[Ffmpeg-devel-irc] ffmpeg-devel.log.20120512

burek burek021 at gmail.com
Sun May 13 02:05:02 CEST 2012


[00:55] <RobertNagy> why was avfilter_get_audio_buffer removed?
[01:00] <michaelni> RobertNagy, the removial came from libav. Feel free to put the function(s) back if you need them
[01:01] <RobertNagy> hm, broke my code :(
[01:01] <michaelni> iam just merging these changes by default
[01:01] <michaelni> it isnt supposed to mean i considert them a good idea (or a bad idea)
[01:02] <michaelni> you have some filters outside libavfilter ?
[01:02] <RobertNagy> yes
[01:02] <RobertNagy> video_decode and audio_decode filters
[01:02] <RobertNagy> similar to ffplay
[01:03] <RobertNagy> thsi broke my audio_decode filter
[01:04] <michaelni> well, feel free to submit a patch that undoes the rename, or if the filter makes sense outside your application you could also submit it itself 
[01:04] <michaelni> so it could use the ff_*
[01:04] <Compn> michaelni : i think some devs think you approve some of these patches before you merge them
[01:05] <Compn> you might want to make it clear you are merging libav without RFC on ffmpeg lists
[01:05] <michaelni> i do skip really silly things when i spot them
[01:05] <Compn> i know :)
[01:05] <RobertNagy> they are not ready for merging into ffmpeg yet... my idea is to have a filter source that output video/audio packets to different outputs which are directed to appropriate decoding fitlers
[01:05] <beastd> it is the damn politics that can't be skipped easily...
[01:05] Action: Compn goes afk
[01:10] <ubitux> RobertNagy: can you describe a bit more the filter source?
[01:12] <RobertNagy> nothing complicated, just read av_read_frame and write the packets as AVFrame->data to appropriate output link
[01:12] <RobertNagy> then have a decoding filter decode it
[01:13] <RobertNagy> and continue from there
[01:13] <RobertNagy> its jsut an expirement, not sure yet if it is a good idea
[01:13] <ubitux> it sounds just like movie/amovie filter to me
[01:13] <ubitux> what's the difference?
[01:13] <RobertNagy> the difference is that you can't have both audio and video from the same file with those filters
[01:14] <RobertNagy> you would ned to open the same file twice
[01:14] <ubitux> ah yeah ok a mixed one
[01:14] <ubitux> i think it was mention of this at some point
[01:14] <RobertNagy> oh? and?
[01:15] <ubitux> i don't remember the outcome :)
[01:15] <RobertNagy> the movie/amovie filters are not very useful as is (atleast for me)
[01:15] <ubitux> with the -filter_complex though, it shouldn't be necessary anymore
[01:15] <RobertNagy> hm, explain?
[01:16] <RobertNagy> another problem with avfilter is that if you a have a movie source filter, and do a seek, there is no way to reset the following filters (as far as I know)
[01:17] <ubitux> see -filter_complex in http://ffmpeg.org/ffmpeg.html
[01:17] <RobertNagy> I don't see how that makes it unecessary?
[01:18] <ubitux> there is no need for the movie/amovie source since the streams are directly mapped
[01:18] <RobertNagy> in ffmpeg yes
[01:18] <RobertNagy> I want a generic solution that works outside of ffmpeg.c as well :)
[01:18] <ubitux> it uses the public api afaik, so it should be doable in any app
[01:19] <RobertNagy> I disagree
[01:19] <ubitux> OTOH, it might indeed be easier to just load a graph and avoid the need of opening inputs
[01:20] <ubitux> you disagree to what?
[01:20] <RobertNagy> that is doable in any app, it is complicated to reproduce the same functionality
[01:20] <ubitux> sure, complicated, but possible
[01:21] <ubitux> don't get me wrong, i think movie/amovie are useful
[01:21] <ubitux> and yes i think a common filter source would be nice
[01:22] <ubitux> if your filter work it could be nice to have it upstream
[01:22] <RobertNagy> how do you mean?
[01:23] <RobertNagy> "upstream"?
[01:23] <ubitux> in FFmpeg
[01:23] <RobertNagy> oh
[01:29] <tg2> guys, any way to encode aac with more than 1 thread?
[01:51] <Daemon404> tg2, if youre being bottlenecked by audio encoding... lol
[02:50] <tg2> yes
[02:50] <tg2> lol
[02:51] <tg2> 800fps on video, but only ~400 on audio
[02:51] <tg2> and ffmpeg -ss works like garbage with audio
[02:51] <tg2> so i split it first
[02:51] <tg2> separate it rather
[02:51] <tg2> but then the video finishes encoding twice as fast as the audio lol
[02:51] <tg2> first world problems
[02:58] <Daemon404> what are you even encoding the audio with?
[03:45] <CIA-63> ffmpeg: 03Michael Niedermayer 07master * r3b5632479e 10ffmpeg/libavcodec/snowenc.c: 
[03:45] <CIA-63> ffmpeg: snow: fix 32x32 encoding
[03:45] <CIA-63> ffmpeg: Fixes ticket1083
[03:45] <CIA-63> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[12:11] <CIA-63> ffmpeg: 03Michael Niedermayer 07master * rd04956344d 10ffmpeg/libavcodec/tiffenc.c: 
[12:11] <CIA-63> ffmpeg: tiffenc: fix out of array read
[12:11] <CIA-63> ffmpeg: Fixes ticket1112
[12:11] <CIA-63> ffmpeg: Found-by: ami_stuff
[12:11] <CIA-63> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[12:11] <CIA-63> ffmpeg: 03Michael Niedermayer 07master * r40d8c89cb6 10ffmpeg/libavcodec/flashsv2enc.c: 
[12:11] <CIA-63> ffmpeg: flashsvenc: fix image sizes below 192x192
[12:11] <CIA-63> ffmpeg: Fixes half of Ticket1105
[12:11] <CIA-63> ffmpeg: Found-by: ami_stuff
[12:11] <CIA-63> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[12:11] <CIA-63> ffmpeg: 03Michael Niedermayer 07master * r2a37ac042f 10ffmpeg/libavcodec/flashsv2enc.c: 
[12:11] <CIA-63> ffmpeg: flashsv2enc: fix division by zero
[12:11] <CIA-63> ffmpeg: Fixes other half of Ticket1105
[12:11] <CIA-63> ffmpeg: Found-by: ami_stuff
[13:12] <CIA-63> ffmpeg: 03Carl Eugen Hoyos 07master * rdb11fc5392 10ffmpeg/libavcodec/flashsv2enc.c: Fix flashsv2 encoder error message.
[17:31] <ubitux> michaelni: i think there is a problem with the #include "audio.h" at the end of lavfi/avfilter.h
[17:31] <ubitux> avfilter.h is a public header
[17:41] <kierank> tg2: profile it
[17:41] <kierank> and see what's slop
[17:41] <kierank> slow*
[17:41] <kierank> a lot of audio encoding is unoptimised
[17:44] <ubitux> michaelni: i think audio.h should be included in internal.h, removed from avfilter.h, and fix the audio filters missing the include "internal.h"
[17:44] <ubitux> or sth like that
[17:47] <ubitux> or just include audio.h in the audio filters
[18:02] <michaelni> ubitux,  including it in internal.h doesnt work as internal isnt included where its needed
[18:03] <ubitux> ok
[18:03] <ubitux> anyway i just sent a patch for this
[18:03] <ubitux> not sure if it's very sexy but it at least fix building the doc/examples
[18:04] <michaelni> patch approved
[18:05] <ubitux> thx
[18:10] <CIA-63> ffmpeg: 03Clément BSsch 07master * r9a19341e6e 10ffmpeg/doc/examples/Makefile: doc/examples: rename LDFLAGS to LDLIBS.
[18:10] <CIA-63> ffmpeg: 03Clément BSsch 07master * r5b55c7f4e9 10ffmpeg/doc/examples/Makefile: doc/examples: add -O2 in CFLAGS.
[18:10] <CIA-63> ffmpeg: 03Clément BSsch 07master * r4522df52aa 10ffmpeg/libavfilter/ (16 files): 
[18:10] <CIA-63> ffmpeg: lavfi: remove audio.h include from avfilter.h.
[18:10] <CIA-63> ffmpeg: avfilter.h is a public header and the unexported audio.h header contains
[18:10] <CIA-63> ffmpeg: only internal prototypes.
[18:10] <CIA-63> ffmpeg: 03Clément BSsch 07master * r9188294d30 10ffmpeg/doc/examples/Makefile: doc/examples: fix typo.
[18:10] <CIA-63> ffmpeg: 03Clément BSsch 07master * r2f197a20cf 10ffmpeg/doc/examples/Makefile: 
[18:10] <CIA-63> ffmpeg: doc/examples: link decoding_encoding and muxing with math lib.
[18:10] <CIA-63> ffmpeg: These two examples use the sin() function.
[18:10] <CIA-63> ffmpeg: 03Clément BSsch 07master * r072c2c08bf 10ffmpeg/doc/examples/decoding_encoding.c: 
[18:10] <CIA-63> ffmpeg: doc/examples: add missing math.h include in decoding/encoding example.
[18:10] <CIA-63> ffmpeg: The header is required for the sin() function.
[18:52] <CIA-63> ffmpeg: 03Michael Niedermayer 07master * r70f133fb54 10ffmpeg/libavcodec/mpegvideo_enc.c: 
[18:52] <CIA-63> ffmpeg: h263enc: check w/h for validity
[18:52] <CIA-63> ffmpeg: Fixes Ticket1063
[18:52] <CIA-63> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:59] <CIA-63> ffmpeg: 03Clément BSsch 07master * r8c3c7dbbb9 10ffmpeg/.gitignore: doc/examples: add programs to .gitignore.
[19:59] <CIA-63> ffmpeg: 03Clément BSsch 07master * r55e5f105d3 10ffmpeg/doc/examples/filtering_audio.c: doc/examples: use buffersrc.h instead of deprecated asrc_abuffer.h.
[19:59] <CIA-63> ffmpeg: 03Clément BSsch 07master * rbe062de9d4 10ffmpeg/doc/examples/Makefile: 
[19:59] <CIA-63> ffmpeg: doc/examples: add libswresample in the libraries.
[19:59] <CIA-63> ffmpeg: It is required for audio filtering.
[21:23] <CIA-63> ffmpeg: 03Michael Niedermayer 07master * r70e9308dec 10ffmpeg/ffmpeg.c: 
[21:23] <CIA-63> ffmpeg: ffmpeg: preserve bits_per_coded sample on stream copy.
[21:23] <CIA-63> ffmpeg: Fixes Ticket1124
[21:23] <CIA-63> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:50] <ubitux> michaelni: it looks like audio filtering is completely broken
[22:50] <ubitux> anything using amovie for instance
[22:50] <ubitux> can you confirm?
[22:51] <ubitux> (ffmpeg -f lavfi -i amovie=in.mp3 -f null -
[22:51] <ubitux> )
[22:58] <burek> ubitux, are you talking about -af ?
[23:00] <ubitux> no, audio filtering in general
[23:00] <ubitux> anything using the audio filtering is broken (ffmpeg, ffplay, api example, ...)
[23:06] <michaelni> ubitux, confirm :(
[23:08] <michaelni> why is this not tested by fate btw ?
[23:08] <burek> how is it broken? I mean audio volume up/down works.. pan works..
[23:08] <michaelni> ubitux, example segafults
[23:09] <ubitux> yes i'm working on the example and that's how i saw it
[23:09] <ubitux> i removed the poll_frame() using the new way (EAGAIN etc) but it didn't help
[23:16] <michaelni> ubitux, do you know which revission broke it ?
[23:16] <ubitux> no but it should be fairly recent
[23:35] <michaelni> ubitux, its related to extended_data
[23:37] <ubitux> ok
[23:37] <ubitux> i'm not on it atm :p
[23:37] <ubitux> (sorry but i'm playing with some users and filters on #ffmpeg)
[23:40] <michaelni> ubitux, fixed segfault with amovie locally
[23:40] <ubitux> awesome thanks :)
[23:40] <michaelni> ill test a little more and then push it
[23:40] <ubitux> can i see the patch?
[23:41] <ubitux> i can test it too
[23:41] <michaelni> if ( ----> ref->extended_data && <----  ref->extended_data != ref->data) { 
[23:42] <michaelni> in avfilter.c
[23:42] <michaelni> avfilter_ref_buffer()
[23:42] <ubitux> thx
[23:44] <ubitux> that seems better indeed :)
[23:45] <ubitux> example is fixed (with the no pulling patch at least)
[23:47] <ubitux> my (simple) tests pass too
[23:49] <michaelni> ok, ill push in a moment, thanks for testing
[23:50] <ubitux> any idea why we need this?
[23:50] <ubitux> libav doesn't have it
[23:51] <michaelni> something doesnt set extended_data i guess
[23:51] <michaelni> that can be fixed but i still think we should check for it being NULL
[23:52] <michaelni> could be changed to a av_assert0() once nothing sets it to null
[23:52] <ubitux> at least it's fixing a huge regression
[23:52] <michaelni> yes
[23:55] <CIA-63> ffmpeg: 03Michael Niedermayer 07master * rad6f006081 10ffmpeg/libavfilter/avfilter.c: 
[23:55] <CIA-63> ffmpeg: avfilter: fix regression since addition of extended_data to audio
[23:55] <CIA-63> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:57] <Daemon404> heh
[23:57] Action: Daemon404 wishes people could write better commit messages :P
[23:57] <Daemon404> cause everyone always forgets why things were done in a particular commit N months later
[23:58] <Daemon404> when i ask
[23:58] <beastd|2> Daemon404: I agree the "why" part is missing way too often
[23:59] <Daemon404> beastd|2, in this case "what" is also missing :P
[23:59] <ubitux> because it's broken obviously
[23:59] <ubitux> :)
[23:59] <Daemon404> "a regression"
[23:59] <ubitux> last thing you can do now is comment on cvslog
[00:00] --- Sun May 13 2012


More information about the Ffmpeg-devel-irc mailing list