[Ffmpeg-devel-irc] ffmpeg-devel.log.20120221

[burek]

[01:13] <bove> I'm trying to understand what determines the coefficients when converting rgb to yuv. I see some mentions in yuv2rgb.c, but I don't know how it's treated the other way
[03:32] <durandal_1707> lol michaelni MatildaHarris is still active and spamming
[03:34] <Compn> on mailing list ?
[03:34] <durandal_1707> wiki on trac
[03:35] <Compn> ah
[05:29] <CIA-17> ffmpeg: 03Panagiotis H.M. Issaris 07master * r2b3d041cdc 10ffmpeg/libavformat/hls.c: (log message trimmed)
[05:29] <CIA-17> ffmpeg: applehttp: Do seeking within segments, too
[05:29] <CIA-17> ffmpeg: Enhance seeking by demuxing until the requested timestamp is
[05:29] <CIA-17> ffmpeg: reached within the segment selected by the seek code using the
[05:29] <CIA-17> ffmpeg: playlist info.
[05:29] <CIA-17> ffmpeg: Some mpegts streams don't have dts set for all packets though,
[05:29] <CIA-17> ffmpeg: this seeking method doesn't work well for that case.
[05:29] <CIA-17> ffmpeg: 03Ronald S. Bultje 07master * rb18f8cbf3d 10ffmpeg/libswscale/ (swscale_internal.h x86/swscale_mmx.c x86/swscale_template.c): (log message trimmed)
[05:29] <CIA-17> ffmpeg: Revert two swscale commits.
[05:29] <CIA-17> ffmpeg: Revert "swscale: update context offsets after removal of AlpMmxFilter."
[05:29] <CIA-17> ffmpeg: (commit a95e3fa90b4190381b65d180eec5a4027075e2da)
[05:29] <CIA-17> ffmpeg: and
[05:29] <CIA-17> ffmpeg: Revert "swscale: Remove some write-only variables related to alpha handling."
[05:29] <CIA-17> ffmpeg: (commit 9d03cb9fc5ddf914920ab0dbe13f19a34c754966).
[05:29] <CIA-17> ffmpeg: 03Ronald S. Bultje 07master * r8fb26950ed 10ffmpeg/libavcodec/x86/h264_deblock.asm: 
[05:29] <CIA-17> ffmpeg: h264: don't use redzone in loopfilter on win64.
[05:29] <CIA-17> ffmpeg: Red zone usage is not allowed in the Win64 ABI.
[05:29] <CIA-17> ffmpeg: 03Anton Khirnov 07master * r11505f39e1 10ffmpeg/libavcodec/zmbvenc.c: zmbvenc: switch to encode2().
[05:29] <CIA-17> ffmpeg: 03Anton Khirnov 07master * r8c8c7b5e37 10ffmpeg/libavcodec/zmbvenc.c: 
[05:29] <CIA-17> ffmpeg: zmbvenc: move header writing to the end of encode_frame().
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * r0b42a9388c 10ffmpeg/ (4 files in 2 dirs): avutil: add av_rescale_q_rnd() to allow different rounding
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * re9cda85351 10ffmpeg/ (doc/APIchanges libavcodec/avcodec.h libavformat/utils.c): 
[05:30] <CIA-17> ffmpeg: avcodec: add duration field to AVCodecParserContext
[05:30] <CIA-17> ffmpeg: This will allow parsers to export the duration of the current frame being
[05:30] <CIA-17> ffmpeg: output, if known, instead of using AVCodecContext.frame_size.
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * rc7f3f1c91e 10ffmpeg/libavcodec/flac_parser.c: flac parser: set duration instead of frame_size
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * r16e54ac725 10ffmpeg/libavcodec/aac_ac3_parser.c: (e)ac3 parser: set duration instead of frame_size
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * r7575ffac8a 10ffmpeg/libavcodec/mpegaudio_parser.c: mpegaudio parser: set duration instead of frame_size
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * r2460b168b4 10ffmpeg/libavcodec/gsm_parser.c: gsm parser: set duration
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * rb3a4c7e0f1 10ffmpeg/libavcodec/mlp_parser.c: mlp parser: set duration instead of frame_size
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * r41ac9bb253 10ffmpeg/libavcodec/adx_parser.c: adx parser: set duration
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * r91a28b0e8e 10ffmpeg/libavcodec/ (internal.h libspeexenc.c libvorbis.c utils.c): avcodec: add ff_samples_to_time_base() convenience function to internal.h
[05:30] <CIA-17> ffmpeg: 03Justin Ruggles 07master * r770a5c6d02 10ffmpeg/ (libavcodec/adpcmenc.c tests/ref/acodec/adpcm_yam): (log message trimmed)
[05:30] <CIA-17> ffmpeg: adpcmenc: Use correct frame_size for Yamaha ADPCM.
[05:30] <CIA-17> ffmpeg: Output packet size should match avctx->block_align. The target output packet
[05:30] <CIA-17> ffmpeg: size is 1024 bytes.
[05:30] <CIA-17> ffmpeg: Before:
[05:30] <CIA-17> ffmpeg: mono - 1024 samples -> 512 bytes
[05:30] <CIA-17> ffmpeg: stereo - 2048 samples -> 2048 bytes
[05:30] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * readd4264ee 10ffmpeg/: (log message trimmed)
[05:30] <CIA-17> ffmpeg: Merge remote-tracking branch 'qatar/master'
[05:30] <CIA-17> ffmpeg: * qatar/master: (36 commits)
[05:30] <CIA-17> ffmpeg:  adpcmenc: Use correct frame_size for Yamaha ADPCM.
[05:30] <CIA-17> ffmpeg:  avcodec: add ff_samples_to_time_base() convenience function to internal.h
[05:30] <CIA-17> ffmpeg:  adx parser: set duration
[05:30] <CIA-17> ffmpeg:  mlp parser: set duration instead of frame_size
[09:39] <AdamWorld> Hi! I run "ffserver.exe" in CentOS with command:"./ffserver -f doc/ffserver.conf" and "ffmpeg.exe" in windows like "ffmpeg -i 1.avi http://192.168.0.128:8090/feed1.ffm" but "http://192.168.0.128:8090/feed1.ffm: Operation not permitted" occurred. What's wrong with me?
[13:25] <ubitux> michaelni: is there a reason the "audio sync method" is not part of libswresample?
[13:25] <Tjoppen> ls
[13:26] <Tjoppen> oops :)
[13:26] <ubitux> :)
[13:27] <ubitux> i guess it needs to much pts stuff&
[13:34] <Tjoppen> btw, the way ffmpeg does sync change depending on stream order. this seems wrong to me
[13:35] <Tjoppen> if I capture my webcam to stream 0 and mic to stream 1 the streams end up ~1 sec off. if I reverse the order (mic = 0, webcam = 1) then it works fine
[13:36] <ubitux> i also expected this issue, wasn't aware about that workaround :)
[13:36] <ubitux> i have a lot of a/v desync when extracting samples too
[13:37] <ubitux> (if i don't re-encode)
[13:37] <Tjoppen> youtube has similar issues
[13:37] <ubitux> isn't youtube using ffmpeg? :p
[13:37] <Tjoppen> yes, they are
[13:38] <Tjoppen> which explains the problems I had a while back with uploading emulator captures that I encoded just to not have to upload raw footage
[13:39] <Tjoppen> the emulator (MESS) doesn't do a particularly good job of its AVI output, which when passed through virtualdub caused various problems
[13:40] <ubitux> nice emulator name
[13:43] <Tjoppen> it's actually a collection of emulators. I use it for capturing and uploading videos of demoscene productions
[13:44] <Tjoppen> I should make a youtube upload guide with some tips and caveats like that sync thing
[13:46] <ubitux> or fix ffmpeg and wait a few years to get it it libav and then youtube
[13:46] <ubitux> or both :)
[13:48] <Tjoppen> I've started to nearest-point supersample every 8-bit thing I upload. gives very crisp video and audio (since for some reason youtube shits on audio bitrate at 240p)
[13:53] <Tjoppen> case in point: www.youtube.com/watch?v=CO_HcjzU4SI
[14:31] <ubitux> it seems the audio sync is not tested at all
[14:32] <ubitux> lcov seems to say nothing ever pass through the sync stuff
[14:33] <Tjoppen> is frame accurate cutting supported yet? my guess is no
[14:34] <ubitux> not afaik
[14:38] <michaelni> ubitux, moving the audio_sync_method code into swr may be interresting, i havnt thought about that ...
[14:39] <michaelni> Tjoppen, we maybe should change the async default, this should fix the stream order sync issue i think
[14:40] <michaelni> i remember considering to change the default a while ago already bit didnt do it cuz we where before a release
[14:40] <michaelni> seems i forgot about it
[14:43] <Tjoppen> michaelni: sounds good
[14:43] <ubitux> michaelni: this is actually giving me headaches while hacking do_audio_out to integrates -af :p
[15:11] <ubitux> it would be nice to add a few -async fate tests :p
[15:12] <ubitux> if the option actually work :p
[16:29] <gnafu> Hehe: <AdamWorld> [...] What's wrong with me?
[16:30] <ubitux> saste: i found the cause of the issue when changing channels, and fixed it
[16:30] <ubitux> saste: so well, -af now "works" with ffmpeg
[16:30] <ubitux> but it breaks a few other things
[16:30] <saste> ubitux: what was the problem?
[16:31] <ubitux> the problem is simply that do_audio_out uses the dec->channels to define the size to send to encode frame
[16:31] <ubitux> but dec->channels is not the number of channels which outputs the filtergraph
[16:32] <ubitux> output of*
[16:32] <ubitux> *gets out* (english beeh)
[16:33] <saste> ah funny
[16:33] <ubitux> i'm going to send the WIP, and all the issues that need to be fixed
[16:33] <saste> i wonder how many of these assumptions there are still in ffmpeg.c
[16:33] <ubitux> this is not an assumption, this is normal
[16:33] <ubitux> do_audio_out() is actually doing the right thing in this case
[16:33] <ubitux> but not when using -af
[16:34] <saste> yes i mean is correct unless you have a filtergraph between the decoder and the encoder
[16:34] <ubitux> not even -af actually, the use of audio filters
[16:34] <ubitux> yes right
[16:34] <ubitux> but using the filtergraph between decoder and encoder causes a lot of problem in do_audio_out :)
[16:35] <ubitux> so i just disabled almost everything in it
[16:35] <ubitux> (hoping the af aresample and aconvert will deal with it fine)
[16:35] <ubitux> but there is still two unsolved problem even if that works:
[16:35] <ubitux> -async
[16:35] <ubitux> and -map_channel (i can deal with this one)
[16:35] <ubitux> -async is the real issue :p
[16:36] <ubitux> also, aresample will need some rework too, iirc it requires a flush
[16:36] <ubitux> a lot of things to do :(
[16:36] <saste> yes a flush API is missing (also for video)
[16:37] <saste> the more you do the more you know there is to do
[16:37] <saste> it's normal
[16:37] <ubitux> well the issue is, the more there is to do to finish just a feature :p
[16:37] <ubitux> it's normal to get more and more things to do, but when they're all blocking one single issue, it's a nightmare :D
[16:38] <ubitux> all of this is just to get the merge in -map_channel ffs! :D
[16:38] <saste> eheh... that's because we're pushing to the extreme the design of ffmpeg.c
[16:39] <ubitux> i'd like to auto insert the volume filter too :-°
[16:46] <CIA-17> ffmpeg: 03Clément BSsch 07master * r241f8465d0 10ffmpeg/libavfilter/asrc_abuffer.c: (log message trimmed)
[16:46] <CIA-17> ffmpeg: lavfi/abuffer: init the data planes with NULL pointers.
[16:46] <CIA-17> ffmpeg: Samples buffer ref is allocated and loaded with the uninitialized data
[16:46] <CIA-17> ffmpeg: pointers:
[16:46] <CIA-17> ffmpeg:  av_asrc_buffer_add_buffer()
[16:46] <CIA-17> ffmpeg:  -> av_asrc_buffer_add_samples()
[16:46] <CIA-17> ffmpeg:  -> avfilter_get_audio_buffer_ref_from_arrays(data, ...)
[16:46] <CIA-17> ffmpeg: 03Clément BSsch 07master * raecf0cf5ed 10ffmpeg/doc/examples/ (Makefile filtering_audio.c): 
[16:46] <CIA-17> ffmpeg: doc/examples: add audio decoding/filtering example.
[16:46] <CIA-17> ffmpeg: Mostly based on doc/examples/filtering.c. lavfi API is still limited to
[16:46] <CIA-17> ffmpeg: "buffer feeding" instead of "frame feeding" at the moment, so this
[16:46] <CIA-17> ffmpeg: example code sticks with it.
[17:34] Action: Daemon404 pokes kierank 
[20:35] <Freakshow> what's a supported format for -loop
[20:39] <Freakshow> nm... I think I found out
[22:39] <kriegerod> Is this a sane idea to implement video filter that gives output pictures with given framerate? So it can force constant framerate from vfr input, or  just change fps.
[22:47] <CIA-17> ffmpeg: 03Clément BSsch 07master * r2bdac29360 10ffmpeg/doc/examples/ (Makefile filtering.c): doc/examples: rename filtering.c into filtering_video.c.
[22:48] <overflow_0f8b> hi
[23:06] <burek> kriegerod, do you mean ffmpeg -i input -vcodec ... -acodec ... -r X -f image2 out%02d.png ?
[23:07] <kriegerod> No, i mean avfilter
[23:07] <iive> i think he means something like filter that enforces constant framerate
[23:08] <kriegerod> That allows to transform, e.g., vfr flv video to constant 25fps
[23:09] <kriegerod> Avfilter will be handy for apps
[23:09] <kriegerod> I need some mentorship to implement it
[23:59] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r68fac5c2b8 10ffmpeg/libavutil/avutil.h: 
[23:59] <CIA-17> ffmpeg: doxy: Disable the main index page.
[23:59] <CIA-17> ffmpeg: Theres no usefull or even remotely complete information on it currently.
[23:59] <CIA-17> ffmpeg: Which just leads to confusion.
[23:59] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:59] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * re6380afa6a 10ffmpeg/libavformat/avidec.c: 
[23:59] <CIA-17> ffmpeg: avidec: Reduce log level for out of index error message.
[23:59] <CIA-17> ffmpeg: Its otherwise spaming every time one tries to seek to outside
[23:59] <CIA-17> ffmpeg: the file.
[23:59] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:59] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r92b5f71a7d 10ffmpeg/libavformat/utils.c: 
[23:59] <CIA-17> ffmpeg: lavf: Reimplement new seek API emulation
[23:59] <CIA-17> ffmpeg: This fixes seeking to before and after files with ffplay.
[23:59] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Wed Feb 22 2012