmpeg: libavcodec/aacenc.c:524: warning: passing argument 2 of deinterleave_input_samples discards qualifiers from pointer target type
[22:45] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r2c5a2958e9 10ffmpeg/: (log message trimmed)
[22:45] <CIA-17> ffmpeg: Merge remote-tracking branch 'qatar/master'
[22:45] <CIA-17> ffmpeg: * qatar/master:
[22:45] <CIA-17> ffmpeg:  h264: Factorize declaration of mb_sizes array.
[22:45] <CIA-17> ffmpeg:  vsrc_buffer: when no frame is available, return an error instead of segfaulting.
[22:45] <CIA-17> ffmpeg:  configure: add dl to frei0r extralibs.
[22:45] <CIA-17> ffmpeg:  dsputil x86: use SSE float instruction instead of SSE2 integer equivalent
[22:45] <CIA-17> ffmpeg: 03Ronald S. Bultje 07master * r7756859a31 10ffmpeg/tests/ (15 files in 3 dirs): fate: add BMP tests.
[22:45] <CIA-17> ffmpeg: 03Ronald S. Bultje 07master * rbc0bdda77b 10ffmpeg/libswscale/ (input.c output.c): 
[22:45] <CIA-17> ffmpeg: swscale: handle complete dimensions for monoblack/white.
[22:45] <CIA-17> ffmpeg: Fixes bug 269.
[23:38] <ubitux> bcoudurier: the -amerge code won't require much change when we will need to move to a better solution; it's likely we will just have to change the "amovie=foo.avi:si=3" into something like "[0:3]" as source (see -filter_complex patch proposition by Anton)
[23:40] <ubitux> and anyway, the "hack" is pretty isolated
[23:40] <ubitux> the visibility of the feature will also be restricted (i removed the changelog entry, and added a warn in the doc)
[23:43] <ubitux> are you proposing yourself to help with the lavfi auto reconfiguration, or audio stretching?
[23:54] <beastd> "-filter_complex" sounds pretty bad from a naming perspective. didn't read the proposal though.
[23:56] <ubitux> it's still a RFC
[23:56] <ubitux> from what i can tell :)
[00:00] --- Fri Apr  6 2012


More information about the Ffmpeg-devel-irc mailing list