[FFmpeg-devel-irc] IRC log for 2010-04-05

Tue Apr 6 02:00:07 CEST 2010

[00:41:27] <pentan0l> mru it stat without optimisations, but crashed in the same place but more late
[07:06:19] * Terminating due to: TERM
[07:06:34] * /join #ffmpeg-devel ...
[07:06:36] *** TOPIC: Welcome to the FFmpeg development channel. That is, development of FFmpeg, not using FFmpeg, nor libav*. | Users should redirect their questions to #ffmpeg | FFmpeg 0.5.1 has been released! | this chat is now publicly logged.
[07:06:36] *** TOPICINFO: superdump!~rob at unaffiliated/superdump, 1267604523
[07:06:56] <kshishkov> looks like this channel is selectively logged now
[07:09:14] <Dark_Shikari> selectively?
[07:09:39] <kshishkov> yep, since bot joined this channel only four minutes ago
[07:11:02] <_av500_> it was on easter vacation
[08:11:02] <superdump> do we have a template for the lgpl header somewhere?
[08:12:45] <peloverde> copy a random source file?
[08:13:26] <kshishkov> yep, either that or ask Diego
[08:51:54] <pentanol> mru here dump this func on i386  http://codepad.org/TZ8xlTMb  will try compile gdb to my camera
[08:52:39] <mru> x86 dump is useless
[09:09:13] <KotH> moin
[09:10:58] <kshishkov> shalom
[09:18:16] <KotH> .o0(stupidity prevails)
[09:24:16] <kshishkov> KotH: that's this world official motto
[09:25:11] <KotH> guess so
[09:25:49] <KotH> a guy asks to remove a mail he send to mplayer-users because he mentioned zshare which seems to be a client of them. and this client doesnt want to be mentioned on mailinglists
[09:26:11] <KotH> i told him politely that it's not worth the effort to remove it because -users is mirrored on a dozen other websites
[09:26:30] <KotH> his answer simply was "i dont care about the other websites, just remove it"
[09:57:51] <pross-au> Hi
[10:00:27] <KotH> haudi!
[10:00:43] <pross-au> Giday
[10:14:33] <kshishkov> pross-au: mate, check this out - http://shishkov.homeunix.net/WILDCAT.avi
[10:19:21] <pross-au> i am on dialup
[10:19:29] <pross-au> will check it out tomorrow
[10:19:38] <kshishkov> it's 900k
[10:19:49] <kshishkov> but no worries
[10:32:13] <pross-au> That video is not familar to me kshishkov
[10:32:32] <kshishkov> yep, but is seems a bit australian to me
[10:33:07] <pross-au> It does
[10:33:20] <pross-au> Bouncng kangaroos and what looks to be a tasmanian tiger
[10:34:10] <kshishkov> that's one of the logos from Discworld II game
[10:34:29] <pross-au> Ah
[10:34:34] <pross-au> I never got into those
[10:34:37] <kshishkov> converted with quick port of decoder from ScummVM to lavc
[10:34:40] <pross-au> Nice
[10:34:56] <kshishkov> but here it's content that matters
[10:35:20] <pross-au> Yep!
[10:35:27] <pross-au> Decoder looks good
[10:35:30] * kshishkov wonders if dangling corks are really good against flies
[10:35:42] <pross-au> Yes
[10:36:44] <pross-au> The flies not do much harm. You can learn to ignore them
[10:36:48] <kshishkov> ok, I'll inform you if there's any progress with BIKb decoder
[10:37:55] <pross-au> No worries
[10:37:59] <kshishkov> :)
[10:38:03] <pross-au> Mate
[10:38:17] <pross-au> I hacked together a Siren7 decoder today
[10:38:23] <kshishkov> good
[10:40:12] * kshishkov curses that larrikin who wrote TAK
[10:42:53] <pross-au> You're doing well with the aussie slang kshishkov
[10:43:13] <pross-au> I reckon you'd ace the citizenship test
[10:43:32] <kshishkov> unlikely - I don't drink beer
[10:43:52] <pross-au> Ah. But do you eat vegemite?
[10:44:19] <kshishkov> never got that stuff here
[10:44:39] <pross-au> Do you want some?
[10:45:13] <kshishkov> nay
[10:46:17] <kshishkov> you seem to grow other good stuff though - like macadamia nuts
[10:46:26] <kshishkov> or beans
[10:46:47] <superdump> vegemite ftw
[10:47:01] <pross-au> superdump knows best!!!
[10:47:05] <superdump> :)
[10:47:13] <superdump> well, we have a variant called marmite in the uk
[10:47:28] <superdump> but i have had vegemite before
[10:47:30] <pross-au> We have that here too superdump. At it tastes like shit.
[10:47:32] <superdump> it's basically the same
[10:47:54] <pross-au> As somebody who eats that stuff daily, there is most certainly a difference
[10:48:57] <superdump> :)
[10:49:09] <superdump> i only had them with a fair degree of separation
[10:49:19] <superdump> what's the difference in your opinion?
[10:49:35] <pross-au> vegemite has a sharper taste, more salty
[10:49:55] <kshishkov> the only known British food here is porridge
[10:50:04] <pross-au> no fish and chips?
[10:50:18] <kshishkov> not at all
[10:50:35] <thresh> you can buy some at pubs here, really overpriced though
[10:51:14] <kshishkov> like everything
[10:52:43] <kshishkov> I'd rather eat Swedish food anyway
[10:53:06] <pross-au> Wouldnt we all
[10:53:28] <kshishkov> looks like Swedes prefer Italian food ;)
[11:04:01] <pJok> köttbullar och stuvade makaroner :)
[11:05:15] <kshishkov> lax eller strÃ¶mming med potatis :P
[11:06:19] <pJok> :)
[11:06:30] <pJok> har inte prövat strömning
[11:08:24] <kshishkov> du bor inte nÃ¤ra Ã–stersjÃ¶n
[11:08:43] <kshishkov> och strÃ¶mming Ã¤r bÃ¤ttre an sill
[11:08:52] <kshishkov> (tror jag)
[11:09:54] <pJok> sill är bra
[11:10:22] <kshishkov> javisst
[11:10:33] <pJok> och jag bor nära innfarten till östersjön... :P
[11:11:51] <kshishkov> men faktiska Ã–stersjÃ¶n ligger mellan Sverige och Finland/Ingermanland
[11:23:38] <superdump> kshishkov: what have you been using to learn swedish?
[11:24:24] <kshishkov> superdump: lexin.nada.kth.se for dictionary and some found textbooks on Swedish gammar
[11:24:28] <kshishkov> *grammar
[11:24:39] <superdump> ah
[11:24:53] <pJok> lexin is nice
[11:25:02] <pJok> i use that for words mostly
[11:26:36] <pJok> grammar i just learn over time
[11:27:07] <pJok> but you're quite good at swedish, kshishkov
[11:27:40] <kshishkov> that's because you've never heard me speaking Swedish :P
[12:16:13] <elenril> http://lwn.net/Articles/381794/ what, svn is still alive?
[12:18:10] <janneg> I've heard svn is still used for the official ffmpeg repo
[12:19:02] <kshishkov> yes, that rumour keeps popping here and there
[12:19:49] <elenril> probably an evil lie
[12:34:31] <KotH> like everyone using utf-8 around here!
[13:48:36] <BBB> nfl: please already put your application in the GSoC database, if you haven't already
[13:48:39] <BBB> spyfeng: same for you ;)
[13:52:15] <BBB> wbs: do you know if our h263 rtp depacketizer outputs rfc2429 or 4629?
[13:59:42] <kshishkov> BBB, have you applied for WVP2 already?
[14:00:26] <BBB> I'm not applying ;)
[14:00:29] <BBB> I'm mentor, not student
[14:00:53] <kshishkov> I'm neither :P
[14:07:08] <BBB> you should apply, still
[14:07:57] <kshishkov> I'm not a student
[15:29:59] <nfl_> i'm supposed to pay for a "state of the art 4Mbps" connection. and this is what i get
[15:30:08] <nfl_> BBB, you were saying?
[15:30:38] <kshishkov> well, that only means your art is not so decent
[15:31:20] <nfl_> apparently :(
[15:32:03] <nfl_> so what did i miss?
[15:32:24] <nfl_> i believe you were saying something concerning me
[15:32:57] <kshishkov> no, there was nothing addressed to you after "<@BBB> nfl: please already put your application in the GSoC database, if you haven't already"
[15:36:13] <nfl> kshishkov, you mean google's database?
[15:36:34] <kshishkov> yes
[15:36:47] <kshishkov> socghop.googleapps.com or something IIRC
[15:38:02] <andoma> mru: ping?
[15:39:10] <BBB> nfl: yes, the google database
[15:39:38] <BBB> nfl: google will allocate preliminary slots in a few days, we'd want to give them a good idea of how many we need
[15:39:49] <BBB> the more applications we have, the better we can justify asking for more slots
[15:40:17] <nfl> does this mean i'm qualified from ffmpeg's side now?!
[15:53:37] <mru> andoma: pong
[16:05:40] <DonDiego> BBB: can you do me a favor and add 4 characters to libavformat rtpdec.h please?
[16:05:49] <BBB> of course
[16:05:56] <DonDiego> write 'void' into the empty parentheses at line 143
[16:06:11] <DonDiego> and then commit
[16:06:16] <DonDiego> this will fix a warning
[16:10:08] <BBB> why don't you do it? ;)
[16:12:32] <BBB> you have my permission to commit :-p
[16:12:53] <kshishkov> BBB: because all FFmpeg devs are lazy, even non-programmers
[16:12:57] <BBB> hm...
[16:12:58] <BBB> ok
[16:12:59] <BBB> will commit
[16:13:00] <BBB> 1 second
[16:13:58] <BBB> fixed
[16:14:40] <CIA-1> ffmpeg: rbultje * r22803 /trunk/libavformat/rtpdec.h:
[16:14:40] <CIA-1> ffmpeg: Add void to a function prototype that takes no arguments. Fixes a potential
[16:14:40] <CIA-1> ffmpeg: compiler warning.
[16:35:13] <DonDiego> BBB: thx
[16:35:34] <BBB> np
[16:35:39] <DonDiego> BBB: my point was also to point out that there was an (ignored) compiler warning in code you maintain..
[16:38:12] <BBB> actually, Luca Abeni is responsible for the RTP code
[16:38:15] <BBB> but I get your point
[16:38:21] <BBB> I didn't actually see the compiler warning
[16:48:27] <DonDiego> doesn't ffmpeg have -Wstrict-prototypes in CFLAGS now?
[16:57:46] <wbs> BBB: the h263 depacketizer is rfc 4629, the other one is deprecated.. don't know how widely used the old one is today
[16:58:35] <BBB> thanks, it was a question on libav-used@
[16:58:39] <wbs> ah
[16:58:40] <BBB> *user
[16:58:56] <wbs> i'm not subscribed to that one, and just browse it every now and then
[17:02:45] <BBB> don't worry, I'll forward relevant questions :)
[17:03:14] <wbs> good, good. :-_)
[17:13:28] <BBB> wbs: oh, and I was trying to tempt you into cleaning up RTP/AAC, but if you don't want, I'll do it ;)
[17:13:47] <wbs> BBB: hah :-) yeah, I've got it on my low priority todo list, too
[17:14:05] <BBB> sounds familiar ;)
[17:14:08] <wbs> it isn't really broken, it's just horribly structured compared to other formats :-)
[17:14:21] <BBB> right
[17:14:35] <wbs> but finding that it allocs and leaks one buffer per packet was quite an interesting surprise
[17:14:36] <BBB> and you're probably right, it's most likely historical, i.e. this code was there before we added rtpdynamicpayloarhandler
[17:14:47] <BBB> if you tested that patch, it's ok for now
[17:14:51] <BBB> I think
[17:14:54] <BBB> let me look at it
[17:15:16] <wbs> tested it with valgrind, seems to work as intended
[17:16:33] <BBB> you shouldn't need au_headers_allocated
[17:16:51] <BBB> oh right, I see
[17:16:55] <BBB> it's for the sequence 4-2-3
[17:17:07] <BBB> so you don't reallocate (unnecessarily) when going from high-to-low-back-to-high
[17:17:30] <BBB> I'd make it a copy of nb_au_headers
[17:17:34] <BBB> so remove the * sizeof(...)
[17:17:39] <BBB> decreases code size
[17:18:09] <BBB> also, does it need a zero check?
[17:18:19] <BBB> I mean, can au_headers_length be zero?
[17:18:26] <BBB> or nb_au_headers
[17:18:31] <wbs> don't think so
[17:18:45] <BBB> ok, so if you remove the * sizeof(), patch is ok with me
[17:18:48] <wbs> ok
[17:26:31] <CIA-1> ffmpeg: mstorsjo * r22804 /trunk/libavformat/ (rtpdec.h rtpdec.c): Fix leaks in the AAC RTP depacketizer
[17:26:32] <CIA-1> ffmpeg: mstorsjo * r22805 /trunk/libavformat/rtpdec.c: Reindent
[17:27:14] <BBB> \o/
[17:27:25] <BBB> I wonder why we never noticed this memleak
[17:27:29] <BBB> it's pretty massive
[17:28:13] <kshishkov> too much free RAM?
[17:28:30] <wbs> luckily, i try to test most of my major (and most minor, too) changes with valgrind
[17:28:50] <wbs> and it's annoying to see leaks and/or errors in other parts than the one I'm working on :-)
[17:33:47] <j-b> hello peopl
[17:36:32] <kshishkov> hej
[17:40:45] <BBB> hi j-b
[18:03:26] <BBB> wbs: good catch by the way, indeed, the whole rtp_.. file isn't needed in this case, just changing the codec payload in the rtp standard table should be enough to make it work
[18:39:36] <relaxed> the ffvhuff license is covered here -> libavcodec/huffyuv.c, correct?
[18:46:07] <astrange> yes
[18:46:18] <astrange> er
[18:46:52] <astrange> if you're distributing ffmpeg then the license of ffmpeg is the important part, try not to copy random parts of code out of it
[18:51:14] <elenril> anybody can apply a few patches for me?
[18:51:30] <Compn> they been tested?
[18:51:43] <Compn> but not me :D
[18:52:13] <elenril> =p
[18:53:33] <relaxed> astrange: right, there was someone asking about it in #ffmpeg. I told them to come here and ask to be sure.
[18:58:18] <BBB> elenril: tell me which and then I can
[18:59:34] <elenril> BBB: thread 'add AV_METADATA_DONT_OVERWRITE flag'
[19:00:09] * elenril hopes nobody bumped lavf version since then
[19:07:03] <dgt84> Does anyone know how easy it would be to get motion vectors for a frame with a single ref frame in a libavfilter filter using existing motion estimation code in libav*?
[19:07:31] <dgt84> Or would it be better to try and rewrite some of that in the filter itself?
[19:07:38] <Dark_Shikari> well imo we need a motion estimation api
[19:07:51] <Dark_Shikari> look at what mplayer's vfs do
[19:07:54] <Dark_Shikari> they ENCODE VIDEO
[19:07:58] <Dark_Shikari> and steal the MVs
[19:08:02] <Dark_Shikari> that's horrible
[19:08:16] <dgt84> yeah that was my thought here, but it does seem horrible
[19:10:27] <astrange> motion estimation for filters might need a "truemotion" mode instead of rd-optimal
[19:11:55] <dgt84> I'm afraid most of this stuff is still above my head - I was hoping for an easy way to get a list of mv's for each block in the frame and then do various things with them in my free time, such as motion compensation
[19:14:24] <Dark_Shikari> astrange: that too
[19:14:29] <Dark_Shikari> mvtools is something good to look at
[19:17:01] <pengvado> Dark_Shikari: you mean other than CODEC_FLAG2_MEMC_ONLY
[19:18:44] <Compn> dgt84 : there is lavc debug modes
[19:19:07] <Compn> dgt84 : with mplayer they can be displayed using -lavdopts debug=...
[19:19:21] <Compn> note that there is little documentation on such things
[19:19:37] <Compn> i seem to remember something about motion or macroblocks or so
[19:38:49] <Dark_Shikari> pengvado: ahhh, didn't realize it did that
[19:43:07] <BBB> elenril: working on it...
[19:47:45] <CIA-1> ffmpeg: rbultje * r22806 /trunk/ (doc/APIchanges libavformat/metadata.c libavformat/avformat.h):
[19:47:45] <CIA-1> ffmpeg: Add a flag to not overwrite existing tags
[19:47:45] <CIA-1> ffmpeg: Patch by Anton Khirnov <wyskas gmail com>.
[19:48:25] <BBB> damn it
[19:51:37] <CIA-1> ffmpeg: rbultje * r22807 /trunk/libavformat/avformat.h:
[19:51:37] <CIA-1> ffmpeg: Actually add flag (somehow got lost in my previous patchset?), fixes broken
[19:51:37] <CIA-1> ffmpeg: r22806.
[19:54:16] <CIA-1> ffmpeg: rbultje * r22808 /trunk/ffmpeg.c:
[19:54:17] <CIA-1> ffmpeg: Don't overwrite user-supplied metadata with metadata mapped from an input file.
[19:54:17] <CIA-1> ffmpeg: Patch by Anton Khirnov <wyskas gmail com>.
[19:55:10] <CIA-1> ffmpeg: rbultje * r22809 /trunk/ffmpeg.c:
[19:55:10] <CIA-1> ffmpeg: Simplify writing stream metadata.
[19:55:10] <CIA-1> ffmpeg: Patch by Anton Khirnov <wyskas gmail com>.
[19:55:20] <elenril> BBB: thanks
[19:55:42] <BBB> np
[19:56:00] <CIA-1> ffmpeg: rbultje * r22810 /trunk/doc/APIchanges: Add actually correct rXXXXX here.
[20:10:35] <mru> btw, anyone in stockholm tomorrow night?