[Ffmpeg-devel-irc] ffmpeg-devel.log.20120201

[burek]

[00:17] <durandal_1707> kierank: is there a way to pick between them?
[00:17] <michaelni> durandal_1707, ffv1s uses a context adaptive coder which is not reset between frames by default
[00:17] <kierank> if you care about latency then use slice threading, if you don't use frame threads :)
[00:17] <kierank> not that simple but meh
[00:59] <walker> Hi folks - I'm having an issue with h263 Mode B packets. From looking at ffmpeg-devel archives, it seems that ffmpeg only supports Mode A for h263 RTP?
[01:00] <walker> Is there any work on support for Mode B?
[01:01] <Compn> walker : good question. might want to ask whoever worked on h263 rtp
[01:02] <durandal_1707> walker: do you have sample?
[01:04] <walker> durandal_1707: I do.
[01:05] <durandal_1707> then open bug on ffmpeg trac explaining your problem and upload it there
[01:05] <walker> Okay.
[01:05] <walker> I just want to know at this point, though, if Mode B is supported at all?
[01:06] <Compn> i dont know :)
[01:06] <Compn> maybe wbs does
[01:06] <Compn> wbs : ping
[01:06] <durandal_1707> if it is not supported how it will get support if you never report it?
[01:07] <durandal_1707> only if someone else reports it or fix it
[01:07] <walker> From https://lists.libav.org/pipermail/ffmpeg-devel/2009-April/067074.html - "It seems reasonable to only implement Mode A"
[01:07] <Compn> walker : got a url that has problems ?
[01:07] <walker> Compn: Lemme stash it somewhere public.
[01:08] <Compn> walker : the guy who resent that patch is wbs , and michaelni is the one who reviewed it
[01:09] <Compn> did that patch ever get applied ?
[01:09] <walker> As far as I can see, it still supports only mode A.
[01:09] <walker> The stupid thing is, this h263 stream contains both mode A and mode B.
[01:09] <Compn> lol
[01:09] <walker> (Commercial videophone.)
[01:09] <Compn> i'm surprised there are still h263 streams
[01:09] <Compn> ah
[01:10] <Compn> h263 is last century maaaan
[01:10] <walker> Tell that to my client! I developed the software for h264 and h263+. Submitted it. They came back with, "It no woooork!"
[01:10] <Compn> haha
[01:12] <walker> http://static.carbonatedblog.com/h263_test.h263
[01:13] <walker> That's the raw h263 rtp stream containing mode A and B packets. Wireshark confirms both A and B in there.
[01:14] <walker> ffmpeg does its best: It takes the mode A packets and renders them to a rather bad video, making my ugly mug even uglier.
[01:14] <walker> But playing back that stream straight to the videophone gets crystal clear video.
[01:15] <Compn> alrighty, well stick around and wbs should come on sooner or later. or take it to the bug tracker
[01:19] <walker> Thanks!
[01:19] <walker> Poking around bugtracker.
[02:27] <xxthink> which video format mp4 does NOT support?
[03:06] <CIA-40> ffmpeg: 03Martin Storsjö 07master * r89f6e8a20c 10ffmpeg/libavformat/flvdec.c: (log message trimmed)
[03:06] <CIA-40> ffmpeg: flvdec: Interpret a toplevel 'object' type metadata item as normal metadata, too
[03:06] <CIA-40> ffmpeg: Previously, we've only passed the key string on to the recursive
[03:06] <CIA-40> ffmpeg: amf_parse_object for the mixedarray type, not for 'object'. By
[03:06] <CIA-40> ffmpeg: passing the key string on, the recursive amf_parse_object can
[03:06] <CIA-40> ffmpeg: store the amf objects as metadata.
[03:06] <CIA-40> ffmpeg: This kind of data was seen in data from XSplit Broadcaster, received
[03:06] <CIA-40> ffmpeg: 03Diego Biurrun 07master * r91bafb52ae 10ffmpeg/libavcodec/x86/ (Makefile rv40dsp.c): x86: Give RV40 init file a more suitable name.
[03:06] <CIA-40> ffmpeg: 03Martin Storsjö 07master * r75ab1e62d4 10ffmpeg/libavformat/mov.c: (log message trimmed)
[03:06] <CIA-40> ffmpeg: movdec: Ignore sample_degradation_priority bits when checking first_sample_flags
[03:06] <CIA-40> ffmpeg: This makes the first packet of a track fragment run to get
[03:06] <CIA-40> ffmpeg: the keyframe flag set properly if sample_degradation_priority
[03:06] <CIA-40> ffmpeg: is nonzero.
[03:06] <CIA-40> ffmpeg: This makes the keyframes flag be set properly for ismv files
[03:06] <CIA-40> ffmpeg: created by Microsoft.
[03:06] <CIA-40> ffmpeg: 03Christophe Gisquet 07master * re5c9de2ab7 10ffmpeg/libavcodec/x86/ (Makefile rv40dsp.asm rv40dsp_init.c): (log message trimmed)
[03:06] <CIA-40> ffmpeg: rv40: x86 SIMD for biweight
[03:06] <CIA-40> ffmpeg: Provide MMX, SSE2 and SSSE3 versions, with a fast-path when the weights are
[03:06] <CIA-40> ffmpeg: multiples of 512 (which is often the case when the values round up nicely).
[03:06] <CIA-40> ffmpeg: *_TIMER report for the 16x16 and 8x8 cases:
[03:06] <CIA-40> ffmpeg: C:
[03:06] <CIA-40> ffmpeg: 9015 decicycles in 16, 524257 runs, 31 skips
[03:06] <CIA-40> ffmpeg: 03Justin Ruggles 07master * reac31dd163 10ffmpeg/libavcodec/mpc7.c: 
[03:06] <CIA-40> ffmpeg: mpc7: align local temp buffer
[03:06] <CIA-40> ffmpeg: DSPContext.bswap_buf() requires aligned output
[03:06] <CIA-40> ffmpeg: 03Justin Ruggles 07master * rddf70db6d7 10ffmpeg/libavcodec/adpcmenc.c: 
[03:07] <CIA-40> ffmpeg: The new name is more appropriate as only golomb functions are tested.
[03:07] <CIA-40> ffmpeg: 03Diego Biurrun 07master * r9e3a2736bd 10ffmpeg/libavcodec/golomb-test.c: golomb-test: K&R formatting cosmetics
[03:07] <CIA-40> ffmpeg: 03Diego Biurrun 07master * r52afc97168 10ffmpeg/tests/ (fate/libavcodec.mak ref/fate/golomb): fate: add golomb-test
[03:07] <CIA-40> ffmpeg: 03Diego Biurrun 07master * rc30b198381 10ffmpeg/libavcodec/x86/ (fmtconvert_mmx.c rv40dsp.c): 
[03:07] <CIA-40> ffmpeg: x86: Place mm_flags variable declaration below the appropriate #ifdef.
[03:07] <CIA-40> (14 lines omitted)
[05:08] <CIA-40> ffmpeg: 03Michael Niedermayer 07master * r4b4d3d7249 10ffmpeg/libavcodec/aacenc.c: 
[05:08] <CIA-40> ffmpeg: aacenc: fix library references
[05:08] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:08] <CIA-40> ffmpeg: 03Michael Niedermayer 07master * rdbb7009ea5 10ffmpeg/libavformat/westwood_aud.c: 
[05:08] <CIA-40> ffmpeg: westwood_aud: Fix project reference
[05:08] <CIA-40> ffmpeg: "Copyright (c) 2003 The ffmpeg Project" and "part of libav"
[05:08] <CIA-40> ffmpeg: is ROTFL
[05:08] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:08] <CIA-40> ffmpeg: 03Michael Niedermayer 07master * rada28b7f72 10ffmpeg/libavcodec/golomb-test.c: 
[05:08] <CIA-40> ffmpeg: h264 golomb: Fix copyright violation.
[05:08] <CIA-40> ffmpeg: Copy and paste does not change authorship
[05:08] <CIA-40> ffmpeg: See 0da71265 for the original checkin
[05:08] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:08] <CIA-40> ffmpeg: 03Michael Niedermayer 07master * r92f9b26c1e 10ffmpeg/doc/muxers.texi: 
[05:08] <CIA-40> ffmpeg: doc: fix convertion tool references
[05:08] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:09] Last message repeated 1 time(s).
[05:09] <CIA-40> ffmpeg: 03Michael Niedermayer 07master * r9fdae05607 10ffmpeg/libavformat/avformat.h: 
[05:09] <CIA-40> ffmpeg: libavformat/avformat.h: fix stray libavformat reference
[05:09] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:09] <durandal_1707> funny stuff
[05:32] <Compn> lavf: add functions for accessing the fourcc<->CodecID mapping tables.
[05:32] <Compn> that is an odd commit
[05:33] <Compn> i dont even know what its doing
[05:33] <Compn> well i know what its doing, just not why
[09:12] <j-b> Compn: http://git.videolan.org/?p=libpostproc.git;a=shortlog
[13:38] <Compn> j-b : whyd libav fork postproc? :P
[13:38] Action: Compn not involved :P
[13:41] <ubitux> :)
[13:52] <JEEB> The CodecID <=> fourcc thing was IIRC done as it was useful for some people implementing their own parsers but wanting to use libavcodec to decode.
[13:53] <nevcairiel> indeed, I use these tables in my mkv demuxer
[13:53] <Compn> riff was put into libavformat.o
[13:54] <Compn> or .a, whichever
[13:54] <Compn> oh nm
[13:54] <Compn> thats needed for new demuxer anyway
[15:52] <CIA-40> ffmpeg: 03Michael Bradshaw 07master * rec0d02e441 10ffmpeg/libavcodec/libopenjpegdec.c: Support decoding gray8a with libopenjpeg.
[15:52] <CIA-40> ffmpeg: 03Michael Bradshaw 07master * rd76864d9ce 10ffmpeg/libavcodec/libopenjpegdec.c: Support decoding rgba64 with libopenjpeg.
[16:13] <ubitux> bcoudu, bcoudurier: i'll push the new timecode api (the fullpatch) pretty soon; if you want to comment, please do it soon
[16:40] <ubitux> btw, i'm going to push the lavfi stuff too
[18:05] <CIA-40> ffmpeg: 03Carl Eugen Hoyos 07master * r93a3ba9bcd 10ffmpeg/libavdevice/v4l2.c: Fix pix_fmt mapping for v4l2 rgb5x5.
[18:05] <CIA-40> ffmpeg: 03Carl Eugen Hoyos 07master * r4632066713 10ffmpeg/libavdevice/v4l2.c: Fix v4l pix_fmt mapping for rgb32: There is no transparency information.
[18:05] <CIA-40> ffmpeg: 03Carl Eugen Hoyos 07master * rb2327a8202 10ffmpeg/libavdevice/v4l2.c: Add some missing v4l2 pix_fmt mappings.
[19:11] <ubitux> btw, i updated the string interpolation hash, and used for (i = 0; i < len; i++) hash = (hash<<5) ^ (hash>>27) ^ key[i]; (starting at hash=0)
[19:12] <ubitux> seems it is more efficient with ascii keys
[19:12] <ubitux> which will certainly be the main kind of input certainly
[19:12] <ubitux> drop the duplicate.
[19:15] <ubitux> i also changed the way of dealing with the formated buffer
[19:35] <CIA-40> ffmpeg: 03Paul B Mahol 07master * r917803babd 10ffmpeg/libavcodec/ffv1.c: 
[19:35] <CIA-40> ffmpeg: ffv1enc: PIX_FMT_YUV440P support
[19:35] <CIA-40> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[19:35] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:35] <CIA-40> ffmpeg: 03Paul B Mahol 07master * rec09230c9a 10ffmpeg/libavcodec/ffv1.c: 
[19:35] <CIA-40> ffmpeg: ffv1dec: PIX_FMT_YUV440P support
[19:35] <CIA-40> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[19:35] <CIA-40> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:32] <Tjoppen> uh.. ffpresets need to be copied into ~/.ffmpeg for them to work?
[20:33] <Tjoppen> that seems.. dumb?
[20:42] <funman> or in /usr/share/ffmpeg ?
[20:42] <Tjoppen> that's where they are, but they don't work
[20:46] <michaelni> Tjoppen, there are 2 mxf patches on ffmpeg-dev
[20:48] <michaelni> #define FFMPEG_DATADIR "/usr/local/share/ffmpeg"
[20:48] <michaelni> maybe its missing to search in both /usr and /usr/local
[20:49] <Tjoppen> ok, will check. have been sick and without internet for a while
[20:49] <funman> michaelni: hmm
[20:49] <funman> you should respect --prefix= option
[20:50] <Tjoppen> my ffpresets are in /usr/share/ffmpeg/  (and ~/.ffmpeg/ of course)
[20:50] <funman> for example i use debian package, and build/install my own with --prefix=$HOME/.local/ffmpeg
[20:50] <michaelni> DATADIR=\$(DESTDIR)$datadir
[20:51] <michaelni> #define FFMPEG_DATADIR "$(eval c_escape $datadir)"
[20:51] <michaelni> i assume this does end up respecting --prefix but ive not checked
[20:51] <funman> probably
[20:51] <michaelni> the quote from before was from my config.h
[23:49] <Daemon404> bcoudurier, https://github.com/dwbuiten/postproc/commit/5a5d853ed3aedcffe7eab235a43460e3ebb67149
[00:00] --- Thu Feb  2 2012