[Ffmpeg-devel-irc] ffmpeg-devel.log.20121223

[burek]

[00:11] <saste> michaelni, resampler options should be documented in ffmpeg-resampler, not in filters
[00:19] <michaelni> saste, what with aresample ?
[00:19] <michaelni> i mean docs
[00:19] <michaelni> they are the same as swr
[00:20] <saste> michaelni, indeed we should avoid duplication
[00:20] <saste> and if we have to document in a place, we should do it in the library
[00:20] <saste> and point to the resampler manual
[00:21] <michaelni> that works but its not that nice to the user
[00:21] <michaelni> asyncts -> see aresample, aresample see swr
[00:21] <saste> michaelni, same problem with scale
[00:21] <saste> but we're not duplicating the scale manual in the filter
[00:22] <saste> it's not optimal but is the best we can get
[00:22] <michaelni> i think at least async should be mentioned in aresample docs
[00:22] <saste> as an example, sure
[00:22] <saste> but i don't want to maintain duplicate docs
[00:23] <cone-495> ffmpeg.git 03Piotr Bandurski 07master:2e6bd22bbf09: rmdec: set bit rate for ra4
[00:25] <michaelni> would be nice if one could just include a section or table from another texi
[00:26] <michaelni> anyway ill move the async addition to the resample docs, any other comments on the patchset ?
[00:28] <saste> uh i believed it was already committed
[00:28] <saste> michaelni, we can use @include
[00:29] <saste> but again i'd prefer to avoid inclusions when unnecessary, since they result in unwieldly manuals
[00:31] <saste> michaelni, i'll review it tomorrow since i'm felling asleep
[00:32] <michaelni> saste, ok no hurry
[03:59] <Abhijit> hello.
[03:59] <Abhijit> I need help understanding how ffmpeg uses vdpau and vaapi?
[03:59] <Abhijit> i need to enable libavg to use vaapi through ffmeg
[03:59] <Abhijit> help please
[04:00] <Abhijit> where can I start?
[04:15] <michaelni> Abhijit, look at applications that use ffmpeg with VDPAU / VAAPI
[04:15] <Abhijit> michaelni, libavg uses vdpau through ffmpeg.
[04:21] <cone-495> ffmpeg.git 03James Darnley 07master:707af43d19e8: lavf/y4m: spellcheck
[04:21] <cone-495> ffmpeg.git 03Michael Niedermayer 07master:ca9eb9305aa2: mpegvideo_enc: fix edge emulation of dimension%16 != 0 for YUV != 420
[04:25] <michaelni> Abhijit, if looking at the source & applications doesnt awnser your questions, look in MAINTAINERS and ask the person listed there (gwenole for VAAPI) but i think he is busy recently
[04:26] <Abhijit> ok
[04:38] <ubitux> michaelni: i get something strange here; it seems pp is (currently) leaking
[04:38] <ubitux> valgrind does show it, but doesn't error out
[04:38] <ubitux> adding --show-reachable=yes displays it clearly
[04:38] <ubitux> but --error-exitcode=1 still doesn't have any effect
[04:41] <ubitux> michaelni, this is fate-lavfi-pp2: http://pastie.org/private/nsi5j2iz7rc1dedvi2gg#3
[04:45] <ubitux> btw, the dering code seems to run some invalid read&
[04:45] <ubitux> (try just pp=fa)
[04:46] <michaelni> ill fix the leak
[04:48] <ubitux> well the leak looks relatively harmless
[04:48] <ubitux> (at least valgrind doesn't worry about it)
[05:41] <cone-495> ffmpeg.git 03Michael Niedermayer 07master:f9d8658d67cf: libpostproc: fix leak of 2kb
[13:49] <JEEB> requesting merging of 68a04b0ccee66f57516e129dd3ec457fd50b4bec from libav to fix ALAC decoding with quite a few QT-created samples, this was borked in both libav and ffmpeg HEAD (one of the samples that got fixed http://www.cccp-project.net/beta/test_files/libavcodec_failing_at_alac.m4a )
[13:50] <wm4> JEEB: wouldn't you expect this to be merged on the next full merge anyway?
[13:51] <JEEB> yeah... just didn't see it for a while so I decided to /r/ it manually
[13:51] <cone-367> ffmpeg.git 03Justin Ruggles 07master:68a04b0ccee6: alacdec: do not be too strict about the extradata size
[13:51] <cone-367> ffmpeg.git 03Justin Ruggles 07master:5778299c7ed3: ttadec: fix last frame handling when seeking
[13:51] <cone-367> ffmpeg.git 03Justin Ruggles 07master:d02202e08a99: opt: avoid segfault in av_opt_next() if the class does not have an option list
[13:51] <cone-367> ffmpeg.git 03Anton Khirnov 07master:418693bdc5ed: lavc: rewrite and extend AVFrame doxy
[13:51] <cone-367> ffmpeg.git 03Michael Niedermayer 07master:2f9903f66220: Merge commit '418693bdc5edad72b0d055f741931612d797eb05'
[13:51] <wm4> lol
[13:52] <JEEB> now that's nice :3
[13:52] <JEEB> nevcairiel, feel free to update your ffmpeg to fix ALAC decoding for a lot of stuff (VLC reportedly had a lot of herp and derp with these)
[14:59] <cone-367> ffmpeg.git 03Anton Khirnov 07master:6e5c44b04383: lavc: make pkt_{p,d}ts doxy less confusing
[14:59] <cone-367> ffmpeg.git 03Anton Khirnov 07master:2ca5ca295760: huffyuv: split encoder and decoder into separate files.
[14:59] <cone-367> ffmpeg.git 03Anton Khirnov 07master:b047c68783aa: 4xm: don't rely on get_buffer() initializing the frame.
[14:59] <cone-367> ffmpeg.git 03Michael Niedermayer 07master:def18e5470bd: Merge commit 'b047c68783aa4042b322af7af043b643d5daf09c'
[14:59] <ubitux> "Idk what top posting is"
[14:59] <ubitux> :/
[14:59] <ubitux> why u no google it :(
[15:10] <cone-367> ffmpeg.git 03Anton Khirnov 07master:bdfa24514e06: aura: cosmetics, reformat
[15:11] <cone-367> ffmpeg.git 03Anton Khirnov 07master:c3a76c3decc7: asvdec: cosmetics, reformat
[15:11] <cone-367> ffmpeg.git 03Anton Khirnov 07master:808b2227d748: ansi: remove an extra return
[15:11] <cone-367> ffmpeg.git 03Anton Khirnov 07master:c64cf2a37bfc: aasc: cosmetics, reformat
[15:11] <cone-367> ffmpeg.git 03Anton Khirnov 07master:b6d7d4efae60: 8bps: cosmetics
[15:11] <cone-367> ffmpeg.git 03Michael Niedermayer 07master:8e09e183fc5f: Merge remote-tracking branch 'qatar/master'
[16:15] <cone-367> ffmpeg.git 03Piotr Bandurski 07master:76f126ba2e38: rmdec: set bit rate for ra3
[16:15] <cone-367> ffmpeg.git 03Piotr Bandurski 07master:9bbfcc267540: rmenc: write correct bytes per minute
[17:04] <burek> how can i properly forward captured video from usb webcam + audio to vlc for further streaming..?
[17:04] <burek> what output format would be the most suitable between ffmpeg and vlc
[17:14] <ubitux> burek: ffmpeg should be able to send anything to vlc; i guess it depends mostly on what vlc can take in input
[17:16] <burek> i tried using the "most obvious way" (to me) to output to -f mpegts udp://...
[17:17] <burek> but i get a lot of errors: DTS 1356279396128793, next:1100000 st:0 invalid dropping
[17:26] <Compn> burek : why not just use vlc to input and output ?
[17:26] <Compn> it does both
[17:26] <Compn> v4l dshow etc
[17:27] <Compn> or use ffserver for both...
[17:27] <Compn> either way
[17:27] <burek> well...
[17:27] <burek> it stops capturing from the webcam
[17:27] <Compn> why people try to chain two different programs using pipes or network hacks :\
[17:27] <burek> without giving me any clue as to why
[17:27] <burek> also
[17:27] <burek> i get a millions of "late buffer for mux input"
[17:27] <burek> no matter how i capture the video
[17:27] <burek> even without the audio
[17:28] <burek> cpu is never over 30%
[17:28] <Compn> git version of vlc ?
[17:28] <burek> and they never replied to anyone who ever asked what does that error mean
[17:28] <burek> so it's pretty obvious why i have to do this gymnastics :)
[17:28] <burek> yes
[17:28] <Compn> it means the buffer from the capture card is late
[17:29] <Compn> the muxer needs data ... its not important probably :D
[17:29] <Compn> what bout ffserver ?
[17:29] <burek> if it's not important, why does it clog my log then :)
[17:29] <burek> and why does it NEVER work
[17:29] <burek> i mean, it always throws that warning
[17:29] <burek> no matter how i capture / save-stream video
[17:30] <burek> anyway, i'll have to get back to ffserver i guess
[17:30] <burek> the problem was it was segfaulting too much
[17:30] <burek> and got me annoyed :)
[18:12] <ubitux> saste: why -help filter= doesn't work?
[18:31] <saste> burek, i discovered what the problem was with the ipod
[18:35] <michaelni> saste, what help text would u suggest for async ?
[18:35] <j-b>  1 ??????,?? 125 ?(+),?? 7 ?(-)
[18:35] <j-b> oops
[18:35] <j-b> sorry :)
[18:36] <cone-367> ffmpeg.git 03Michael Niedermayer 07master:85e83e7a5813: doc/ffmpeg-resampler: mention that the options also work with the aresample filter
[18:55] <burek> saste cool :) something specific to ipod, i guess? :)
[18:56] <saste> burek, it was the usual yuv444p silly thing
[18:56] <saste> had to force the format to yuv420p and worked smoothly
[18:56] <burek> oh :)
[18:57] <burek> good :)
[18:57] <burek> btw is it possible to port ffmpeg to android somehow?
[19:06] <cone-367> ffmpeg.git 03Clément BSsch 07master:e36ba9884f26: fate/mp=pp: expand subfilters with dr (dering) in them.
[19:06] <cone-367> ffmpeg.git 03Clément BSsch 07master:72380d15ae8d: fate/mp=pp: remove dering subfilter test.
[19:06] <cone-367> ffmpeg.git 03Clément BSsch 07master:9b8de930366d: lavfi: add pp filter.
[19:06] <cone-367> ffmpeg.git 03Clément BSsch 07master:768c47ec3713: lavfi: remove mp=pp filter.
[19:28] <cone-367> ffmpeg.git 03Michael Niedermayer 07master:3e1577533318: x86/ac3dsp_init: try to workaround ICC failure.
[19:28] <ubitux> saste: i have no idea how i could achieve the representation you are proposing :(
[19:30] <saste> ubitux, no problem
[19:30] <saste> but you may simplify colors, add text annotations, and move input pads to the left
[19:30] <saste> it is already a nice improvement
[19:33] <ubitux> the input pads are declared first, i think graphviz is moving them to "the best place"
[19:33] <ubitux> maybe i can change that&
[19:33] <ubitux> about the color, i like having the filters src, normal and sink filters in a different color
[19:34] <ubitux> it helps looking for all teh input and output of a graph
[19:34] <ubitux> (especially when they are moved all around the graph)
[19:35] <ubitux> i'll try to improve all of this
[19:48] <cone-367> ffmpeg.git 03Michael Niedermayer 07master:d33c939bac31: docs: add aresample/async example
[19:50] <mfg> ubitux I assume that was your review on the mailing list. I've responded if you want to have a look. Also, thanks for looking!
[19:56] <ubitux> sounds like you ran into another options scope issue
[19:57] Action: ubitux looks deeply in the eyes of saste 
[19:58] <mfg> I guess you could call it that
[20:04] <ubitux> haha michaelni 
[20:13] <mfg> ubitux I gtg, but I'd like to get this commited, so I'll be watching my email! (and I'll be back here soon enough)
[20:14] <ubitux> ok, will try to find some time to understand the issue well
[20:14] <ubitux> though, i'm really not the best for this task :(
[20:16] <mfg> ok, can we call in someone who is?
[20:17] <ubitux> he's running away regularly!
[20:17] <mfg> ?
[20:18] <ubitux> also you might have notice we kind of lack manpower
[20:18] <mfg> yes, as do all OSS projects :(
[20:18] <ubitux> mfg: 20:03:36 ::: saste [~saste___ at dynamic-adsl-78-15-171-139.clienti.tiscali.it] has quit [Ping timeout: 250 seconds]
[20:18] <ubitux> :(
[20:18] <mfg> oh, ok.
[20:18] <ubitux> anyway, i'll try to have a look
[20:18] <mfg> ok, TY -- anything I can do I will
[20:18] <mfg> bbl
[20:18] <ubitux> generally speaking, if you see inactivity after 3 days
[20:18] <ubitux> feel free to ping the thread
[20:18] <mfg> ok!
[20:25] <kierank> Compn: interesting stuff on the ml
[20:25] <cone-367> ffmpeg.git 03Clément BSsch 07master:57e99ea44d37: doc/ffserver: fix a few typo.
[20:28] <Compn> which ml ?
[20:29] <Compn> oh , you keep talking to the person
[20:30] <Compn> who cant understand a license, but wants to write a decoder ...
[20:43] <ubitux> Compn: you should paste him a link about top posting first
[20:44] <ubitux> seems you're skipping to much concepts at once
[20:44] <ubitux> too*
[20:48] <Compn> what
[20:48] <Compn> ubitux : that guy is the devel of perian
[20:48] <Compn> er
[20:49] <Compn> no, someone that tried to fork perian 
[20:51] <Compn> i'm internet stalker for sure
[20:51] <ubitux> :)
[21:01] <Compn> kierank : oops, did i mix up dts and dolby?
[21:01] <kierank> yes
[21:30] <durandal_1707> ubitux: ping
[21:30] <ubitux> durandal_1707: pong
[21:31] <durandal_1707> ubitux: i have unfinished histogram filter but i fail to get it to work properly with overlay
[21:32] <durandal_1707> ubitux: you can find branch on my github fork
[21:32] <ubitux> does it work with another filter?
[21:33] <durandal_1707> what another filter?
[21:33] <ubitux> ffplay randomvideo.mp4 -vf 'split[b], pad=iw:ih*2[src], [b]edgedetect, [src]overlay=0:h'
[21:34] <ubitux> if this work, does it still work with your filter?
[21:34] <durandal_1707> my filter pass frame unchanged and optionaly creates extra pad to render histogram
[21:36] <durandal_1707> so it creates video out0 and out1
[21:36] <ubitux> building it
[21:36] <ubitux> 'going to have a look
[21:36] <durandal_1707> so simple overlay should still work
[21:37] <durandal_1707> but i also get buffer queque overflow
[21:37] <durandal_1707> if i switch outputs for overlay i get no window but same message
[21:37] <durandal_1707> *s, it is spam
[21:40] <ubitux> your histogram doesn't show up here
[21:41] <durandal_1707> you use overlay?
[21:42] <ubitux> nope
[21:42] <ubitux> ./ffplay foobar.avi -vf 'histogram[a][out],[a]nullsink'
[21:42] <ubitux> just doing this
[21:42] <ubitux> pts are broken
[21:43] <durandal_1707> i get black frame here
[21:43] <durandal_1707> and it eats cpu...
[21:44] <ubitux> yes blackframe too
[21:44] <durandal_1707> and how to fix pts?
[21:45] <ubitux> you could copy it from the input
[21:45] <ubitux> any reason you are duplicating the stream btw?
[21:45] <ubitux> i mean users can just use split
[21:46] <ubitux> split, histogram, overlay sounds like a simpler way
[21:46] <durandal_1707> so i remove null_get_buffer ?
[21:47] <durandal_1707> actually creating only one pad
[21:47] <ubitux> yeah i guess, and you just unref the input frame when you generated your histogram one
[21:48] <ubitux> make your filter simple
[21:48] <ubitux> i see you are doing something similar to show{spectrum,waves} and ebur128, but they are a mix of audio and video
[21:48] <ubitux> so it's a special case
[21:48] <ubitux> IMO in the histogram case, just make one output stream with the histogram
[21:49] <ubitux> the split will be pretty trivial in the filtergraph
[21:49] <ubitux> and the filter will be more reliable
[21:49] <ubitux> and simpler :p
[21:51] <durandal_1707> so in this case i will not need to set pts?
[21:52] <ubitux> i think you'll still need to copy a few props from the inpicref
[21:52] <ubitux> look at what avfilter_copy_buffer_ref_props() does
[21:53] <ubitux> or maybe vf scale
[21:53] <ubitux> vf scale is doing a avfilter_copy_buffer_ref_props and then re-setting the proper settings for the output
[21:53] <ubitux> afaiu
[21:57] <ubitux> saste is going to enjoy latest luca changes to hlsenc
[21:58] <ubitux> let's fix the off by one but mmh& differently
[21:58] <durandal_1707> hmm i get assert now when i use rgb24
[21:59] <durandal_1707> i set different pix ftm for input and output
[21:59] <ubitux> mmh isn't vf scale handling such case?
[22:00] <durandal_1707> i use drawutils for output
[22:01] <durandal_1707> are you saying that formats should be same?
[22:02] <ubitux> wtf is this broken av_basename
[22:02] <ubitux> durandal_1707: i don't think so
[22:02] <ubitux> durandal_1707: vf scale is handling different formats right?
[22:02] <durandal_1707> i dont do conversion
[22:03] <durandal_1707> i read one and output always rgb24 for example
[22:03] <ubitux> yes but in vf scale you have a different format between in and out
[22:03] <ubitux> so the problem should be similar
[22:10] <durandal_1707> ubitux: if graph output is disabled should i still output video?
[22:20] <cone-367> ffmpeg.git 03Paul B Mahol 07master:98fed176cf28: alsdec: improve warning message when invalid channel position is found
[22:26] <cone-367> ffmpeg.git 03Paul B Mahol 07master:4f5440adc2aa: NIST SPHERE demuxer
[22:28] <ubitux> durandal_1707: why should it be disabled?
[22:28] <ubitux> otherwise sure you can make it a sink filter
[22:28] <ubitux> < cone-367> ffmpeg.git Paul B Mahol master:4f5440adc2aa: NIST SPHERE demuxer // ffmpeg can now encode planetes \o/
[23:36] <saste> michaelni, your ascii art... LOL
[00:00] --- Mon Dec 24 2012