[Ffmpeg-devel-irc] ffmpeg-devel.log.20120809

burek burek021 at gmail.com
Fri Aug 10 02:05:02 CEST 2012


[00:22] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r0325e01fe7 10ffmpeg/doc/demuxers.texi: doc/demuxers: itemize image2 examples and put them in a dedicated subsection
[00:23] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r3a06ea8436 10ffmpeg/ (3 files in 2 dirs): 
[00:23] <CIA-41> ffmpeg: lavf/img2dec: add and document pattern_type option
[00:23] <CIA-41> ffmpeg: Allow to override the default 'glob_sequence' value, which is deprecated
[00:23] <CIA-41> ffmpeg: in favor of the new 'glob' and 'sequence' options.
[00:23] <CIA-41> ffmpeg: The new pattern types should be easier on the user since they are more
[00:23] <CIA-41> ffmpeg: predictable than 'glob_sequence', and do not require awkward escaping.
[00:23] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r7794ddeae3 10ffmpeg/doc/ffmpeg.texi: (log message trimmed)
[00:23] <CIA-41> ffmpeg: doc/ffmpeg: do not show examples making use of the deprecated glob_sequence pattern type
[00:23] <CIA-41> ffmpeg: The syntax of the 'glob_sequence' pattern type is explained in the image2
[00:23] <CIA-41> ffmpeg: manual section, so there is no need to replicate it here.
[00:23] <CIA-41> ffmpeg: Also discourage the use of such deprecated pattern type by not showing
[00:23] <CIA-41> ffmpeg: examples employing it.
[00:23] <CIA-41> ffmpeg: Replace the syntax explanation with an use example of the 'glob' pattern
[00:23] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r083754f2be 10ffmpeg/doc/indevs.texi: 
[00:23] <CIA-41> ffmpeg: doc/indevs: rework options description in the x11 grab section
[00:23] <CIA-41> ffmpeg: Create a dedicated options subchapter, rather than a subchapter for each
[00:23] <CIA-41> ffmpeg: option.
[00:23] <CIA-41> ffmpeg: This is more consistent with the rest of the docs. Also apply misc minor
[00:23] <CIA-41> ffmpeg: fixes for improving rendering.
[00:23] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * rf09a06908e 10ffmpeg/doc/indevs.texi: 
[00:23] <CIA-41> ffmpeg: doc/indevs: use separate @example blocks for separate examples
[00:23] <CIA-41> ffmpeg: Improve rendering.
[00:23] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * rc93b796c3e 10ffmpeg/libavdevice/x11grab.c: 
[00:23] <CIA-41> ffmpeg: lavd/x11grab: apply misc consistency fixes to options definition
[00:24] <CIA-41> ffmpeg: In particular, do not upcase first word, do not use final dot, use a verb
[00:24] <CIA-41> ffmpeg: to specify what the option does, sort entries by name, apply random
[00:24] <CIA-41> ffmpeg: vertical align.
[00:24] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * rbe2b885738 10ffmpeg/doc/indevs.texi: 
[00:24] <CIA-41> ffmpeg: doc/indevs: document missing options in x11grab section
[00:24] <CIA-41> ffmpeg: Add documentation for options draw_mouse, framerate, and video_size.
[00:24] <CIA-41> ffmpeg: In particular, fix trac ticket #1314.
[00:24] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r4270d8c04d 10ffmpeg/libavdevice/x11grab.c: 
[00:24] <CIA-41> ffmpeg: lavd/x11grab: deprecate "nomouse" specification in filename
[00:24] <CIA-41> ffmpeg: Favor draw_mouse option. Should simplify parsing, and make it more
[00:24] <CIA-41> (1 lines omitted)
[00:26] <ubitux> michaelni: i've commented all i actually could in the 6-patches patchset from nicolas (subs related)
[00:27] <ubitux> so don't wait for more review from me if you want this applied
[00:28] <ubitux> saste: showspectrum delayed; too tired to work on it now
[00:28] <ubitux> but feel free to review afain edgedetect so i can make my mind at peace ;)
[00:28] <saste> yes no hurry
[00:29] <ubitux> (again*)
[00:35] <ubitux> saste: you know you can re-order commits with git rebase -i?
[00:36] <saste> ubitux: I do that all the time
[00:36] Action: ubitux regularly use ``git rebase -i master`` from his dev branches
[00:36] <ubitux> ok
[00:36] <ubitux> :)
[00:36] <saste> but if the order changes locally, the order of the posted patches won't
[00:36] <saste> unless i resend the whole patchset, which i find awkward on the reviewer
[00:37] <ubitux> ok
[00:37] <saste> (mutt shows the thread nicely, and I'm used to review the whole discussion sometimes, splitting the thread doesn't help)
[00:37] <saste> it's not like I don't know how to use git, it is that this workflow works fine for me
[00:37] <saste> the main drawback is that many times you can't git apply the patch on master
[00:38] <saste> but for docu patches i don't think that's really an issue
[00:41] <ubitux> ok ok :)
[00:47] <ubitux> 'night folks
[01:02] <CIA-41> ffmpeg: 03Derek Buitenhuis 07master * r4ab26cb4cc 10ffmpeg/configure: 
[01:02] <CIA-41> ffmpeg: configure: Add support for e500mc CPU
[01:02] <CIA-41> ffmpeg: Signed-off-by: Derek Buitenhuis <derek.buitenhuis at gmail.com>
[01:02] <CIA-41> ffmpeg: 03Mans Rullgard 07master * rec9d2c15c1 10ffmpeg/ (3 files in 2 dirs): 
[01:02] <CIA-41> ffmpeg: ARM: use Q/R inline asm operand modifiers only if supported
[01:02] <CIA-41> ffmpeg: Some compilers do not support the Q/R modifiers used to access
[01:02] <CIA-41> ffmpeg: the low/high parts of a 64-bit register pair. Check for this
[01:02] <CIA-41> ffmpeg: and disable all uses of it when not supported.
[01:02] <CIA-41> ffmpeg: Fixes bug #337.
[01:02] <CIA-41> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[01:02] <CIA-41> ffmpeg: 03Samuel Pitoiset 07master * r12127b65b0 10ffmpeg/libavformat/rtmpproto.c: 
[01:02] <CIA-41> ffmpeg: rtmp: Factorize simultaneous URLProtocol / AVClass declarations with a macro
[01:02] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[01:02] <CIA-41> ffmpeg: 03Samuel Pitoiset 07master * rf9e77c1762 10ffmpeg/libavformat/rtmpproto.c: 
[01:02] <CIA-41> ffmpeg: rtmp: Add support for subscribing live streams
[01:02] <CIA-41> ffmpeg: When streaming live streams using the Akamai, Edgecast or Limelight CDN,
[01:02] <CIA-41> ffmpeg: players cannot simply connect to the live stream. Instead, they have to
[01:02] <CIA-41> ffmpeg: subscribe to it, by sending an FC Subscribe call to the server.
[01:02] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[01:02] <CIA-41> ffmpeg: 03Mans Rullgard 07master * rdfd9159f3d 10ffmpeg/ (.gitignore Makefile common.mak library.mak): (log message trimmed)
[01:02] <CIA-41> ffmpeg: build: change checkheaders to use regular build rules
[01:02] <CIA-41> ffmpeg: Many compilers need special flags to compile *.h files as regular
[01:02] <CIA-41> ffmpeg: source code, if they will do so at all. Rather than hoping all
[01:02] <CIA-41> ffmpeg: compilers will have such a flag and adding mappings for it, create
[01:02] <CIA-41> ffmpeg: wrapper .c files for test building single headers.
[01:02] <CIA-41> ffmpeg: This allows using the regular rule for compiling C files without the
[01:02] <CIA-41> ffmpeg: 03Diego Biurrun 07master * ra2b4d1df52 10ffmpeg/library.mak: build: Remove all installed headers and header directories on uninstall
[01:03] <CIA-41> ffmpeg: g723_1: use all LPC vectors in formant postfilter
[01:03] <CIA-41> ffmpeg: Due to some mistake LPC vector for the first subframe was used for all
[01:03] <CIA-41> (47 lines omitted)
[01:06] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * rf6580b50ed 10ffmpeg/ (6 files in 3 dirs): 
[01:06] <CIA-41> ffmpeg: lavfi: add asetpts audio filter
[01:06] <CIA-41> ffmpeg: Based on an idea by Andrey Utkin <andrey.krieger.utkin at gmail.com>.
[01:16] <saste> michaelni: should I close #1353?
[01:21] <durandal11707> michaelni: that aasc crash should be fixed with my patch that adds 8bit support
[01:25] <michaelni> saste, you already closed it :)
[01:46] <durandal11707> michaelni: something is fishy with aasc samples we have on fate - it reports pix_fmt change - which is nonsense
[01:49] <durandal11707> patch i wrote introduce such strange behaviour ... 
[01:52] <durandal11707> lol what a FAIL = vs ==
[01:53] Action: michaelni is blind
[01:54] <michaelni> i should have spoted that = vs == thing ...
[01:55] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rae199ab4ae 10ffmpeg/libavutil/integer.c: 
[01:55] <CIA-41> ffmpeg: integer: fix typos in asserts
[01:55] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:55] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rf79f25e948 10ffmpeg/libavutil/integer.c: 
[01:55] <CIA-41> ffmpeg: integer: switch to av_assert
[01:55] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:55] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r4ec03d1386 10ffmpeg/libswresample/resample.c: 
[01:55] <CIA-41> ffmpeg: bessel: check that the function terminates as expected by av_assert2().
[01:55] <CIA-41> ffmpeg: A clear abort is better than wrong output and a possible crash.
[01:55] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:02] <CIA-41> ffmpeg: 03Paul B Mahol 07master * r8a57ca5c6a 10ffmpeg/libavcodec/aasc.c: 
[02:02] <CIA-41> ffmpeg: aasc: fix out of array write
[02:02] <CIA-41> ffmpeg: Closes #1619.
[02:02] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[02:02] <CIA-41> ffmpeg: 03Paul B Mahol 07master * rbd971ddaf4 10ffmpeg/libavcodec/aasc.c: 
[02:02] <CIA-41> ffmpeg: aasc: 8bit support
[02:02] <CIA-41> ffmpeg: Closes #1319.
[02:02] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[03:58] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rb41d481aa4 10ffmpeg/common.mak: 
[03:58] <CIA-41> ffmpeg: common.mak: add forgotten hunk from last merge
[03:58] <CIA-41> ffmpeg: Found-by: jamal <jamrial at gmail.com>
[03:58] <CIA-41> ffmpeg: Author of the hunk: Mans Rullgard <mans at mansr.com>
[03:58] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:09] <CIA-41> ffmpeg: 03jamal 07master * r5d8e54fe1d 10ffmpeg/common.mak: 
[05:09] <CIA-41> ffmpeg: checkheaders: Rename HEADEROBJS to HOBJS
[05:09] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:09] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r2cdb0810f1 10ffmpeg/ffmpeg.c: 
[05:09] <CIA-41> ffmpeg: ffmpeg: reduce difference to qatar by a few lines
[05:09] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[08:10] <CIA-41> ffmpeg: 03Clément BSsch 07master * rcc650cf029 10ffmpeg/ (23 files in 2 dirs): 
[08:10] <CIA-41> ffmpeg: Remove "Error parsing options string [...]" messages.
[08:10] <CIA-41> ffmpeg: This reduces from 3 to 2 messages for the same syntax error in ffprobe,
[08:10] <CIA-41> ffmpeg: and from 4 to 3 in filters.
[09:32] <ubitux> saste: maybe the documentation could be splitted into new man pages?
[09:33] <ubitux> something like man ffmpeg-indevs, man ffmpeg-syntax, etc
[09:33] <ubitux> (and we would add a reference within each tool documentation)
[09:35] <saste> ubitux: that was my first proposal and was rejected
[09:36] <ubitux> ah? do you have the thread for this?
[09:36] <saste> so i hacked texi2pod.pl to support includes
[09:36] <ubitux> :)
[09:38] <saste> http://thread.gmane.org/gmane.comp.video.ffmpeg.devel/108669/
[09:38] <ubitux> thank you :)
[10:30] <highgod> Hi,all.I want ask a question about avfilter
[10:34] <saste> michaelni: is there a particular reason for which we have both AVFrame accessors and options?
[10:34] <saste> highgod: don't ask to ask just ask
[10:34] <saste> also API/usage questions => #ffmpeg-user
[10:35] <highgod> when I use command "ffmpeg -i ./testfile/test_vc1.wmv -vf "hqdn3d=240:2:2:2,unsharp=3:3:5,yadif,deshake,tinterlace=5,scale=720:480" ./testfile/output.mkv" to test filter,I found that the function "avfilter_end_frame" run very slowly 
[10:35] <highgod> almost 100ms
[10:36] <highgod> I read the function,I just found that it just release buffer operation.But if I run a single filter such as just run yadif filter, it run very fast
[10:36] <highgod> Thanks
[10:37] <saste> highgod: hqdn3d, unsharp, yadif, deshake, scale
[10:37] <highgod> yes
[10:37] <saste> they are all heavy filters, the more you have the slower it gets
[10:39] <highgod> but I don't know way the function "avfilter_end_frame"run so slowly.
[10:41] <highgod> I use tools to get the function running duration.and I found that it is the most time cost part 
[10:41] <ubitux> because some filters do all the processing in the end_frame callback?
[10:41] <highgod> 157599ms
[10:42] <highgod> sorry
[10:42] <highgod> 157599us
[10:44] <saste> highgod: hints: check every filter one by one, enable optims (some filter have asm optimizations), remove filters until you get a reasonable performace/quality ratio
[10:50] <highgod> @saste:I have enable optims.If I run yadif single,the function avfilter_end_frame just cost 8.9862us
[10:52] <ubitux> so?
[10:55] <highgod> @ubitux:So I'm so confused that why? Can you teach me anything about the "end_frame" callback?
[10:56] <ubitux> the more filters you queue, the more slower it gets
[10:56] <ubitux> in most libavfilter filters, you will have a end_frame callback
[10:56] <ubitux> git grep end_frame in libavfilter/
[10:56] <highgod> Yes I use tools to analyses the squence charts about function run order
[10:57] <ubitux> and in that callback sometimes most of the processing is done by the filter
[11:08] <saste> what was the point of "exploding" ffmpeg.c into small pieces?
[11:08] <ubitux> making noise
[11:10] <highgod> @ubitux:Thanks,I am not very familar with filter pipeline.I saw that some filters has start_frame draw_slice end_frame,how dose ffmpeg.exe calls them?
[11:11] <ubitux> check doc/filters_design.txt, doxy, and source code
[11:12] <highgod> Thanks
[11:33] <CIA-41> ffmpeg: 03Nicolas George 07master * raed032c25b 10ffmpeg/ (ffmpeg.c libavcodec/dvbsub.c): (log message trimmed)
[11:33] <CIA-41> ffmpeg: dvbsub: fix encoding of termination packets.
[11:33] <CIA-41> ffmpeg: The old code generates a termination packet with the same regions as the
[11:33] <CIA-41> ffmpeg: start packet and page_state set to "only what changed"; the result is
[11:33] <CIA-41> ffmpeg: that the termination packet is decoded as identical to the start packet.
[11:33] <CIA-41> ffmpeg: The new code does as found in some DVB broadcasts: produce a packet with
[11:33] <CIA-41> ffmpeg: no regions. This is done by expecting num_rects to be 0 rather than
[11:33] <CIA-41> ffmpeg: 03Nicolas George 07master * re13df05fdb 10ffmpeg/libavcodec/dvbsub.c: dvbsub: reindent after last commit.
[12:03] <j-b> michaelni: ping
[12:04] <j-b> http://streams.videolan.org/ffmpeg/ all the files that were in the incoming and that made some kind of sense
[12:14] <CIA-41> FFmpeg: 03Michael Niedermayer 07master * rb41d481 10/ common.mak : common.mak: add forgotten hunk from last merge - http://git.io/B0BMDg
[12:14] <CIA-41> FFmpeg: 03Michael Niedermayer 07master * r2cdb081 10/ ffmpeg.c : ffmpeg: reduce difference to qatar by a few lines - http://git.io/mtdwfQ
[12:14] <CIA-41> FFmpeg: 03jamal 07master * r5d8e54f 10/ common.mak : checkheaders: Rename HEADEROBJS to HOBJS - http://git.io/nM97Kw
[13:09] <saste> should I add AV_OPT_FLAG_FILTERING_PARAM?
[13:13] <michaelni> saste, AVOptions are needed to allow generic access (like listing/enumerating what fields are available)
[13:13] <saste> michaelni: i'm trying to list internal options stored in filters
[13:14] <saste> opt_list will show the options only if they have at least one flag set (opt->flags != 0)
[13:14] <michaelni> i dont think the accessors are really nneeded, AVOptions would be enough, i think nicols or someone else preferred the accessors 
[13:15] <saste> michaelni: ok but now we have two distinct interfaces
[13:15] <saste> which are not synched
[13:15] <saste> indeed i was not even aware of AVField options
[13:16] <saste> *AVFrame options
[13:16] <michaelni> they are desync ?
[13:16] <michaelni> which one?
[13:19] <saste> for example i recently added nb_channels
[13:19] <michaelni> btw, off topic but our  texi2pod.pl seems to be run without anything checking that perl is installed
[13:19] <michaelni> thats why haiku is red on fate ...
[13:20] <saste> oh...
[13:22] <michaelni> maybe some test could be written to check that no (compatible) field misses an AVOption
[13:22] <michaelni> or that no accessor misses an avoption
[13:23] <michaelni> would as a sideeffect also enforce some strict naming convention so easy grep compare is possible
[13:24] <michaelni> that is having part of the accessor name match the field name
[13:32] <saste> that's not always true, also sometimes we give two names to the same field (aliasing)
[13:35] <saste> why were AVFrame accessors favored over AVOptions in the first place?
[13:36] <saste> i suppose one of the reasons is that they're fast, and bypass all the avoptions bloat, which in case of AVFrame is relevant (not so much for generic component options)
[13:39] <michaelni> I dont mind if the accessors are deprecated/remvoed where it isnt meaningfull to keep them for speed reasons
[13:42] <saste> lunchtime
[15:32] <j-b> https://trac.videolan.org/vlc/ticket/6000 btw
[15:36] <Compn> atrac sample crash eh
[15:42] <CIA-41> ffmpeg: 03Anton Khirnov 07master * ra4f5011065 10ffmpeg/avconv.c: (log message trimmed)
[15:42] <CIA-41> ffmpeg: avconv: maintain sync on lavfi outputs.
[15:42] <CIA-41> ffmpeg: Before this commit, poll_filters() reads all frames available on each
[15:42] <CIA-41> ffmpeg: lavfi output. This does not work for lavfi sources that produce
[15:42] <CIA-41> ffmpeg: an unlimited number of frames, e.g. color and similar.
[15:42] <CIA-41> ffmpeg: With this commit, poll_filters() reads from output with the lowest
[15:42] <CIA-41> ffmpeg: timestamp and returns to wait for more input if no frames are available
[15:42] <CIA-41> ffmpeg: 03Anton Khirnov 07master * rd64901f99d 10ffmpeg/avconv.c: avconv: remove commented out cruft.
[15:42] <CIA-41> ffmpeg: 03Anton Khirnov 07master * rbed43d58b0 10ffmpeg/avconv.h: avconv: remove unused variable from InputFile.
[15:42] <CIA-41> ffmpeg: 03Anton Khirnov 07master * r0b26ef4228 10ffmpeg/ (avconv.c avconv.h): 
[15:42] <CIA-41> ffmpeg: avconv: replace no_packet array in transcode() with a var in InputStream
[15:42] <CIA-41> ffmpeg: This simplifies splitting code for reading from input out of
[15:42] <CIA-41> ffmpeg: transcode().
[15:42] <CIA-41> ffmpeg: 03Anton Khirnov 07master * rc0fbf9713c 10ffmpeg/avconv.c: 
[15:42] <CIA-41> ffmpeg: avconv: cosmetics
[15:42] <CIA-41> ffmpeg: Replace for (;foo == 0;) with while (!foo)
[15:42] <CIA-41> ffmpeg: This is prettier.
[15:42] <CIA-41> ffmpeg: 03Anton Khirnov 07master * r8391602913 10ffmpeg/avconv.c: 
[15:42] <CIA-41> ffmpeg: avconv: simplify transcode().
[15:42] <CIA-41> ffmpeg: Operate with a pointer to InputFile instead of its index in input_files.
[15:42] <CIA-41> ffmpeg: 03Anton Khirnov 07master * rf154ef1ae5 10ffmpeg/avconv.c: avconv: send EOF to lavfi even if flushing the decoder fails
[15:42] <CIA-41> ffmpeg: 03Anton Khirnov 07master * rcc2133b66a 10ffmpeg/avconv.c: avconv: get rid of pointless temporary variable.
[15:42] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rb7e9eea31f 10ffmpeg/: (log message trimmed)
[15:42] <CIA-41> ffmpeg: Merge commit 'f154ef1ae5b03f288dd8c025dab1884b4cb20c1a'
[15:42] <CIA-41> ffmpeg: * commit 'f154ef1ae5b03f288dd8c025dab1884b4cb20c1a':
[15:42] <CIA-41> ffmpeg:  avconv: send EOF to lavfi even if flushing the decoder fails
[15:42] <CIA-41> ffmpeg:  avconv: get rid of pointless temporary variable.
[15:42] <CIA-41> ffmpeg:  avconv: simplify transcode().
[15:42] <CIA-41> ffmpeg:  avconv: cosmetics
[18:25] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rfc97f8e0e2 10ffmpeg/ffmpeg.c: 
[18:25] <CIA-41> ffmpeg: ffmpeg: factorize process_input() out
[18:25] <CIA-41> ffmpeg: Based-on:
[18:25] <CIA-41> ffmpeg:  commit 0c00fd80ee4791bd70b634084307fc9f179e0412
[18:25] <CIA-41> ffmpeg:  Author: Anton Khirnov <anton at khirnov.net>
[18:25] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:43] <CIA-41> ffmpeg: 03Anton Khirnov 07master * r0c00fd80ee 10ffmpeg/avconv.c: avconv: split the code for processing input packets out of transcode()
[18:43] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * re1a983e601 10ffmpeg/: (log message trimmed)
[18:43] <CIA-41> ffmpeg: Merge commit '0c00fd80ee4791bd70b634084307fc9f179e0412'
[18:43] <CIA-41> ffmpeg: Just for metadata, this change has been integrated by hand already
[18:43] <CIA-41> ffmpeg: * commit '0c00fd80ee4791bd70b634084307fc9f179e0412':
[18:43] <CIA-41> ffmpeg:  avconv: split the code for processing input packets out of transcode()
[18:43] <CIA-41> ffmpeg: Conflicts:
[18:43] <CIA-41> ffmpeg:  ffmpeg.c
[19:35] <CIA-41> ffmpeg: 03Anton Khirnov 07master * r57d2422559 10ffmpeg/ (avconv.c avconv.h): 
[19:35] <CIA-41> ffmpeg: avconv: rename OutputStream.is_past_recording_time to finished.
[19:35] <CIA-41> ffmpeg: The new name is shorter and more accurate, since this variable is no
[19:35] <CIA-41> ffmpeg: longer used only for checking recording time constraint.
[19:35] <CIA-41> ffmpeg: 03Martin Storsjö 07master * rc864e461d2 10ffmpeg/libavformat/rtpdec_h264.c: 
[19:35] <CIA-41> ffmpeg: rtpdec: Remove an unused include
[19:35] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[19:35] <CIA-41> ffmpeg: 03Martin Storsjö 07master * rdf8cf076c8 10ffmpeg/libavformat/ (rtsp.c rtsp.h): 
[19:35] <CIA-41> ffmpeg: rtsp: Support receiving plain data over UDP without any RTP encapsulation
[19:35] <CIA-41> ffmpeg: EvoStream Media Server can serve data in this format, and
[19:35] <CIA-41> ffmpeg: VLC/live555 already supports it.
[19:35] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[19:35] <CIA-41> ffmpeg: 03Samuel Pitoiset 07master * rf89584ca44 10ffmpeg/libavformat/rtmpproto.c: 
[19:35] <CIA-41> ffmpeg: rtmp: Add message tracking
[19:35] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[19:35] <CIA-41> ffmpeg: 03Martin Storsjö 07master * r1243c72251 10ffmpeg/libavformat/ (rtsp.c rtsp.h version.h): 
[19:35] <CIA-41> ffmpeg: rtsp: Support mpegts in raw udp packets
[19:35] <CIA-41> ffmpeg: This is basically the same way as mpegts packets are parsed in
[19:35] <CIA-41> ffmpeg: rtpdec.c.
[19:35] <CIA-41> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[19:35] <CIA-41> ffmpeg: 03Mans Rullgard 07master * rc318626ce2 10ffmpeg/ (30 files in 7 dirs): 
[19:35] <CIA-41> ffmpeg: x86: rename libavutil/x86_cpu.h to libavutil/x86/asm.h
[19:35] <CIA-41> ffmpeg: This puts x86-specific things in the x86/ subdirectory where they
[19:35] <CIA-41> ffmpeg: belong.
[19:35] <CIA-41> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[19:35] <CIA-41> ffmpeg: 03Alex Converse 07master * r122d5c526a 10ffmpeg/libavcodec/aacdec.c: 
[19:46] <impiaaa> I made a small patch: http://pastebin.com/zq24pb2K
[19:48] <impiaaa> Any way I could go about getting it in trunk?
[19:54] <michaelni> impiaaa, best submit a patch to ffmpeg-devel
[19:54] <michaelni> ffmpeg-devel mailing list that is
[19:55] <impiaaa> okay, thanks
[20:26] <michaelni> j-b, Compn vlc/ticket/6000 no crash and nothing in valgrind || asan with ffmpeg ... -f null -
[20:38] <Compn> cant trust them vlc guys to make proper bugreport :P
[21:13] <kierank> michaelni: does vf_yadif require padded buffers somehow?
[21:42] <michaelni> kierank, i dont remember, but it should not, is that asm or plain C case ?
[21:42] <kierank> asm
[21:44] <kierank> http://pastebin.ca/2178764
[21:44] <kierank> (admittedly using the libav version)...
[21:47] <kierank> lemme see if i can make ffmpeg do the same
[22:07] <kierank> michaelni: happens with ffmpeg as well: http://pastebin.ca/index.php
[22:07] <michaelni> kierank, index.php ?
[22:08] <kierank> bah
[22:08] <kierank> sorry
[22:09] <kierank> http://pastebin.ca/2178771
[22:09] <kierank> this is using av_image_alloc with no padding around the edges
[22:10] <michaelni> is width and stride aligned to something like 16 ?
[22:12] <kierank> width is mod 16 and I would assume av_image_alloc would set a decent stride
[22:13] <michaelni> you mean width mod 16 == 0 i assume
[22:13] <kierank> yes
[22:14] <michaelni> how can i reproduce this ?
[22:14] <kierank> not sure. I'll see if it can be done with a simple test case
[22:15] <michaelni> also "X bytes before a block of size 3,114,256 alloc'd" for X=1,2,3
[22:15] <michaelni> this is odd too
[22:16] <michaelni> can you check what the actual strides are that are pased to yadif ?
[22:17] <kierank> hmmm i think i know why it breaks actually
[22:17] Action: michaelni has a suspicoun too 
[22:18] <kierank> it's because avframe and co has up to 8 pointers per plane
[22:18] <kierank> but in x264 it's still only 4
[22:18] <j-b> Compn: michaelni: probably fixed in the meantime
[22:27] <michaelni> kierank, i think theres a bug in yadif
[22:27] <kierank> yeah it's not to do with the pointers per plane thing
[22:29] <kierank> "linesize -603845600" dunno why i get that
[22:30] <kierank> whoops my mistake
[22:32] <kierank> linesize 1920
[22:32] <kierank> and 960 for the chroma as i would expect
[22:33] <kierank> michaelni: are you 100% sure it's a yadif bug?
[22:44] <michaelni> kierank, according to asan, yes
[22:44] <michaelni> but it could be yours is something else than what i found
[22:45] <kierank> fwiw mine happens in the c as well
[22:45] <kierank> i need to try some more things
[22:46] <michaelni> kierank, i think i know what it is, give me half an hour or so and ill fix it
[22:46] <kierank> cool, thanks. if there's anything i can do to help please ping me
[23:13] <ubitux> saste: i'll this the 255/256 thing, and the few other things on edgedetect this week end; about the generic gaussian stuff i'll pass for now
[23:13] <ubitux> s/this/fix/
[23:13] <ubitux> i'd like to get done with showspectrum
[23:13] <saste> ubitux: k
[23:13] <ubitux> (asetnsamples is sooo useful!)
[23:15] <ubitux> saste: about the FATE test for asetpts, i think it could be in filter.mak from the beginning since it doesn't require the random slicing stuff
[23:15] <saste> yes
[23:15] <ubitux> also if it is in lavfi-regression.sh it will likely be ported at some point, so it save some pain for the one doing it in the future
[23:16] <saste> yes but i still have to wrap my head around the ever changing fate stuff
[23:16] <ubitux> check the fate-filter-amix-transition
[23:16] <ubitux> i mean every amix tests actually
[23:17] <ubitux> they have some explicit ffmpeg calls with -ss
[23:41] <saste> i don't know how to generate the same output file produced by the lavfi-regression.sh tests
[23:41] <saste> i didn't look it very hard i confess
[23:41] <saste> and I don't want to validate again all the files "by eyes"
[00:00] --- Fri Aug 10 2012


More information about the Ffmpeg-devel-irc mailing list