[FFmpeg-devel-irc] IRC log for 2010-08-12

Fri Aug 13 02:00:44 CEST 2010

[00:05:32] <Dark_Shikari> someone broke the build
[00:05:45] <Dark_Shikari> lol
[00:05:49] <Dark_Shikari> BBB forgot mms.h
[00:06:10] <pasteeater> it's 24779
[00:06:57] <Dark_Shikari> I know
[00:34:05] <CIA-93> ffmpeg: bcoudurier * r24781 /trunk/libavfilter/vf_scale.c: 100L, fix vf_scale, since copy_ref_props now copy w and h, we must update them
[00:36:08] <CIA-93> ffmpeg: bcoudurier * r24782 /trunk/libavformat/utils.c: Fix segv when stream copy and validate_tag fails, st->codec->codec is not set
[00:47:31] <Dark_Shikari> Yuvi: any idea why vp8 rac init gets 16 bits instead of 8?
[00:58:24] <Dark_Shikari> Is there a number in ffmpeg that is 4 when the system is 32-bit, and 8 when it's 64-bit?
[00:58:30] <Dark_Shikari> Or some similar way of knowing if we're 64-bit
[00:58:38] <Dark_Shikari> or an ifdef
[00:58:50] <roxfan> sizeof(int)?
[00:58:53] <Dark_Shikari> wrong
[00:59:01] <Dark_Shikari> sizeof(void*) would be more correct
[00:59:05] <Dark_Shikari> but I assume ffmpeg already has one
[01:12:23] <CIA-93> ffmpeg: darkshikari * r24783 /trunk/libavcodec/ (x86/vp56_arith.h vp56.h vp56rac.c):
[01:12:23] <CIA-93> ffmpeg: VP5/6/8: ~7% faster arithmetic decoding
[01:12:23] <CIA-93> ffmpeg: Grab from the bitstream in 16-bit chunks instead of 32-bit chunks.
[01:12:23] <CIA-93> ffmpeg: TODO: grab in 32-bit chunks on 64-bit systems.
[01:18:56] <alexkay> does anyone else see compile errors on svn trunk? http://pastebin.com/yMM1EsZX
[01:18:58] <Dark_Shikari> er, oops.  I screwed up the commit message.
[01:19:03] <Dark_Shikari> alexkay: yes
[01:25:34] <alexkay> Dark_Shikari: thanks, should have read the backlog
[01:46:19] <Dark_Shikari> wtf
[01:46:25] <Dark_Shikari> why does be24 reading suck so much on x86
[01:50:07] <j0sh> vp8 is a scalable codec?!?
[01:51:07] <Dark_Shikari> yes, it synergizes paradigms to improve the bottom line
[01:51:48] <j0sh> lol
[01:52:07] <j0sh> is it scalable as in svc-bitrate-peeling?
[01:52:12] <Dark_Shikari> no
[02:02:01] <BBB> "oops"
[02:02:27] <CIA-93> ffmpeg: rbultje * r24784 /trunk/libavformat/mms.h: Add another missing file from r24799.
[02:03:45] <j0sh> gracias, been waiting for that :)
[02:06:08] <BBB> I apologize ...
[02:27:52] <BBB> hi spyfeng
[02:28:00] * BBB about to go offline
[07:28:32] <Tjoppen> god morgon
[07:28:50] <spaam> god morgon
[07:31:27] <kshishkov> goda morgnar
[07:32:40] <Tjoppen> the ffmpeg doesn't work!
[07:33:08] <lu_zero> moin
[07:33:28] <kshishkov> Tjoppen: that'a blatant lie!
[07:34:25] <Tjoppen> which version the FFMPEG doesn't not support?
[07:35:13] <kshishkov> your sentence was found to be invalid and was discarded
[07:52:54] <pJok> hehe
[07:52:58] <pJok> god morgon, kshishkov
[07:53:31] <KotH> moikka moi
[07:54:17] <pJok> servus!
[07:55:55] <kshishkov> okay, while this channel is mostly Swedish, can someone tell me how to order, say, sixpack of Trocadero?
[07:57:34] <pJok> en sexpack Trocadero tack
[07:57:36] <pJok> iirc
[07:58:06] <kshishkov> most people here won't understand that :(
[07:58:43] * kshishkov looks at http://vasterbottensost.com/webbshop/produkter/ though
[08:01:14] <kshishkov> hmm, no shipping outside .se :(
[08:01:18] <pJok> mmmh, swedish cheese
[08:03:37] <Tjoppen> send money to random swedish ffdev and they can send you the cheese? buying it is simple enough - they have it at normal stores :)
[08:04:16] <kshishkov> Tjoppen: I still have to figure how to send money
[08:04:27] <Tjoppen> btw, there's a trocadero candy nowadays. or at least about a year ago
[08:04:48] <KotH> what's Trocadero?
[08:04:55] <kshishkov> ask Diego
[08:05:05] <KotH> which one?
[08:05:16] <kshishkov> the one in Italy, of course!
[08:05:25] <KotH> ^^'
[08:05:32] <KotH> both of them are in .it currently
[08:05:44] <Tjoppen> http://en.wikipedia.org/wiki/Trocadero_%28drink%29
[08:05:44] <kshishkov> that's why I said that
[08:06:03] <thresh> looks like 'Ð¶Ð¸Ð²Ñ‡Ð¸Ðº'
[08:06:06] <pJok> ffcheese?
[08:06:20] * KotH dips kshishkov in a big pot of dave's
[08:06:37] <kshishkov> thresh: Ð²ÐºÑƒÑ ÑÐ¾Ð²ÐµÑ€ÑˆÐµÐ½Ð½Ð¾ Ð½Ðµ Ñ‚Ð¾Ñ‚. ÐÑ…Ð¸Ð½Ð°Ñ†ÐµÑŽ Ð½Ðµ Ð´Ð¾Ð±Ð°Ð²Ð»ÑÑŽÑ‚, Ð½Ð°Ð¿Ñ€Ð¸Ð¼ÐµÑ€.
[08:06:53] <KotH> looks like the swedish variant of ramune
[08:07:11] <kshishkov> you can't compare it
[08:07:11] <thresh> kshishkov: Ñ Ñ‚ÑƒÑ‚ Ð½Ð°ÑˆÐµÐ» Ð¼Ð°Ð³Ð°Ð·Ð¸Ð½, Ñ‚Ð¾Ñ€Ð³ÑƒÑŽÑ‰Ð¸Ð¹ "Ð¶Ð¸Ð²Ñ‡Ð¸ÐºÐ¾Ð¼".. Ð¿Ð»Ð¾Ñ…Ð¾, Ñ‡Ñ‚Ð¾ Ð¾Ð½ Ð² 10ÐºÐ¼ Ð¾Ñ‚ Ð¼Ð¾ÐµÐ³Ð¾ Ð´Ð¾Ð¼Ð°
[08:07:16] <thresh> Ð¶Ð¸Ð²Ñ‡Ð¸Ðº ÑÑ‚Ð¾ Ð½Ð°ÑˆÐµ Ð²ÑÑ‘
[08:07:41] <kshishkov> thresh: Ð´Ð° Ð½Ñƒ? Ð Ð±Ð¾Ð´ÑÐ¶Ð°Ñ‚ ÐµÐ³Ð¾ Ð² ÐšÐ¸ÐµÐ²Ðµ Ð²ÐµÐ´ÑŒ
[08:08:05] <thresh> kshishkov: Ð½Ñƒ Ñ Ñ€Ð°Ð½ÑŒÑˆÐµ Ð½Ð¸Ð³Ð´Ðµ Ð½Ðµ Ð²Ð¸Ð´ÐµÐ» Ð² Ð Ð¾ÑÑÐ¸Ð¸, Ñ‚Ð¾Ð»ÑŒÐºÐ¾ ÐºÐ¾Ð³Ð´Ð° Ð±Ñ‹Ð» Ð² Ð£ÐºÑ€Ð°Ð¸Ð½Ðµ Ð¿Ð¸Ð» Ð¿Ð¾ÑÑ‚Ð¾ÑÐ½Ð½Ð¾ :))
[08:08:48] <kshishkov> thresh: Ð´Ð°, Ð½Ð°Ð¿Ð¸Ñ‚Ð¾Ðº Ð½Ðµ Ð¿Ð»Ð¾Ñ…Ð¾Ð¹, Ñ‚Ð¾Ð»ÑŒÐºÐ¾ Ð·Ð° Ð¿Ñ€ÐµÐ´ÐµÐ»Ð°Ð¼Ð¸ ÐšÐ¸ÐµÐ²Ð° Ð¿Ð¾Ñ‡ÐµÐ¼Ñƒ-Ñ‚Ð¾ Ñƒ Ð½ÐµÐ³Ð¾ Ð²ÐºÑƒÑ Ð´Ñ€ÑƒÐ³Ð¾Ð¹
[08:09:27] <Tjoppen> damn you and your unicode :)
[08:09:56] <KotH> Tjoppen: 's isch müesam, gäll?
[08:10:05] <spaam> Tjoppen: inge fel pÃ¥ unicode ;)
[08:10:17] <kshishkov> spaam: precis
[08:10:30] * elenril wonders Ð³Ð´Ðµ Ð´Ð¾ÑÑ‚Ð°Ñ‚ÑŒ Ð¶Ð¸Ð²Ñ‡Ð¸Ðº Ð² Ñ‡ÐµÑ…Ð¸Ð¸
[08:11:07] * KotH sees ÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐ's
[08:11:19] <Tjoppen> I just see \u's
[08:11:24] <Tjoppen> \fuuuuuu
[08:11:32] <mchinen> ãƒ©ãƒ ãƒã£ã¦ï¼Ÿ
[08:11:33] <KotH> \fuuuuuuck ?
[08:11:57] <spaam> Tjoppen: why do you use iso8859-1 ?
[08:12:18] <KotH> 'cause all cool guys do :)
[08:12:22] <kshishkov> elenril: well, there are two guys contributing FFmpeg patches from Kiev. Try convincing tipok to send you some ;)
[08:12:30] <Tjoppen> because our school does. I had a stab at using utf-8 a while back, but so many other channels messed up
[08:12:59] * elenril uses utf-8 without any problems :/
[08:13:06] <wbs> Tjoppen: you still can configure irssi to recode whatever is input to the charset your terminal uses
[08:13:26] <elenril> kshishkov: i can get "some" without much trouble
[08:13:29] <elenril> but not enough
[08:13:54] <wbs> you won't be able to see the fancy utf8 chars that can't be recoded to latin1, you'll just see '?' instead of them, but it won't mess things up at least
[08:14:23] <kshishkov> mchinen:ã€€æ—¥æœ¬ã®é£²ã¿ç‰©ãŒé¢ç™½ã„ãªã„
[08:14:23] <elenril> â˜ƒ!
[08:14:25] <KotH> wbs: and that's exactly what i dont want
[08:14:48] <KotH> wbs: i want the raw data
[08:15:02] * kshishkov tries to remember that XKCD with basilisk stare unicode symbol
[08:15:04] <wbs> well in that case, don't enable it :-)
[08:15:25] <KotH> dont worry, i wont :)
[08:16:23] <mchinen> kshishkov: any country with drinks whose names end with "sweat" are bound to be of some interest
[08:16:51] <KotH> mchinen: which are those?
[08:16:54] * KotH knows only one
[08:17:21] * mchinen too
[08:17:26] <kshishkov> sounds like some stuff with alcohol
[08:17:39] <astrange> it's a loan word
[08:17:49] <astrange> "sweat" means "exercise" in japanese now
[08:18:10] <KotH> astrange: and a few other things ;)
[08:18:20] * kshishkov still thinks it means "bodily fluid seen after exercise"
[08:18:41] <astrange> and they call corn dogs "american dogs"
[08:18:42] <mchinen> wow didn't know that
[08:19:10] <kshishkov> astrange: they may have a point there
[08:22:13] <mchinen> is there a way to get ffplay to display more than a few frames of video for large file?
[08:23:03] <kshishkov> it should display all frames by default
[08:23:10] <lu_zero> pocari is wonderful
[08:23:38] <KotH> another addict!
[08:24:06] <astrange> large resolution or large filesize?
[08:24:09] <mchinen> for 720p mine always stops after a few but the audio continues
[08:24:21] <astrange> it works for me
[08:24:24] <mchinen> sorry, resolution
[08:24:27] <astrange> what about audio off?
[08:24:31] <mchinen> i'm on a slower cpu
[08:24:40] <astrange> the sdl video is REALLY slow so it's probably dropping frames
[08:24:47] <lu_zero> KotH: you too?
[08:25:03] <astrange> ffplay cpu use is like 90% sdl 10% ffmpeg
[08:25:15] <lu_zero> uhmm
[08:25:16] <KotH> lu_zero: i'm more the gogo no kocha type... the one with lemon :)
[08:25:19] <kshishkov> Tjoppen: do they drink Trocadero in your region?
[08:25:35] <wbs> mchinen: tried -noframedrop?
[08:25:38] <mchinen> audio off gives a bunch more frames, but then stop until i seek
[08:25:38] <lu_zero> lemon reminds me CC
[08:25:44] <KotH> mchinen: you are aware that you're asking user questions?
[08:25:47] <lu_zero> yet another beverage I like
[08:26:04] <kshishkov> KotH: at least he was at LinuxTag
[08:26:05] <lu_zero> (another is the yuzu juice)
[08:26:23] <mchinen> wbs: that works thanks
[08:26:43] <lu_zero> noframedrop seems to fix a lot lately...
[08:27:02] <KotH> kshishkov: i was too... just not this year
[08:27:11] <KotH> kshishkov: and unlike you, i was at fosdem ;)
[08:27:21] <lu_zero> btw
[08:27:22] <mchinen> KotH: yeah, if thats not desired here I'll moderate myself in the future
[08:27:36] <lu_zero> news regarding september?
[08:28:15] <kshishkov> KotH: unlike you I spent more time in Ukraine
[08:28:37] <spaam> haha
[08:29:23] <kshishkov> KotH: also I attended Swedish FFmpeg Developers Conference in spring 2009
[08:38:17] <merbzt> do we have mms support now ?
[08:38:41] <kshishkov> ask that in ~5-6 hours
[08:38:47] <mru> only if all the files have been added
[08:39:13] <Tjoppen> kshishkov: yes
[08:39:33] <Tjoppen> since umeå is in norrland. in fact, finding trocadero is say stockholm is fairly hard
[08:39:41] <Tjoppen> *in
[08:39:44] <kshishkov> nope
[08:39:52] <kshishkov> I had no troubles with that
[08:40:12] <kshishkov> finding *good* Trocadero is a different story though
[08:40:25] <Tjoppen> odd. I lived there for two years, and the stores very rarely had it
[08:40:42] <kshishkov> ask merbzt
[08:42:08] * kshishkov bought Trocadero in different regions of Stockholm - VÃ¤llingby, SkÃ¤rholmen, Norrmalm
[08:42:34] <merbzt> Tjoppen: it's easy to find now
[08:42:55] <kshishkov> merbzt: not here though :(
[09:39:17] <Tjoppen> there's even a light trocadero nowadays. heresy!
[09:40:12] <merbzt> er what
[09:42:02] * kshishkov refuses to believe that
[09:43:16] <kshishkov> next thing you can order Surstromming sandwich with it
[11:28:40] <lu_zero> <@kshishkov> next thing you can order Surstromming sandwich with it
[11:28:43] <lu_zero> why not?
[11:29:39] <kshishkov> I think we've discussed Surstromming here
[11:30:05] <kshishkov> it's traditional food not to be spoiled by fast food chains
[11:30:37] * kshishkov looks at traditional pie of Italian poor people
[11:30:39] <lu_zero> theheh
[11:30:48] <lu_zero> which pie?
[11:31:05] <lu_zero> btw which kind of cheese do you like?
[11:31:33] <kshishkov> pie made from leftovers
[11:31:41] <thresh> a pie is a lie
[11:31:45] <lu_zero> ah
[11:31:47] <lu_zero> brb
[11:32:07] <kshishkov> and I like different kinds of cheeses - maybe except for too runny or mouldy (i.e. majority of French cheeses)
[11:32:45] <kshishkov> thresh: not that one
[11:34:39] <pJok> kshishkov, danablu!
[11:34:52] <pJok> nice mouldy cheese
[11:35:33] <kshishkov> pJok: exactly, that's kind of cheese I don't like much
[11:36:49] <pJok> kshishkov, french cheeses have a way of overdoing it... i find danish mouldy cheeses a lot easier going
[11:37:34] <kshishkov> pJok: could be. Personally I like Bavaria Blu
[11:39:09] <KotH> o_0
[11:39:16] <KotH> that's awfull...
[11:39:24] <KotH> cheap, industry cheese
[11:39:47] <KotH> or rather cheese surogat
[11:41:07] <kshishkov> good troll
[11:41:49] * kshishkov force feeds a slice of Ukrainian cheese to KotH
[11:45:11] <KotH> i'm sorry that i grew up in two countries that good dairy food highly
[11:45:20] <KotH> that value*
[11:46:02] <kshishkov> so is Ukraine. That's why I ate cheeses from Baltic countries in my childhood
[12:02:50] <lu_zero> kshishkov: Parmigiano is within your taste?
[12:02:59] <lu_zero> or Gorgonzola ?
[12:04:00] <kshishkov> lu_zero: the first one is, the second is probably not
[12:06:15] * lu_zero take a note to try to bring some the next time
[12:06:30] <merbzt> does ffserver support streaming over rtsp to android devices ?
[12:06:45] <lu_zero> merbzt: try
[12:07:00] <lu_zero> android stack is _really_ picky regarding codecs
[12:07:15] <lu_zero> what are you trying to do?
[12:07:56] <merbzt> I guess any source -> streaming -> to android
[12:08:16] <kshishkov> bbl
[12:20:25] <janneg> merbzt: ask hyc. there were difficulties
[12:20:44] <merbzt> hyc: ^^
[12:21:50] <lu_zero> merbzt: the best way is trying =)
[12:25:15] <janneg> merbzt: http://forum.xda-developers.com/showpost.php?p=6495551&postcount=8
[12:26:41] <janneg> merbzt: https://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2010-May/088706.html
[12:27:28] <janneg> merbzt: so the answer seems to be yes
[12:29:32] <merbzt> k thanks
[13:40:30] <CIA-93> ffmpeg: rbultje * r24785 /trunk/libavformat/rtsp.c:
[13:40:30] <CIA-93> ffmpeg: Send OPTIONS request at a regular basis to standard RTSP servers as well,
[13:40:30] <CIA-93> ffmpeg: this prevents a time-out which closes the TCP connection and kills our
[13:40:30] <CIA-93> ffmpeg: session.
[13:40:30] <CIA-93> ffmpeg: see "Re: [FFmpeg-devel] [PATCH] rtsp.c: keep-alive" thread on mailinglist.
[13:45:36] <BBB> j0sh: one thing I'd like you to do, and it's maybe not quite what you'd expect, but can you make a list of all depayloaders we have, and put it in doc/something, and make a comparison of what others have, e.g. whatever lib mpalyer/vlc/xine uses, gstreamer, etc?
[13:46:15] <BBB> shouldn't take you too long, and it'd help in convincing the vlc people to use ffmpeg instead of their own homegrown solution
[13:47:03] <kshishkov> I think the only thing that may convince VLC devs is a big pointy stick
[13:48:20] <BBB> well, we'd have to ask j-b
[13:48:23] <j-b> ?
[13:48:27] <BBB> hi j-b :)
[13:48:33] <BBB> let's talk at ovc in october
[13:48:38] <j-b> VLC sucks, what else?
[13:48:48] <Dark_Shikari> j-b: matroska seeking is broken
[13:48:56] <j-b> Dark_Shikari: duh?
[13:49:38] <j-b> 15:47 <@kshishkov> I think the only thing that may convince VLC devs is a big pointy stick
[13:49:43] <j-b> kshishkov: about?
[13:49:44] <BBB> j-b: I can help you fix it if you make it use rtsp-ffmpeg by default
[13:50:13] <j-b> Dark_Shikari: I thought it was fixed in 1.1.x
[13:50:33] <j-b> BBB: if it is tested, why not;
[13:50:48] <j-b> live555 is the worse thing ever
[13:51:08] <wbs> indeed
[13:51:10] <Tjoppen> is there any particular reason why av_open_input_stream() can't probe/needs fmt?
[13:51:58] <wbs> Tjoppen: since av_open_input_file opens the file for you, it also can probe at the same time (all probes and protocols don't want you to open a file at all)
[13:52:26] <Tjoppen> so? just read some data, probe it and rewind the buffer
[13:52:40] <wbs> that's what av_open_input_file does
[13:52:47] <BBB> j-b: ok, I'll ask j0sh to make a patch for you then (it's officially part of his soc and he did most of the work to make it awesome)
[13:52:58] <kshishkov> j-b: VLC should use FFmpeg demuxers and network handlers
[13:53:03] <Tjoppen> yes, but it should be possible to probe a stream as well
[13:53:11] <Tjoppen> for instance, it can probe stdin
[13:53:21] <wbs> Tjoppen: you can use av_open_input_file with streams too
[13:53:27] <BBB> vlc should not do anything
[13:53:28] <Tjoppen> you can?
[13:53:47] <BBB> we should convince them that our handlers are better and make vlc convert 1-by-1 where we've shown to be better
[13:53:53] <wbs> yes, streams and files are exactly the same at ByteIOContext/URLContext level
[13:53:57] <wbs> some are seekable, some not
[13:54:06] <Dark_Shikari> BBB: but then it wouldn't be VLC.
[13:54:21] <wbs> the difference is that with av_open_input_stream, the caller allocates the ByteIOContext, with av_open_input_file, it is done internally (if needed)
[13:54:23] <BBB> ssshhhhhh
[13:54:30] <BBB> let's keep the taking-over-the-world a secret
[13:54:35] <Dark_Shikari> BBB: What makes VLC VLC is its brokennness
[13:54:36] <Tjoppen> yes, but how exactly? some parameter to av_open_input_file() or do you just replace AVFormatContext::pb?
[13:54:44] <BBB> hehe :)
[13:54:52] <BBB> Dark_Shikari: will you come to openvideoconference?
[13:54:55] <Tjoppen> ah, and perhaps set AVFMT_NOFILE
[13:55:01] <Dark_Shikari> BBB: no, I'm not interested in xiphcon
[13:55:30] <wbs> Tjoppen: well, if you set AVFMT_NOFILE, av_open_input_file won't open a ByteIOContext for you
[13:55:31] <Dark_Shikari> I don't throw away lots of money on plane flights to go to xiph press conferences
[13:55:42] <wbs> Tjoppen: what are you trying to do, exactly?
[13:56:10] <Tjoppen> I need to build a QuickTime dref file and thus need proper file offsets to essence data
[13:56:38] <Tjoppen> I was relying on pkt->pos until now, since that works for a few demuxers (but far from all). now I'm just going to look for the demuxed packet data in the stream
[13:56:40] <j-b> kshishkov: the issue by doing that is the time to do support and regressions...
[13:57:05] <j-b> kshishkov: moreover, so far network handlers of ffmpeg haven't been the best on all platform
[13:57:06] <Tjoppen> so basically scan the last megabyte or so for pkt->data, and put its offset in a table
[13:57:44] <j-b> kshishkov: not to say, that we still support older versions of libav* because linux distributions are not up to date, and that is annoying
[13:57:53] <Tjoppen> I want to avoid opening say an HTTP stream more than once, so I want a custom ByteIOContext to keep a buffer. or something like that
[13:57:59] <kshishkov> j-b: ouch
[13:58:37] <Tjoppen> there might be other use cases as well, where I need to tee off a stream
[13:58:44] <BBB> j0sh: also, one thing I'd like to be done, not sure if you're personally interested in this, is regression tests for those cases where we support both packetizing and depacketizing
[13:58:51] <wbs> Tjoppen: well, then perhaps av_open_input_stream is the right, where you allocate pb yourself. but I guess you'd have to duplicate a bit of the probing logic from av_open_input_file
[13:58:54] <kierank> if xiph can have xiphcon we should have ffmpegcon
[13:59:02] <j-b> kshishkov: but as you might have seen, we force people to not write demuxers in VLC anymore.
[13:59:14] <BBB> Dark_Shikari: we can change it into x2ffcon if enough of us attend
[13:59:17] <j-b> cf .rso, dxva2 and a few other
[13:59:17] <Dark_Shikari> kierank: by definition, no
[13:59:26] <Dark_Shikari> conventions are about talking instead of doing things
[13:59:27] <Dark_Shikari> we do things, they talk
[13:59:28] <Tjoppen> wbs: yes, that's what I suspect. but I also seem to recall someone passing pb as one of the params to av_open_input_file()?
[13:59:30] <Dark_Shikari> therefore, they have conventions.
[13:59:37] <kierank> s/talking/drinking
[13:59:40] <kierank> then you are correct
[13:59:47] <BBB> Dark_Shikari: don't forget we created the fastest video decoder for an "open video format" as per their definition
[13:59:52] <BBB> Dark_Shikari: therefore, we rule
[13:59:56] <BBB> we are awesome
[14:00:02] <kierank> but we're not free
[14:00:05] <Tjoppen> or maybe it simply was _stream(9
[14:00:09] <Dark_Shikari> BBB: again, $500 for a plane flight to go to xiphcon?
[14:00:09] <Dark_Shikari> no.
[14:00:22] <kierank> you can apply for sponsorship iirc
[14:00:23] <wbs> Tjoppen: yeah, since _file doesn't take such a parameter
[14:00:31] <BBB> what if you got sponsored?
[14:00:55] <Tjoppen> in other words: refactoring time
[14:00:58] <Dark_Shikari> but why would anyone want to sponsor me?
[14:00:58] <lu_zero> BBB: how do you plan to do regtest on rtp?
[14:01:01] <Dark_Shikari> there's nothing to talk about.
[14:01:02] <j-b> BBB: why do you want to go to xiphcon?
[14:01:04] <kierank> hmm maybe not
[14:01:04] <Dark_Shikari> Unless we finish xvp8 by then.
[14:01:04] * BBB has to somehow start reviewing sebastian's mod patches now...
[14:01:11] <BBB> j-b: it's next door
[14:01:13] <Dark_Shikari> Actually, I'll go if we finish xvp8 first.
[14:01:21] <BBB> lu_zero: dunno yet, interested in ideas
[14:01:24] <Dark_Shikari> We can announce it there.
[14:01:28] <BBB> hehehe :)
[14:01:31] * BBB kicks Dark_Shikari 
[14:01:36] * Dark_Shikari kicks BBB to start working
[14:01:38] <Tjoppen> that would mean that the fmt param could be removed from av_open_input_stream(), or at least be made optional
[14:01:38] <j-b> xpv8 ? encoder?
[14:01:39] <BBB> you're not very patient :-p
[14:01:48] <Dark_Shikari> patient?  you said you'd start work two weekends ago =p
[14:01:51] <kierank> haha that announcement would be epic
[14:01:57] <Dark_Shikari> j-b: yes
[14:02:02] <Dark_Shikari> kierank: thatstheidea.jpg
[14:02:05] <Tjoppen> apart from version bump etc.
[14:02:12] <j-b> Dark_Shikari: oh, god, please do so.
[14:02:15] <lu_zero> BBB: one way is adding a local socket proto and make ffmpeg fill it and then consume it
[14:02:25] <BBB> lu_zero: that sounds nice
[14:02:31] <kierank> BBB: put yourself down as the ffmpeg rep: http://www.openvideoconference.org/speakers/
[14:02:34] <Dark_Shikari> j-b: that was my reaction too
[14:02:44] <BBB> lu_zero: we could even make a bitexact version that for "udp-over-local-socket" messes up packet order in a prefefined way
[14:02:54] <j-b> OVC is going to be fun this year
[14:03:00] <BBB> kierank: I don't want to speak
[14:03:03] <Dark_Shikari> j-b: tell BBB to hurry up
[14:03:06] <lu_zero> BBB: given that once we got a local socket we can just fill it
[14:03:22] <BBB> kierank: speaking is for wimps... real men code
[14:03:28] <lu_zero> ok, I have something to do for the day
[14:03:31] <Dark_Shikari> +1.  now get coding.
[14:03:31] <kierank> BBB: you don't have to speak. it's just a list of people from particular projects at the bottom
[14:03:33] <j-b> Dark_Shikari: so I can remove libvpx from my build
[14:03:37] <Dark_Shikari> j-b: YES
[14:03:46] <Dark_Shikari> j-b: we'll make it a runtime option
[14:03:49] <Dark_Shikari> param->b_vp8 = 1
[14:03:59] <Dark_Shikari> param_apply_profile( h, "vp8" );
[14:04:03] <Dark_Shikari> or whatnot.
[14:04:03] <BBB> Dark_Shikari: loink to your patch again?
[14:04:09] <BBB> I'll apply it locally and get starte
[14:04:11] <BBB> I have time today
[14:04:20] <Dark_Shikari> awesome, we can work through it in realtime
[14:04:30] <Dark_Shikari> Let's start a dedicated channel.
[14:04:32] <Tjoppen> btw, qt ref files are pretty cool. they work with essence in any format that keeps the data uncompressed and in byte alignment
[14:04:33] <Dark_Shikari> #xvp8
[14:05:34] <BBB> peloverde: can you join also?
[14:08:58] <BBB> peloverde: #x264dev
[14:09:10] <Dark_Shikari> yeah, we're moving development there so we can involve existing devs/etc
[14:09:11] <kierank> lol
[14:12:01] <lu_zero> btw
[14:12:22] <Tjoppen> wbs: I'll propose a patch on the ML later. I imagine more people could use a function like av_open_probe_input_stream()
[14:13:06] <lu_zero> I'm thinking about hijacking part of a local cloudcamp in december for ffmpeg purposes if enough people are willing to go there
[14:15:32] <kierank> which cloudcamp?
[14:23:13] <lu_zero> kierank: some being organized in Torino
[15:06:57] <CIA-93> ffmpeg: stefano * r24786 /trunk/ (5 files in 2 dirs):
[15:06:58] <CIA-93> ffmpeg: Implement av_get_image_linesize() and use it in
[15:06:58] <CIA-93> ffmpeg: ff_get_plane_bytewidth().
[15:06:58] <CIA-93> ffmpeg: The new implementation is more generic, more compact and more correct.
[15:06:58] <CIA-93> ffmpeg: stefano * r24787 /trunk/ (4 files in 2 dirs):
[15:06:58] <CIA-93> ffmpeg: Deprecate ff_get_plane_bytewidth() in favor of
[15:06:59] <CIA-93> ffmpeg: av_get_image_linesize().
[15:10:35] <CIA-93> ffmpeg: stefano * r24788 /trunk/doc/APIchanges:
[15:10:36] <CIA-93> ffmpeg: Add APIchanges entry after the addition of av_get_image_linesize() of
[15:10:36] <CIA-93> ffmpeg: r24787.
[15:11:41] <lu_zero> BBB: uhm
[15:11:56] <lu_zero> unix socket as tcp-like or udp-like?
[15:12:19] * lu_zero is for tcp-like and/or seqpacket if available
[15:12:29] <BBB> tcp-like, I think
[15:12:44] <Honoome> tcp-like is not portable
[15:12:52] <Honoome> iirc
[15:13:00] <lu_zero> iirc is the udp-like non portable
[15:13:15] <lu_zero> and the seqpacket nearly exotic
[15:27:50] <Honoome> lu_zero: never as exotic as mqueue
[15:28:54] <lu_zero> Honoome: mqueue seemed neat
[15:29:24] <Honoome> lu_zero: and I can seem a serious person, in a photo
[15:32:23] <kshishkov> Honoome: photoshop is your friend
[15:32:34] <Honoome> definitely it is not
[15:32:39] <Honoome> but it's friendlier than libev
[15:36:42] <Honoome> lu_zero: I got good news
[15:40:27] <Compn> there isnt any way to remove the timestamps/timecodes and start from zero in an mpegts file is there ?
[15:40:49] <Compn> e.g. frame 1 starts at 62000 seconds
[15:40:59] <Compn> can we make it start at 0 seconds ?
[15:42:43] <BBB> "../ffmpeg-svn/ffmpeg_g -i movie.webm -f yuv4mpegpipe -" <- why doesn't that owork?
[15:42:52] <BBB> if I change "-" to a regular file it works
[15:43:02] <kierank> -y ?
[15:43:22] <BBB> nope
[15:43:23] <Compn>   Duration: 00:00:06.86, start: 74257.767711, bitrate: 11941 kb/s
[15:43:43] * Compn wants start: 0
[15:43:48] <Compn> BBB : try /dev/fd0 ?
[15:43:56] <Compn> or whatever stdin is, fd3 ?
[15:44:15] <BBB> fd1
[15:44:23] <Compn> try putting it ffmpeg - -i blah
[15:44:25] <BBB> fd0 is stdin, I want to write to stdout
[15:44:44] <Compn> ffmpeg cares about order of options
[15:45:33] <BBB> - before -f fails
[15:45:40] <BBB> "unable to find output format for ..."
[15:46:32] <elenril> worksforme
[15:47:23] <elenril> blame your shell?
[15:47:28] <BBB> ?
[15:47:32] <BBB> no, /dev/stdout fails also
[15:48:44] <elenril> /dev/stdout is bashism iirc
[15:49:00] <elenril> and what do you mean by 'fails'
[15:49:09] <BBB> wc -b dies immediately
[15:49:20] <BBB> I think wc -b is stupid
[15:49:23] <BBB> x264 encodes fine
[15:49:24] <Dark_Shikari> what stupid OS are you using?
[15:50:14] <elenril> sounds funny coming from you =p
[15:51:22] <Dark_Shikari> even windows can pipe
[15:52:11] <Compn> BBB : what shell version?
[15:52:27] <kierank> BBB: just use mkfifo video.y4m and use that as apipe
[15:52:43] <BBB> nah, it works a little
[15:52:55] * BBB goes do random stuff to make x264 output vp8 now
[15:53:38] * elenril wonders when did everybody turn into a troll here
[15:53:52] <Compn> lol
[15:54:13] <kierank> people read too much tvtropes
[15:54:32] <Compn> well i think after about 5 mails/replies to -devel , the outer shell hardens and the inner troll metamorphases into a large troll
[15:54:43] <kshishkov> sounds plausible
[15:54:54] <kshishkov> Compn: s/devel/user/
[15:55:11] <Compn> it can be quicker if the first reply is something like 'i did in 2 lines what you made a 100 line patch to do'
[15:58:05] <kshishkov> during my initial contribution phase I've not interacted with FFmpeg ML at all
[15:58:30] <elenril> that's why you don't troll so much?
[15:58:56] <kshishkov> could be
[15:59:22] <kshishkov> or I'm just too lazy
[16:15:54] <lu_zero> \o/
[16:49:53] <j-b> BBB: ping
[16:49:58] <BBB> yeah
[18:03:08] <Dark_Shikari> oh wow. vc-1's inter MV prediction is a bit interesting.
[18:10:30] <kshishkov> why?
[18:11:34] <merbanan> kshishkov: there ?
[18:11:47] <kshishkov> merbanan: yes
[18:12:29] <Dark_Shikari> kshishkov: median4
[18:12:39] <kshishkov> ah, that
[18:14:06] <kierank> do any codecs have a special postfilter for lfe?
[18:15:01] <kshishkov> what for?
[18:15:12] <merbanan> kierank: dca has a separate filter
[18:17:21] <kshishkov> well, DCA is a special case
[18:17:50] <kshishkov> they use 4 or 2 coefficients to make 256 samples of LFE data
[18:18:29] <kshishkov> so input is processed with IIR or FIR filter (depending on codec settings) to reconstruct it that way
[18:19:03] <kshishkov> AC3, for example, just stores 7 out of 256 possible coeffs for LFE channel
[18:20:58] <kshishkov> AAC (IIRC) just forbids switching to short window mode or so
[18:21:15] <kshishkov> no special postprocessing anywhere
[18:21:51] <CIA-93> ffmpeg: stefano * r24789 /trunk/libavfilter/avfilter.h:
[18:21:51] <CIA-93> ffmpeg: Cosmetics: put "{" on the same line of the "struct" keyword in struct
[18:21:51] <CIA-93> ffmpeg: definitions.
[18:21:51] <CIA-93> ffmpeg: More consistent with the rest of FFmpeg.
[18:29:50] <Dark_Shikari> mru: why is get_be24 slower than get_be32 >> 8?
[18:29:57] <Dark_Shikari> aka why does intreadwrite suck so much
[18:30:30] <kshishkov> maybe because three byte accesses may be slower than one 32-bit
[18:30:40] <Dark_Shikari> then why are there 3 byte accesses?
[18:30:45] <Dark_Shikari> get_be16 doesn't do 2 byte accesses
[18:30:49] <Dark_Shikari> so why does get_be24 do 3?
[18:32:37] <kshishkov> it's not that widely used and it's not arch-optimized
[18:34:01] <kshishkov> so r/w 3 bytes is straightforward and nobody has complained about it until now
[18:34:39] <Dark_Shikari> I wanted to make the vp8 rac read 3 bytes instead of 2
[18:34:59] <Dark_Shikari> then again, I tested and get_be32 >> 8 is exactly as fast as reading 2 bytes instead of 3
[18:35:02] <Dark_Shikari> no speed gain :/
[19:43:03] <Compn> arg
[19:43:21] <Compn> trying to find samples for a100 wav tag so it can be tested with that g723.1 patch
[19:43:30] <Compn> cant find anything , even binary codec
[19:43:40] <Compn> 0xA100 Comverse Infosys Ltd. G723 1
[20:29:33] <BBB> nfl: one more thing, make sure to use get_bits_left() at strategically relevant places to prevent overruns on your input buffer
[20:33:27] <nfl> BBB: you're referring to decode_frame()?
[20:34:48] <BBB> yes
[20:35:00] <wbs> BBB: regarding rtpdec_latm, no, didn't commit it yet, janneg wasn't really decided on whether we still should use a chained demuxer or standalone code
[20:35:03] <BBB> whenever you use get_bits(), make sure there's enough data left in your input buffer
[20:35:19] <BBB> wbs: commit it, there's no reason to wait, rtp isn't fully latm-compliant anyway
[20:35:26] <BBB> unless janneg can give us a good reason to wait
[20:35:36] <nfl> BBB: wouldn't a size check in the beginning (as of now) be enough?
[20:35:44] <BBB> nfl: oh, it's there already?
[20:35:49] <BBB> nfl: I guess I missed it
[20:35:58] <BBB> nfl: yes, that's good enough also
[20:36:08] <nfl> ok
[20:40:59] <wbs> BBB: ok
[20:41:02] <wbs> janneg: ping?
[20:46:06] <janneg> wbs: pong
[20:47:48] <wbs> janneg: any objections to committing the standalone rtp/latm code?
[20:48:45] <j0sh> wbs: is luca A averse to the 'reply all' button? :)
[20:49:35] <wbs> j0sh: he's not subscribed to -soc
[20:50:07] <j0sh> ahh alright
[20:50:14] <wbs> so I guess all his mails are in some moderation queue :-)
[20:50:36] <j0sh> heh
[20:56:52] <lu_zero> ^^;
[20:57:13] <lu_zero> good $time_that_I_cannot_define
[20:57:20] * lu_zero just woke up...
[20:58:21] <j0sh> it's always 'good morning' somewhere in the world
[20:59:35] <janneg> wbs: no
[21:01:11] <wbs> janneg: ok, nice
[21:03:12] <j-b> http://www.mainmediasoft.com/products/f4vencoder/  ships ffmpeg with h264 decoding and x264
[21:03:17] <j-b> hall of shame?
[21:03:34] <Dark_Shikari> I just sent them a legal threat
[21:03:44] <Dark_Shikari> From the official-looking x264.com email address.
[21:04:54] <lu_zero> oh
[21:05:22] <lu_zero> nice
[21:06:02] <j-b> what notiorious parts of ffmpeg are still gpl-only?
[21:07:54] <j-b> BBB: HERE
[21:08:10] <CIA-93> ffmpeg: mstorsjo * r24790 /trunk/ (6 files in 2 dirs): Add RTP depacketization of MP4A-LATM
[21:08:22] <BBB> j-b: x264, some x264 asm, some h264 asm marked with x264 in the string, some xvid asm with xvid int he string, libmpeg2 asm with that string
[21:08:25] <BBB> j-b: that's about it
[21:08:33] <BBB> grep x264 grep xvid grep libmpeg2
[21:08:38] <BBB> if these are negative, it's likely lgpl
[21:08:53] <j-b> x264 is in it
[21:08:56] <BBB> gpl
[21:09:00] <BBB> where are they located?
[21:09:19] <j-b> http://pastebin.com/Uy0tPCZn
[21:09:22] <j-b> USA, TX
[21:09:38] <BBB> good, will FW to SFLC
[21:09:41] <BBB> money money money
[21:09:42] <BBB> oops
[21:09:46] <j-b> sure.
[21:10:11] <j-b> BBB: http://mailman.videolan.org/pipermail/vlc-devel/2010-August/076625.html initial report
[21:17:35] <lu_zero> BBB: theheh
[21:18:02] <lu_zero> btw nobody said anything about the ffmerch ideas =P
[21:18:16] <merbanan> lu_zero: didn't I ?
[21:20:08] <lu_zero> merbanan: you didn't say what we should do next =)
[21:20:19] * lu_zero should be more precise ^^
[21:21:21] <merbanan> my though was to find a cup merchant that could take care of orders for us
[21:22:02] <lu_zero> something like cafepress?
[21:22:42] <lu_zero> it shouldn't take me much effor have an ffmpeg eshop setup
[21:23:28] <lu_zero> uhmm
[21:23:52] * lu_zero has is new fun protocol to try... how?
[21:25:38] <pJok> hrm
[21:25:59] <pJok> was there still a need for interlaced mpeg2?
[21:29:22] <lu_zero> BBB: btw which name would be better, local or unix?
[21:29:55] <lu_zero> ffmpeg -i something.mov -vcodec copy -acodec copy -f mov unix:/path/to/my/socket
[21:29:57] <pJok> as in samples
[21:29:59] <lu_zero> unix
[21:33:13] <BBB> unix probably
[21:47:54] <lu_zero> wbs: willing to test a feng branch for us?
[21:59:35] <spaam> lu_zero: so wbs is your test subject? ;)
[22:09:22] <lu_zero> more or less ^^;
[22:16:52] <lu_zero> spaam: do you want to act as subject as well?
[22:16:57] <lu_zero> merbzt: you?
[22:17:00] <lu_zero> ^^