[Ffmpeg-devel-irc] ffmpeg-devel.log.20120608

[burek]

[01:37] <ubitux> http://googleresearch.blogspot.in/2012/05/video-stabilization-on-youtube.html  anyone tried this with our vf deshake?
[01:40] <ubitux> default settings doesn't look efficient :(
[01:44] <ubitux> michaelni: thx for changing the default description in the trac :)
[01:53] <ubitux> (deshake ’ Ticket1430)
[02:01] <Compn> i remember when you could read blogs without javascript :P
[02:05] <ubitux> yeah&
[02:05] <ubitux> that kind of blog sucks so much&
[02:55] <Zeranoe> I'm wondering if anyone familiar with git can tell me how I can see the modification date of the last change. git log is currently reporting the last change was in may
[02:58] <michaelni> Zeranoe, git log --pretty=fuller
[02:59] <llogan> Zeranoe: did you 'git pull'?
[03:01] <Zeranoe> llogan: Yes, apparently git log --date=iso displays the AuthorDate not the CommitDate
[03:01] Action: llogan is still a git noob
[03:11] <Zeranoe> michaelni: Thanks, that did the trick
[04:06] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r80485a4f9a 10ffmpeg/libavformat/mpegtsenc.c: 
[04:06] <CIA-119> ffmpeg: mpegtsenc: make pts/dts offseting optional
[04:06] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[07:50] <ubitux> Daemon404: i just used the "side-by-side feature" of lavfi we were talking about
[07:50] <ubitux> -vf 'split[a][b]; [a]pad=iw*2:ih[src]; [b]deshake[filt]; [src][filt]overlay=w'
[07:50] <ubitux> here is the magic
[08:03] Action: ubitux now wonders how it could work with a frame dropping filter...
[11:45] <ubitux> mmh
[11:45] <ubitux> [src @ 0x2ebb040] Changing frame properties on the fly is not supported.  
[11:45] <ubitux> this looks like a regression to me
[11:45] <ubitux> (ffplay -f lavfi -i cellauto)
[11:46] <ubitux> huh, life too looks broken
[11:54] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rd5d30d56ed 10ffmpeg/libavcodec/sinewin_tablegen.h: 
[11:54] <CIA-119> ffmpeg: Revert "sinewin_tablegen: use av_assert"
[11:54] <CIA-119> ffmpeg: using as_assert() in the table generator would make the generators
[11:54] <CIA-119> ffmpeg: have to be linked to libavutil.
[11:54] <CIA-119> ffmpeg: This reverts commit c3b6efa4cb25400d7ff8853ee6ebd95c29b744a6.
[12:04] <michaelni> ubitux, is that ffplay specific ?
[12:19] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r5b51efdc3a 10ffmpeg/ (libavutil/bprint.c tests/ref/fate/bprint): 
[12:19] <CIA-119> ffmpeg: fate-bprint: try to workaround platform specific value.
[12:19] <CIA-119> ffmpeg: this should fix some fate failures
[12:19] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[12:50] <CIA-119> ffmpeg: 03Paul B Mahol 07master * rf4e0ec814b 10ffmpeg/libavutil/pixdesc.c: 
[12:50] <CIA-119> ffmpeg: pixdesc: BGRA64 is rgb pixel format
[12:50] <CIA-119> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[13:59] <ubitux> michaelni: i didn't look further
[14:00] <ubitux> it seems to pass with ffmpeg
[14:00] <ubitux> mmh the ffmpeg output doesn't play
[14:01] <ubitux> ah wait my bad, forget this last sentence
[14:15] <michaelni> ubitux, regression since merging 67339f6eb41299096dd7de94f557b487ec3477ff
[14:16] <michaelni> that is if one reverts that things work
[14:36] <michaelni> ubitux, fixed locally
[14:37] <ubitux> michaelni: ok :)
[14:38] <michaelni> ill send a patch in a moment
[15:05] <CIA-119> ffmpeg: 03Paul B Mahol 07master * re11020a145 10ffmpeg/libavcodec/Makefile: 
[15:05] <CIA-119> ffmpeg: Fix building libopencore_amrnb decoder without libopencore_amrnb encoder
[15:05] <CIA-119> ffmpeg: Fixes #1210.
[15:05] <CIA-119> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[15:23] <durandal11707> michaelni: why rgb2rgb.c is not killed and all functions moved to rgb2rgb_template.c and made static?
[15:26] <durandal11707> and adding sws_ will make it global instead local as it is now...
[15:34] <durandal11707> hmm version files list functions which no longer exist
[15:36] <michaelni> durandal11707, swsi_ for "sws internal" could be used
[15:37] <iive> are masters done with mangling of swscale?
[15:41] <michaelni> and yes rgb2rgb.c and rgb2rgb_template.c could be merged
[15:42] <durandal11707> but that could cuase headache when merging anything useful from qatar....
[15:48] <CIA-119> ffmpeg: 03Carl Eugen Hoyos 07master * rb8e2dcd266 10ffmpeg/libavformat/Makefile: 
[15:48] <CIA-119> ffmpeg: Fix jacosub muxer Makefile dependencies.
[15:48] <CIA-119> ffmpeg: Found, analysed and tested by trac user Jamal.
[15:48] <CIA-119> ffmpeg: Fixes part of ticket #1428.
[15:48] <CIA-119> ffmpeg: 03Carl Eugen Hoyos 07master * rb8fe9f08be 10ffmpeg/ (configure libavcodec/Makefile): 
[15:48] <CIA-119> ffmpeg: Fix amv encoder dependencies.
[15:48] <CIA-119> ffmpeg: Found, analysed and tested by trac user Jamal.
[15:48] <CIA-119> ffmpeg: Fixes part of ticket #1428.
[15:48] <CIA-119> ffmpeg: 03Carl Eugen Hoyos 07master * rb35d38bd04 10ffmpeg/configure: 
[15:48] <CIA-119> ffmpeg: Add iac build dependency.
[15:48] <CIA-119> ffmpeg: Found and analysed by trac user Jamal.
[15:48] <CIA-119> ffmpeg: Fixes the remaining part of ticket #1428.
[15:48] <CIA-119> ffmpeg: 03Carl Eugen Hoyos 07master * r37a9cdfc2c 10ffmpeg/ (configure libavcodec/Makefile): 
[15:48] <CIA-119> ffmpeg: Fix dirac decoder build dependencies.
[15:48] <CIA-119> ffmpeg: Found, analysed and tested by trac user Jamal.
[15:48] <CIA-119> ffmpeg: Fixes part of ticket #1428.
[15:48] <CIA-119> ffmpeg: 03Carl Eugen Hoyos 07master * r8a880d6918 10ffmpeg/libavformat/Makefile: 
[15:48] <CIA-119> ffmpeg: Fix latm muxer Makefile dependencies.
[15:48] <CIA-119> ffmpeg: Found, analysed and tested by trac user Jamal.
[15:48] <CIA-119> ffmpeg: Fixes part of ticket #1428.
[15:49] <durandal11707> omg
[15:49] <durandal11707> i need to revert last commit
[15:56] <CIA-119> ffmpeg: 03Paul B Mahol 07master * r134d0f7119 10ffmpeg/libavcodec/Makefile: 
[15:56] <CIA-119> ffmpeg: Revert "Fix building libopencore_amrnb decoder without libopencore_amrnb encoder"
[15:56] <CIA-119> ffmpeg: Reverting "Fix" because it disables amrwb decoder.
[15:56] <CIA-119> ffmpeg: This reverts commit e11020a1459de122834055be93a70566ce064bcc.
[16:25] <michaelni> ubitux, do you want / have time to review a srtenc patch ?
[16:26] <michaelni> i mean [FFmpeg-devel] [PATCH] Make srtenc accept text-only (type SUBTITLE_TEXT) subtitles
[16:29] <ubitux> afaik we didn't decide what to do about this; i don't have much opinion on it
[16:30] <ubitux> the patch is also adding a ASS case (??)
[16:30] <ubitux> (if we are talking about http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2012-May/125115.html)
[16:30] <ubitux> i don't have much time to look much more into this 
[16:38] <michaelni> ubitux, ok, i was just asking as wolfram "pinged" it
[16:42] <ubitux> yep just saw it
[18:30] <CIA-119> ffmpeg: 03Mike Scheutzow 07master * r1be7bfdd2b 10ffmpeg/ffmpeg.c: 
[18:30] <CIA-119> ffmpeg: ffmpeg.c: fix erroneous "Output file is empty" message.
[18:30] <CIA-119> ffmpeg: This message is printed whenever the RAWVIDEO codec is used, whether the
[18:30] <CIA-119> ffmpeg: output file contains any frames or not. Test command:
[18:30] <CIA-119> ffmpeg: ./ffmpeg -i infile.y4m outfile.y4m
[18:30] <CIA-119> ffmpeg: Signed-off-by: Mike Scheutzow <mjs973 at optonline.net>
[18:30] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:30] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r718607be28 10ffmpeg/libavcodec/rv10.c: 
[18:30] <CIA-119> ffmpeg: rv10: change assert() to av_assert()
[18:30] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:03] <RobertNagy> what is AVFrame->pkt_pos useful?
[19:03] <RobertNagy> when*
[19:31] <michaelni> RobertNagy, might be usefull for seeking by byte position
[19:31] <michaelni> might be usefull if you want to know where some damage is in a file that you detected in a frame
[19:32] <RobertNagy> ah
[19:32] <RobertNagy> michaelni: on a different note, I'm having some issues woth AVFrame->opaque and lavfi
[19:33] <RobertNagy> it is a bit complicated to keep the AVFrame->opaque through a filter
[19:33] <RobertNagy> right now I'm forced to hijack pkt_pos
[19:34] <RobertNagy> any suggestions?
[19:34] <RobertNagy> I guess I could create a dictionary based on pkt_pos and then re-map based on pkt_pos when I get frames back from the filter graph
[19:41] <michaelni> maybe lavfi could be extended to support passing user data through the filter graph attached to the frames
[19:48] <RobertNagy> so basicaly somethign like an opaque field in AVFilterBuffer?
[19:49] <RobertNagy> though I'm unsure hwo that would work when dropping or creating new frames inside a filter
[19:54] <michaelni> In principle one could add a drop/clone callback to handle that but maybe thats overkill, what exactly do you use that opaque field for ?
[19:56] <RobertNagy> I keep track of how many codec flushes has occured after a frame is decoded
[19:58] <michaelni> not sure i understand you, what do you mean by "codec flushes" ? 
[19:59] <RobertNagy> avcodec_flush_buffers
[19:59] <RobertNagy> frame->flush_count = codec_flush_count++;
[19:59] <RobertNagy> nvm the last one
[20:00] <michaelni> hmm, and why do you need to keep trackl of that and pas it through the filter layer ?
[20:00] <RobertNagy> when syncing video and audio
[20:00] <RobertNagy> I need to know whether there has been a time rollover
[20:00] <RobertNagy> or that the audio is just behind the video
[20:02] <RobertNagy> so if the flush_count for audio and video is not the same, I know that I have to keep decoding frames
[20:02] <RobertNagy> the time rollover can occur when seeking
[20:03] <michaelni> but what if theres simply a timestamp discontinuity ?
[20:03] <michaelni> you would have the same effect but no ++
[20:03] <RobertNagy> I hope that there is no such discontinuity :)
[20:04] <RobertNagy> not sure what kind of discontinuities can occur
[20:04] <michaelni> any that differ from 1/fps IIRC :)
[20:05] <RobertNagy> but can e.g. the next time_stamp be earlier before the current, without any seek being involved
[20:05] <RobertNagy> if not then I don't see any problem
[20:05] <michaelni> it absolutely can
[20:05] <RobertNagy> hm
[20:06] <RobertNagy> this is quite tricky
[20:06] <michaelni> i remember a file (MPEG-PS IIRC) that had timestamp resets to 0 every minute or so
[20:06] <RobertNagy> not sure what to do then
[20:06] <michaelni> not sure how common that is out in the wild
[20:06] <RobertNagy> previously I detected the rollower
[20:06] <RobertNagy> with something like
[20:06] <RobertNagy> next_pts - cur_pts < -20
[20:07] <RobertNagy> but then I noticed problems with mpegts streams
[20:07] <RobertNagy> where audio and video start at different timestamps
[20:07] <RobertNagy> and had to instead the treshold
[20:07] <RobertNagy> but if I do that I cannot seek into that range
[20:07] <RobertNagy> increase the treshold*
[20:08] <RobertNagy> I think I will simply not support streams with such annoying discontinuities
[22:05] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rdfde3d497e 10ffmpeg/libswresample/rematrix.c: 
[22:05] <CIA-119> ffmpeg: swr: add inrermediate sample type for rematrixing
[22:05] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:05] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rca30ae12c5 10ffmpeg/libswresample/ (rematrix.c swresample_internal.h): 
[22:05] <CIA-119> ffmpeg: swr: add all in one rematrixing function ptr
[22:05] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:05] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rb0348158a0 10ffmpeg/libswresample/ (rematrix.c rematrix_template.c): 
[22:05] <CIA-119> ffmpeg: swr: 5.1->stereo double, float and s16
[22:05] <CIA-119> ffmpeg: 2.5 times faster than the generic code
[22:05] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:34] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * re016e3c9d7 10ffmpeg/libswresample/rematrix_template.c: 
[22:34] <CIA-119> ffmpeg: swr: add mix8to2
[22:34] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Sat Jun  9 2012