[Ffmpeg-devel-irc] ffmpeg-devel.log.20120403

[burek]

[01:23] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * rcfd64eec4e 10ffmpeg/: 
[01:23] <CIA-17> ffmpeg: Merge remote-tracking branch 'qatar/master'
[01:23] <CIA-17> ffmpeg: * qatar/master:
[01:23] <CIA-17> ffmpeg:  FATE: Add RALF decoding test
[01:23] <CIA-17> ffmpeg: Conflicts:
[01:23] <CIA-17> ffmpeg:  tests/fate/real.mak
[01:23] <CIA-17> ffmpeg: Merged-by: Michael Niedermayer <michaelni at gmx.at>
[01:23] <CIA-17> ffmpeg: 03Derek Buitenhuis 07master * r0aaa45ef69 10ffmpeg/tests/ (fate/real.mak ref/fate/ralf): 
[01:23] <CIA-17> ffmpeg: FATE: Add RALF decoding test
[01:23] <CIA-17> ffmpeg: Signed-off-by: Derek Buitenhuis <derek.buitenhuis at gmail.com>
[07:05] <ramiro> hi
[07:07] <ramiro> michaelni: asf.c:ff_asf_simple_index_header is duplicated in asfdec.c:index_guid
[07:15] <Compn> i have dumb question... i need to access an ipad app but dont have an ipad. can i run ipad os in virtualbox or something ?
[07:19] <ramiro> Compn: if you have the source code you can try to run it on Xcode, you still need some kind of mac for that
[07:22] <ubitux> btw, added a fate instance with THREADS=8
[07:22] <ubitux> morning ppl.
[07:24] <ramiro> do you also test fate on this system?: http://dmitry.co/index.php?p=./04.Thoughts/07.%20Linux%20on%208bit
[07:25] <ubitux> :)
[07:31] <Compn> ramiro : theres no ipad vbox ?
[07:31] <ramiro> Compn: vbox doesn't emulate arm
[07:37] <Daemon404> qemu oy
[07:37] <Daemon404> yo*
[07:38] <Daemon404> but apple puts stuf in place to stop virt
[07:38] <Daemon404> tl;dr gl;hf
[07:50] <Compn> ah
[07:50] <Compn> crap
[09:48] <michaelni> ramiro, thx, fixed locally
[09:52] <ramiro> michaelni: thanks
[09:53] <ramiro> michaelni: do you know anything about asf created by ffmpeg being unseekable by wmp? (it takes seconds to start playing again)
[09:53] <ramiro> whereas asf created by windows can seek to pretty much any second
[09:56] <ubitux> Anton is doing a good work on lavfi afaict; we will need to rework a few things though :p
[10:00] <michaelni> ramiro, i dont remember such asf issue, no.
[10:00] <michaelni> and i have no windows box so i cant test ...
[10:00] <ramiro> ok...
[10:04] <av500> asf files created by lavf have issues
[10:05] <av500> some play without sound here on my XP machine
[10:05] <av500> I have not yet found the pattern, its not all of them
[11:03] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * rb61e596a14 10ffmpeg/libavformat/asfdec.c: 
[11:03] <CIA-17> ffmpeg: asf: remove duplicated index_guid
[11:03] <CIA-17> ffmpeg: Found-by:ramiro
[11:03] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[11:03] <CIA-17> ffmpeg: 03Tomas Härdin 07master * r24e90d0c05 10ffmpeg/libavformat/mxfenc.c: 
[11:03] <CIA-17> ffmpeg: mxfenc: Don't allow muxing audio-only since it's not supported
[11:03] <CIA-17> ffmpeg: This fixes muxing audio-only output resulting in SIGFPE due to lack of EditRate.
[11:03] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[11:22] <ubitux> i wonder how i'm supposed to make the seek work with a swf compressed file
[11:22] <ubitux> i see two solutions:
[11:23] <ubitux> - allocate the whole decompressed file in memory
[11:23] <av500> 0) install flash :)
[11:23] <ubitux> - redecode everything from the beginning
[11:23] <ubitux> (each time we request a "smart" seek)
[11:23] <ubitux> seek is needed in case of -loop for instance
[11:23] <ubitux> which is recurrent within a flash file
[11:24] <av500> bot -loop is a seek back to 0
[11:24] <av500> but
[11:24] <av500> no?
[11:24] <ubitux> what if you loop the video but not the sound (i mean if they loop async)
[11:24] <ubitux> ?
[11:24] <av500> no idea :)
[11:25] <ubitux> there is a third ugly solution: 2 pass, first pass outputs to a /tmp/file
[11:25] <ubitux> and 2nd pass plays it :p
[11:26] <ubitux> i wonder what flash does&
[11:27] <ubitux> or even swfdec/gnash/lightspark
[12:42] <vivienschilis> anybody there to help me with the segment format?
[13:27] <ubitux> another user trying to fetch sources from the svn
[13:27] <ubitux> maybe we should disable it
[13:29] <burek> why is it up if it's not being used anyway :)
[13:29] <burek> it's just a security hole :)
[13:32] <michaelni> burek, mplayer still uses svn
[13:35] <ubitux> we should at least remove the svn line in the webpage
[13:39] <mrec> Hi, I guess no-one knows in #ffmpeg, av_register_protocol2 got removed, however it seems like I cannot find the replacement for it, should it be ffurl_register_protocol?
[13:39] <ubitux> afaik no alternative is available
[13:40] <mrec> I wonder why did it get removed then?
[13:40] <ubitux> i'm asking the author of the drop
[13:57] <nevcairiel> The answer will probably be "because"
[13:58] <nevcairiel> Someone also decided that supporting external demuxers was apparently bad and started hiding all sorts of symbols needed for it
[14:02] <mrec> okay what I want to do is simply to write the decoder output to memory
[14:02] <mrec> and read the input from memory
[14:05] <wbs> mrec: using a custom aviocontext is usually easier and more straightforward than using a custom urlprotocol
[14:09] <ubitux> the issue is discussed here: http://bugzilla.libav.org/show_bug.cgi?id=224
[14:09] <ubitux> mrec: if you don't agree with this, it can be discussed on ffmpeg dev mailing-list
[14:13] <ubitux> mrec: what's your use-case?
[14:21] <iive> mrec: I think you should talk with reimar (send him email). He have to solve a similar problem related to the URL protocol, when used in mplayer.
[14:34] <mrec> ubitux: I'm encoding analogTV frames + pcm audio
[14:34] <mrec> not using video4linux2
[14:34] <mrec> the input is read via usbfs on application level
[14:34] <ubitux> can't you make use of the avio api?
[14:37] <mrec> need to check that one
[15:25] <mrec> ubitux: I don't quite understand how this can be done with the avio API, do you have any example?
[15:26] <mrec> the problem is I would like to write the encoded output into my own buffer
[15:26] <mrec> not into a file
[15:27] <nevcairiel> with avio you can implement the read and write functions the way you want them
[15:27] <ubitux> i'd say you avio_alloc_context() to redefine the read/write callback
[17:23] <zimbatm> hey guys
[17:23] <zimbatm> any input on that ticket ? : https://ffmpeg.org/trac/ffmpeg/ticket/1164
[17:23] <zimbatm> it's not urgent but i'm curious to understand or if i can do something more to help
[17:26] <gnafu> j-b: http://en.wikipedia.org/wiki/Jean-Baptiste_Charcot ?
[17:26] <j-b> nope
[17:27] <j-b> gnafu: http://www.lepoint.fr/chroniqueurs-du-point/guerric-poncet/exclusif-vlc-saisit-la-hadopi-pour-ouvrir-les-verrous-des-blu-ray-03-04-2012-1448011_506.php
[17:27] <j-b> gnafu: the first version of the article was wrong :)
[17:27] <j-b> and named me Charcot
[17:28] <gnafu> j-b: Aah, hehe.
[17:28] <gnafu> Hehe, from Google's translation: "VideoLAN users are eager to read all the Blu-ray."
[17:29] <gnafu> "[READ] ALL THE THINGS."
[18:03] <ubitux> should we add fate instances with specific cpu flags?
[18:04] <Compn> zimbatm : pasting the full uncut ffmpeg output in the bugreport will help. as will using git ffmpeg
[18:16] <zimbatm> Compn: thanks working on that
[18:27] <zimbatm> Compn: thanks, actually it works with the latest x264 and ffmpeg :/
[18:27] <zimbatm> sorry for the noise
[18:44] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r79a1d98676 10ffmpeg/libavfilter/vf_lut.c: 
[18:44] <CIA-17> ffmpeg: vf_lut: optimize planar case
[18:44] <CIA-17> ffmpeg: 2500 -> 1900 cpu cycles
[18:44] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:44] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r002d9ded4a 10ffmpeg/libavfilter/vf_lut.c: 
[18:44] <CIA-17> ffmpeg: vf_lut: optimize rgb codepath
[18:44] <CIA-17> ffmpeg: 19k -> 15k cpu cycles
[18:44] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:44] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r38477e1981 10ffmpeg/libavfilter/vf_lut.c: 
[18:44] <CIA-17> ffmpeg: vf_lut: avoid rgb component indexing in the inner loop
[18:44] <CIA-17> ffmpeg: 15k->10k cpu cycles
[18:44] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:47] <j-b> curious, what is vf_lut?
[18:52] <nevcairiel> it can compute a lookup table based on some expression syntax to map the input image to other values, i imagine it can be used for gamma or contrast adjustments of sorts
[18:53] <j-b> ok
[19:24] <CIA-17> ffmpeg: 03Clément BSsch 07master * r588eaa106d 10ffmpeg/libavformat/ (swf.h swfdec.c): swfdec: support compressed swf.
[20:48] <ramiro> hi
[20:49] <ramiro> what's the fastest trick to do this calculation: set x to be at least the value of y, but in multiples of BLOCK_SIZE
[20:57] <nefrir> is BLOCK_SIZE a power of 2 ?
[21:01] <nefrir> if yes -> x = (y + BLOCK_SIZE - 1) & ~(BLOCK_SIZE - 1)
[21:01] <nefrir> otherwise, you will need a modulo
[21:01] <ramiro> hm, it's randomly defined to 600, I can randomly define it to 1<<9
[21:01] <ramiro> thanks
[22:07] <ubitux> michaelni: 'got a strange behaviour with a sample in av_packet_split_side_data()
[22:08] <ubitux> it seems there is a use of uninit value
[22:08] <ubitux> wget http://z0r.de/L/z0r-de_1009.swf && valgrind ./ffprobe_g *.swf
[22:12] <ubitux> if you have no idea at first look i'll open an issue
[22:12] <ubitux> but that code looks like it was written by you so& :)
[23:55] <michaelni> ubitux, from a 2 min look id say something stores uninitialized data in the packet or at least valgrind thinks so
[23:56] <ubitux> i had weird behaviour (i got a frame output at some point)
[23:56] <ubitux> but couldn't reproduce
[23:56] <ubitux> so i guess the uninit is real
[00:00] --- Wed Apr  4 2012