[FFmpeg-devel-irc] IRC log for 2010-02-21

[irc at mansr.com]

[00:11:37] <CIA-90> ffmpeg: conrad * r21927 /trunk/libavcodec/vp3.c:
[00:11:38] <CIA-90> ffmpeg: Use memset to set the runs partially coded superblocks
[00:11:38] <CIA-90> ffmpeg: Much faster for long runs (e.g. nearly uncoded frames), slightly faster
[00:11:38] <CIA-90> ffmpeg: for the general case.
[00:11:38] <CIA-90> ffmpeg: conrad * r21928 /trunk/libavcodec/vp3.c: Make the special 4129 case for long-run bit strings a #define and explain it
[00:11:38] <CIA-90> ffmpeg: conrad * r21929 /trunk/libavcodec/vp3.c:
[00:11:39] <CIA-90> ffmpeg: Decode fully coded superblocks in the same manner as partial superblocks and qpi
[00:11:39] <CIA-90> ffmpeg: No speed difference, but it will simplify the special 4129 case.
[00:11:40] <CIA-90> ffmpeg: conrad * r21930 /trunk/libavcodec/vp3.c:
[00:11:40] <CIA-90> ffmpeg: Handle Theora's continued runs in superblock coding.
[00:11:41] <CIA-90> ffmpeg: This doesn't really matter yet since 4:2:0 1080p has only 3060 superblocks,
[00:11:41] <CIA-90> ffmpeg: but larger resolutions or 4:4:4 1080p could hit this case.
[00:11:42] <CIA-90> ffmpeg: conrad * r21931 /trunk/libavcodec/vp3.c:
[00:11:42] <CIA-90> ffmpeg: Simplify determing whether fragments are coded
[00:11:43] <CIA-90> ffmpeg: No measurable speed difference
[01:28:00] <Dark_Shikari> Yuvi: what other patches do you have locally?
[01:29:36] <Dark_Shikari> for theora
[01:31:41] <Yuvi> simplifying mv decode, macroblock-based mc, some intra-only special cases, DC and 10-element idct, and some optimizations to the sse2 idct
[01:31:56] <Yuvi> but my branch is pretty diverged from head
[01:44:05] <Dark_Shikari> you need to get it merged :/
[01:44:18] <Dark_Shikari> guess you're working on that though
[01:54:13] <pentanol> g'afternoon, anyway dsputil.h:449: error: expected ';', ',' or ')' before 'v1' with gcc version 4.4.3 20100108
[02:43:52] <mru> gaaaaaah
[02:47:36] <CIA-90> ffmpeg: mru * r21932 /trunk/libavcodec/vc1dec.c: VC1: fix missing include h263.h
[02:52:47] <astrange> zero/sign_extend can use builtin_constant_p and avoid any unnecessary asm
[02:53:04] <astrange> note gcc evaluates builtin_constant_p after inlining but llvm doesn't. i forgot to write a bug about that
[02:53:28] <mru> so llvm builtin_constant_p is essentially useless
[02:56:39] <Yuvi> could be a macro
[02:56:59] <mru> so 10% useful
[03:02:44] <astrange> they probably just didn't notice it
[03:02:50] <mru> tossers
[03:24:19] <CIA-90> ffmpeg: ramiro * r21933 /trunk/libavdevice/oss_audio.c: Indent.
[03:26:07] <Honoome> 10%? if it's a macro I'd expect the constant propagation to happen automatically… I'd say 5%
[03:28:16] <astrange> it's an asm macro
[03:40:34] <pentanol> hm , looks like my camera... Linux 2.6.25 on a armv5tejl based always received http://codepad.org/tIeHLYih
[03:41:03] <pentanol> weid, but why they seeding compilers for cris
[03:42:04] <Honoome> astrange: no I was more thinking along the lines about Yuvi's remark…
[04:13:29] <astrange> git merge doesn't work at all when there's conflicts
[05:05:14] <astrange> Yuvi: did you change the vp3 decoder to do luma and chroma at the same time, or is it only calling draw_horiz_band after the last plane?
[05:06:57] <Yuvi> after the last plane
[05:12:42] <ohsix> can you deencap stuff in flv's from rtmp? the source is just an mp4 for this particular thing and afaik they just shuttle it through fms
[05:12:56] <ohsix> remux them back into something else
[05:15:45] <astrange> hmm, i was using pixel_addresses_initialized. guess it didn't need it
[06:22:33] <astrange> hmm, merging h264 worked fine but vp3 is still messed up. maybe i'll not bother until the next patch
[06:52:10] * elenril pokes Honoome 
[08:43:17] <elenril> when converting utf8 to utf16, is there a better estimate of output length than strlen(input)*4?
[08:44:08] <kshishkov> no
[08:44:18] <kshishkov> but it's safe upper bound
[08:45:00] <kshishkov> unless you write utf8-aware strlen
[08:48:07] <elenril> well for 1-byte utf-8 characters you always get 2-byte utf16
[08:48:34] <elenril> so i hoped there'd be a lower upper bound
[08:49:17] <elenril> bbl
[08:49:19] <kshishkov> if you can't have 4-byte chars then upper bound is strlen()*2
[09:18:04] <_av500_> what about surrogate pairs?
[09:18:52] <jai> hm
[09:18:55] <jai> ramiro: ping
[09:19:26] <kshishkov> _av500_: presumably it will take a lot of bytes on UTF-8 input too
[09:23:04] <_av500_> they should have made utf8 a turing complete bytecode interpreter
[09:23:22] <kshishkov> why? It's simple VLC
[09:23:38] <_av500_> too restricted
[09:24:04] <kshishkov> hmm?
[09:25:19] <kshishkov> 1,114,111 characters should be enough for now
[09:25:34] <_av500_> i mean you can find the length of a string in finite time, no fun in that :)
[09:26:16] <kshishkov> yeah, take Markov chain
[09:39:03] <kshishkov> mate!
[09:39:06] <pross-au> Gday
[09:40:04] <kshishkov> unless somebody objects, I think I'll commit Bink video decoder tomorrow
[09:40:37] <Dark_Shikari>  \o/ \o/ \o/ \o/
[09:42:14] <pross-au> kshishkov: see my mail
[09:50:35] <KotH> /o\ /o\ /o\ /o\
[09:50:53] <pross-au> What's a \o/ ?
[09:51:05] <thresh> backslash o slash
[09:51:07] <kshishkov> KotH: is that Australian rejoice or facepalming?
[09:51:13] <pross-au> a happy person?
[09:51:25] <kshishkov> or praying one
[09:51:41] <KotH> kshishkov: it's ducking for cover :)
[09:53:10] * pross-au hands kshishkov a VB
[09:53:25] * kshishkov has it, including VB for DOS
[09:55:40] <pross-au> 'Victorian Bitter'
[09:56:01] <kshishkov> sounds like a beer name
[09:57:07] <pross-au> Man, you're smart cookie kshishkov
[09:59:51] <kshishkov> okay, give VB to KotH along with cookies with RFID embedded
[10:11:58] <KotH> .o0(someone wants to know where i'm going)
[10:12:23] <kshishkov> not at all, not even ICBM
[10:19:34] <pross-au> ICBMs?
[10:19:48] <kshishkov> intercontinental ballistic missiles
[10:21:03] <pross-au> Ukraine has those??
[10:21:29] <kshishkov> yes
[10:21:39] <pross-au> Cool
[10:22:11] * kshishkov should point out that 2 out of 3 Soviet rocket designers were Ukrainian. The last one was Polish.
[10:22:12] <pross-au> No nuclear hear, well except for the infite supply of uranium ore
[10:25:19] <KotH> pross-au: everyone has them
[10:25:22] <KotH> pross-au: even .ch :)
[10:34:17] <twnqx> morning
[10:43:29] <KotH> grüezi twnqx
[10:46:00] <pross-au> .ch ?
[10:46:12] <pross-au> oh right
[10:48:49] <KotH> you know, neutral, but heavly armed :)
[10:51:04] <pross-au> What do you percieve australia as? Neutral?
[10:51:18] <pross-au> *perceive
[10:51:24] <Dark_Shikari> censored
[10:51:25] <KotH> nah... .au is a backwaters continent :)
[10:51:40] <pross-au> High Five KotH
[10:52:06] * KotH Z-Fives pross-au 
[10:52:17] * pross-au goes back listening to the Beach Boys
[10:52:40] <pross-au> Back on topic: Bink looks good
[10:53:32] <Dark_Shikari> do we have any way of comparing performance to the official decoder?
[10:57:32] <CIA-90> ffmpeg: stefano * r21934 /trunk/doc/ (4 files):
[10:57:32] <CIA-90> ffmpeg: Put all the options shared amongst the ff* tools under a dedicated
[10:57:32] <CIA-90> ffmpeg: section "Generic options".
[11:13:09] <kshishkov> pross-au: sorry
[11:13:27] <kshishkov> Dark_Shikari: official Bink players can print some stats
[11:13:39] <Dark_Shikari> why not make it the goal to beat them?
[11:13:53] <Dark_Shikari> I would be willing to write x86 simd for you.
[11:13:58] <kshishkov> because there's no video decoder in FFmpeg
[11:14:07] <Dark_Shikari> ?
[11:14:08] <kshishkov> heh, even I can write that SIMD
[11:14:17] <Dark_Shikari> I meant, after you commit it
[11:14:18] <kshishkov> for x86/PPC/ARM
[11:14:24] <kshishkov> ah, OK
[11:14:36] <Dark_Shikari> the 8x8 -> 16x16 one actually has a very smart way of doing it
[11:14:40] <Dark_Shikari> movq mm0, [8x8 src]
[11:14:41] <Dark_Shikari> pshufb
[11:14:49] <Dark_Shikari> er, xmm0
[11:14:50] <Dark_Shikari> movdqa [16x16 dst], xmm0
[11:15:02] <Dark_Shikari> One non-loadstore op per line
[11:15:20] <kshishkov> yes, rather obvious
[11:15:20] <Dark_Shikari> hmm actually you don't even need that
[11:15:25] <Dark_Shikari> punpcklbw xmm0, xmm0
[11:15:40] <Dark_Shikari> a total of 32 ops
[11:15:46] <Dark_Shikari> might as well unroll the whole thing
[11:15:52] <kshishkov> try to optimize fill_block next ;)
[11:15:58] <Dark_Shikari> lol
[11:17:15] <Dark_Shikari> and of course unclamped put pixels
[11:17:15] <Dark_Shikari> etc
[11:17:21] <Dark_Shikari> you should really use the x264 simd strategy
[11:17:21] <Dark_Shikari> i.e.
[11:17:26] <Dark_Shikari> "this looks simdable, make a dsp function!"
[11:17:26] <kshishkov> still, I think it may be Bink audio that is slow
[11:17:55] <kshishkov> someone should organize dsputil
[11:18:10] <Dark_Shikari> probably a good idea at some point
[11:18:53] <pross-au> what do you mean 'organise' ?
[11:19:03] <Dark_Shikari> separate into codecs
[11:19:11] <Dark_Shikari> clean up
[11:19:13] <Dark_Shikari> make less messy
[11:19:17] <Dark_Shikari> port to yasm, remove inline asm
[11:19:22] <pross-au> dsputil.c is a big file to compile
[11:19:29] <kshishkov> yes
[11:19:33] <kshishkov> and very messy
[11:19:40] <_av500_> libavdsputil
[11:19:46] <Dark_Shikari> lol
[11:19:47] <pross-au> this little ppc takes a few minutes to compile it
[11:19:49] <iive> it would be better to be split by functional criteria
[11:19:59] <iive> e.g. dct at one place, mc at another
[11:20:02] <kshishkov> pross-au: try motionest.c ;)
[11:20:14] <_av500_> even better, it could be split into ffcc intrinsics
[11:20:17] <Dark_Shikari> well motion_est is hilarious
[11:20:32] <Dark_Shikari> it has single functions larger than x264's entire ME module
[11:20:33] <pross-au> kshishkov: that too
[11:20:47] <Dark_Shikari> I think motion_est is larger than x264's entire analysis module, doubled
[11:20:50] <Dark_Shikari> or tripled
[11:21:15] <iive> any volunteers?
[11:21:37] <Dark_Shikari> oh and of course it has functions larger than the L1 cache of any known cpu
[11:21:42] <kshishkov> iive: yes, right after they finish making good documentation for FFmpeg
[11:23:35] <_av500_> Dark_Shikari: i am sure intel can be convinced to 1mb l1
[11:23:42] <_av500_> to add
[11:23:44] * kshishkov thinks that maybe he can commit decoder even today. To start/end week on a bright side.
[11:24:25] <kshishkov> _av500_: ask them for 256 additional 512-bit SIMD registers while you're on it
[11:25:13] <_av500_> kshishkov: but then they put h264 decoder in chipsets now, so maybe not...
[11:25:38] <kshishkov> _av500_: ask them for HW Theora decoder then
[11:26:39] <kshishkov> so you can write in yasm " pdecodetheora m0, m1"
[11:28:33] <_av500_> kshishkov: i think the will rather put flash in hw :)
[11:29:03] <kshishkov> heh
[11:29:24] <pross-au> kshishkov: agree
[11:29:55] <Dark_Shikari> lol cisc
[11:30:04] <pross-au> i suspect there is a quality regression in the bink audio decoder
[11:30:08] <twnqx> rather... commit the mov/mp4 tag changes :D
[11:30:12] <pross-au> certain files sound poor wishy-washy here
[11:30:19] <kshishkov> here too
[11:30:23] <kshishkov> mostly DCT ones
[11:30:49] <pross-au> also i had a seeking patch for the demuxer
[11:30:55] <pross-au> *have
[11:31:31] <kshishkov> feel free to apply ;)
[11:31:57] <kshishkov> you can even register yourself as its maintainer
[11:36:07] <CIA-90> ffmpeg: stefano * r21935 /trunk/ffplay.c: Use the official FFmpeg spelling: "ffmpeg" -> "FFmpeg".
[11:43:50] <CIA-90> ffmpeg: stefano * r21936 /trunk/ (Makefile configure doc/ffprobe-doc.texi ffprobe.c Changelog): Add FFprobe tool.
[12:05:05] <elenril> kshishkov: i think length of output <= strlen(input)*2 for all characters
[12:05:24] <kshishkov> yes
[12:06:03] <elenril> good
[12:32:35] <mru> kshishkov: all latin1 characters map into <= 2 bytes of utf8
[12:33:08] <kshishkov> it was not matter of utf8 here, it was mostly the matter of output
[12:33:25] <kshishkov> for utf8 strlen() is always <= length of string
[12:33:44] <mru> >=
[12:33:59] <kshishkov> yes, the other way round
[12:47:30] <iive> strlen is always returning number of bytes, so it doesn't include the '\0' so it is always <
[12:49:25] <kshishkov> in this case length of string meant number of Unicode characters in that string, terminator is irrelevant
[12:51:49] <iive> :)
[12:52:08] <mru> "terminator is irrelevant" ... tell that to arnie
[12:53:27] <kshishkov> studios told him already
[12:53:49] <mru> so they did
[12:53:50] <kshishkov> last movie hasn't featured him
[12:54:06] <mru> only as CGI
[12:55:14] <Kovensky> he probably isn't convinced yet
[12:55:28] <Kovensky> maybe it was just because they thought he was too busy as the governator, etc
[12:57:45] <iive> i just pretend that the last 2 movies doesn't exist.
[12:58:34] <Kovensky> heh
[12:58:45] * Kovensky never watched any terminator other than the 2nd one
[12:58:50] <kshishkov> reminds me of XKCD saying that Matrix was a good movie
[12:59:16] <Kovensky> Matrix was, too bad they never made a sequel
[12:59:32] <iive> they did, it's called animatrix
[12:59:37] <twnqx> :D
[12:59:40] <Kovensky> o rly
[12:59:46] * elenril <3 animatrix
[12:59:48] <Kovensky> I thought those were side stories
[13:00:16] <kshishkov> Kovensky: I also saw the third. It featured female terminator able to whistle as modem (and consequently dialup via cellphone). That's all I remember about it.
[13:01:29] <mru> t3 was the worst of them
[13:01:30] <iive> actually in 3 and 4'th movies the terminator characters are quite good hit. too bad nothing else is on their level.
[13:01:43] <mru> t4 is better
[13:01:49] <mru> t2 still the best
[13:01:59] <iive> kshishkov: you forgot the automatic breast augmentation!
[13:01:59] <mru> regular version, not director's cut
[13:02:07] <mru> iive: :-)
[13:02:46] <kshishkov> iive: I don't remember that. Nor don't want really.
[13:03:41] * kshishkov likes movies directed by Olle Hellbom
[13:08:10] <iive> i'm not sure i've watched anything from him. He is swedish , isn't he?
[13:08:36] <kshishkov> of course
[13:08:39] <mru> name sounds swedish
[13:08:57] <kshishkov> all in Wikipedia ;)
[13:10:05] <mru> I've found something disturbing in lavc
[13:10:12] <mru> shifts by negative amounts
[13:10:39] <kshishkov> where?
[13:10:42] <twnqx> that... works?
[13:10:47] <mru> libavcodec/bitstream.c:164
[13:10:51] <mru> libavcodec/golomb.h:172
[13:11:15] <mru> I hacked an emulator to trap on invalid shifts
[13:12:43] <mru> the first one is mostly harmless
[13:12:47] <mru> but dangerous still
[13:12:54] <mru> the result of that shift is never used
[13:13:55] <mru> but a negative shift has undefined behaviour
[13:14:13] <mru> it's perfectly valid for it to crash
[13:17:27] <DonDiego> h264.h:816: warning: unused variable 'mb_xy'
[13:17:33] <DonDiego> how lame is that, guys?
[13:17:54] * mru blames michael
[13:18:04] <mru> he seems to build with warnings off
[13:18:13] <mru> or else his eyes should be bleeding by now
[13:19:37] <DonDiego> well, somebody just fix this..
[13:19:49] <mru> you have write access...
[13:19:50] <DonDiego> Yuvi: you also introduced a few fresh warnings .)
[13:22:45] * kshishkov prepares for conducting lavc minor bump
[13:23:12] <Kovensky> -w ftw
[13:23:25] * Kovensky usually builds his own code with -pedantic
[13:23:32] <mru> that's silly
[13:23:39] <mru> that's bordering on lint-level whinging
[13:24:43] <DonDiego> -Werror has done wonders to the projects i work on at uni..
[13:24:45] <twnqx> -Wall -Werror ftw
[13:25:00] <mru> -Wextra too
[13:25:20] <twnqx> try to compile a gentoo with it and start crying :P
[13:25:29] <mru> -Werror can be useful if you always use the same compiler
[13:25:50] <mru> if the code will be built with lots of different compilers, as released code will, it's a very bad idea
[13:25:51] <twnqx> for me it meant "omg, lots of things changed with gcc 4... let's go to work
[13:26:18] <twnqx> anyway. still the trivial fixes for mp4/mov tagging aren't in? :X
[13:26:19] <mru> like when gcc4 started throwing warnings on 95% of all str*() calls
[13:26:37] <Kovensky> why did it do that?
[13:26:46] <mru> pointer target types differ in signedness
[13:27:19] <mru> it really shouldn't do that for char
[13:27:26] <mru> since signedness of char is undefined
[13:27:38] <mru> or unspecified for the pedants
[13:29:02] <kshishkov> the bomb has been dropped
[13:29:45] <CIA-90> ffmpeg: kostya * r21937 /trunk/ (11 files in 3 dirs): Bink video decoder
[13:30:15] * Kovensky waits for the boom
[13:30:33] <mru> 5
[13:30:34] <mru> 4
[13:30:36] <mru> 3
[13:30:38] <mru> 2
[13:30:39] <mru> 1
[13:30:42] <mru> BOOM
[13:31:03] <Kovensky> oh you
[13:31:19] <kshishkov> Compn: you around?
[13:31:36] <iive> i thought fate needs more time.
[13:32:17] <kshishkov> it's Mike's problem
[13:33:12] <DonDiego> bink..
[13:33:36] <kshishkov> s/bink/yet another obscure game codec/
[13:34:07] <mru> let's celebrate: http://imagebin.ca/view/x6Tdu5C.html
[13:34:14] <kshishkov> DonDiego: http://shishkov.homeunix.net/bink-mplayer.patch
[13:35:05] <thresh> compiles here fine on osx
[13:35:17] <thresh> no boom at all :(
[13:35:30] <DonDiego> kshishkov: what are you waiting for, commit it :)
[13:36:11] <kshishkov> ok, wait a minute
[13:40:27] <pentanol> hi, I need suggestion so as to open rtsp stream  AVInputFormat *fmt = NULL;  pd->filename = "rtsp://mystream"; av_probe_input_format(pd, 0);  then do what? rtsp_read_play()?
[14:27:58] <KotH> am i correct that the power density spectrum is the square of the fourier transformation of a signal?
[14:31:28] <mt> KotH: Yes I believe so.
[14:33:15] <KotH> hmm... no it's not... it's only in the case that the fourier transformation is real
[14:36:04] <mt> Ah. I was wrong then. I read something about it months ago but it didn't mention anything about the transform being complex or real. But iirc, all the transforms on that page were real so probably why it wasn't mentioned.
[14:38:25] <KotH> well.. it's been ages since i last had a look at signal theory, so i forgot most of it
[14:40:22] <kshishkov> KotH: lucky you, you have something to forget
[14:40:34] <KotH> ^^'
[14:40:44] <KotH> i forgot too many things already
[14:41:05] * elenril thought projects L^2(R)->L^2(R)
[14:41:11] <elenril> *FT
[14:42:27] <CIA-90> ffmpeg: vitor * r21938 /trunk/libavformat/idcin.c: Fix memory leak for truncated packets in idCin demuxer
[14:43:43] <Vitor1001> kshishkov: so avcodec_check_dimensions() is needed after all?
[14:45:13] <kshishkov> Vitor1001: it seems to be still present in other decoders, so unless someone can guarantee it's safe to remove I'll leave it there as well
[14:46:17] <Vitor1001> kshishkov: looking at libavcodec/utils.c it looks like avcodec_encode_video() check it.
[14:46:33] <Vitor1001> not very important anyway
[14:46:41] <kshishkov> it's decode
[14:46:57] <Vitor1001> oops, also in  avcodec_decode_video2()
[14:47:01] <Vitor1001> well, have to go now
[14:47:13] <Vitor1001> Congrats for the commit \o/
[14:47:29] <kshishkov> I had no problems committing it :P
[14:51:30] <kshishkov> ok, looks like the rest of FFmpeg 0.6 pendiang features are not of my concern (except maybe AMR-NB)
[15:10:15] <Honoome> elenril: yes?
[15:10:51] <kshishkov> Honoome: IIRC you've promised him to apply his patch at weekend
[15:11:27] <Honoome> kshishkov: I actually sent a patch for review, I should be able to apply part of it if I can solve Michael's rebus :P that I will do after a coffee
[15:12:51] <CIA-90> ffmpeg: mru * r21939 /trunk/libavcodec/bitstream.c:
[15:12:51] <CIA-90> ffmpeg: Avoid negative shifts in build_table()
[15:12:51] <CIA-90> ffmpeg: A shift by a negative amount has undefined behaviour. Even though
[15:12:51] <CIA-90> ffmpeg: the result of this shift is never used, the shift itself could
[15:12:51] <CIA-90> ffmpeg: cause an exception of some kind.
[15:12:52] <CIA-90> ffmpeg: mru * r21940 /trunk/libavcodec/bitstream.c: indent
[15:21:18] <elenril> Honoome: i saw your comment here https://bugs.freedesktop.org/show_bug.cgi?id=17431
[15:21:31] <elenril> did you figure out how to fix it?
[15:21:50] <elenril> kshishkov: no, nobody promised me to apply my patches
[15:21:56] <elenril> you're all being lazy slackers =p
[15:21:57] <Honoome> elenril: sorta… my solution was to simply change SDL to unbundle the library :)
[15:22:16] <elenril> so the problem is in SDL?
[15:22:24] <kshishkov> elenril: good engineer must be lazy. So we are both lazy and proud of it!
[15:22:33] <kshishkov> (at least some of us)
[15:22:37] <Honoome> yeah they bundle an old copy of Xxf86vm library and that is causing the problem
[15:22:50] * elenril facepalms
[15:23:04] <mru> still should not make the server crash
[15:23:20] * Kovensky is reminded of sisimedia
[15:23:27] <mru> a client should never be able to crash the server
[15:23:43] <Kovensky> not only it comes with a broken configure script but it uses headers that were deprecated on 7.5 :(
[15:23:57] <Honoome> mru: yeah I know, but me and Xorg devs don't get much along
[15:24:13] <mru> heh
[15:24:42] <Honoome> I went into a pissing contest with Eric Anholt when I worked on Gentoo/FreeBSD because I insisted into making Xorg work with FreeBSD by default
[15:24:48] <Honoome> he didn't care because ports had all the patches needed…
[15:25:19] <mru> typical bsd attitude
[15:25:24] <Kovensky> and then some people complain that downstream is uncooperative and keep all their patches to themselves...
[15:25:44] <Honoome> Kovensky: he's both _up_ and _down_ stream for Xorg/FreeBSD…
[15:26:03] <Kovensky> D:
[15:26:24] <mru> a vortex
[15:26:36] <Honoome> when they split Xorg, FreeBSD was still on the old merged one and Gentoo/FreeBSD was trying to move to split (with the rest of Gentoo)… I reported bugs and his answer was “untrue, Xorg works perfectly on FreeBSD… use monolithic”…… d'oh! >_<
[15:26:59] <mru> I actually liked the monolithic one better
[15:27:04] <mru> it worked
[15:27:21] <mru> the split one is version hell all over the place
[15:27:23] <Kovensky> <@mru> typical bsd attitude <-- not really, it isn't related to freebsd other than using the same kernel
[15:27:41] <Honoome> Kovensky: what are you referring to?
[15:27:47] <Kovensky> gentoo/fbsd
[15:27:47] <mru> it's typical bsd attitude to shove the ports tree full of patches and refuse to talk
[15:27:57] <Honoome> Kovensky: gentoo/freebsd is not the same as debian's gnu/kfreebd
[15:28:08] <Honoome> Kovensky: gentoo/freebsd uses same kernel, same libc, same userland commands
[15:28:28] <Kovensky> only with portage instead of ports?
[15:29:01] <Honoome> yep
[15:29:13] <Honoome> [and unbundled packages that are otherwise part of freebsd base system, such as libarchive]
[15:30:09] <CIA-90> ffmpeg: diego * r21941 /trunk/libavcodec/h264.h:
[15:30:09] <CIA-90> ffmpeg: Remove unused variable, fixes warnings of the type:
[15:30:09] <CIA-90> ffmpeg: libavcodec/h264.h:816: warning: unused variable `mb_xy'
[16:17:04] <mru> oh no, fate is going yellow again
[16:17:29] <kshishkov> somebody "fixed" idcin?
[16:17:34] <mru> yep
[16:24:58] <CIA-90> ffmpeg: reimar * r21942 /trunk/ffprobe.c:
[16:24:58] <CIA-90> ffmpeg: Avoid using log2, it is not available everywhere.
[16:24:58] <CIA-90> ffmpeg: Should fix compilation on FreeBSD.
[16:34:21] * Kovensky wonders why it isn't in fbsd anyway
[16:34:28] <Kovensky> not log2 or log2f <_<
[16:35:14] <kshishkov> because they are idiots
[16:36:39] <mru> they are bsd
[16:36:43] <mru> they don't do modern stuff
[16:37:06] <kshishkov> they always did it their way even in times of System V
[16:37:32] <mru> sysv and bsd forked a long time back
[16:37:43] * elenril wonders how did asfenc ever work
[16:38:22] * kshishkov considers SysV to be the True UNIX
[16:38:48] <mru> Tru64 is of course the true one
[16:40:11] <kshishkov> well, since it's from DEC...
[16:41:47] * Kovensky pats elenril
[16:42:27] * elenril still wonders why he's wasting his free time on this
[16:42:33] <elenril> it's not like i use asf
[16:53:51] <KotH> elenril: masochism
[16:55:31] <elenril> http://tvtropes.org/pmwiki/pmwiki.php/Main/BileFascination
[16:59:12] <Honoome> elenril: rotfl
[17:09:36] <elenril> what's the best way to check if an array is all zero?
[17:09:49] <kshishkov> flag and loop
[17:10:29] <elenril> only 5 pointers in it
[17:11:04] <kshishkov> do | and check the result
[17:11:05] <mru> array of 5 pointers?
[17:11:19] <elenril> yes
[17:11:37] <mru> what kshishkov said
[17:11:52] <elenril> hmm, i hoped there'd be some nice black magi
[17:11:54] <elenril> c
[17:11:54] <mru> unless you're on a 64-bit machine with 32-bit pointers and the array has suitable alignment
[17:12:47] <mru> probably not worth the trouble though
[17:13:28] <elenril> right
[17:40:41] * justlooking always thought the AT&T Unix System V Release 4 for the Amiga computer family was the true UNIX ? being the First ONE  ordinary end users got access to on 1990 ...http://en.wikipedia.org/wiki/Amiga_Unix
[17:55:50] <kierank> So, i need to use half sized KBD windows. Is it worth me writing an "ff_half_kbd_window_init" function or just using the current one then overwriting half of the output?
[18:03:05] <CIA-90> ffmpeg: vitor * r21943 /trunk/ (6 files in 3 dirs):
[18:03:05] <CIA-90> ffmpeg: AMR-NB floating-point based decoder.
[18:03:05] <CIA-90> ffmpeg: Code produced during SoC by Robert Swain and Colin McQuillan.
[18:04:30] <mru> congrats
[18:04:44] * mru waits for fate to turn red
[18:04:48] <iive> \o/
[18:05:39] <Vitor1001> congrats to Robert and Colin, who did the bulk of the work
[18:06:24] <Vitor1001> In these last few days, the pipeline of important things remaining to get committed got significantly shorter :D
[18:07:34] <Vitor1001> FFprobe, WMAVoice, AMR-NB, ALS, Bink...
[18:08:28] <kshishkov> yes, now we have only bunch of lossless audio codecs, some Windows Media codecs, Indeo 4, some Apple codecs and whatever
[18:13:59] <Vitor1001> I agree, but those are not really low-hanging fruits
[18:14:00] <mru> VP8
[18:14:09] <Vitor1001> LAME hostile takeover ;)
[18:14:17] * Vitor1001 runs away from the flames
[18:14:17] <kshishkov> mru: I said "whatever"
[18:15:02] <Vitor1001> Is any working code written for indeo 4?
[18:15:03] <kshishkov> Vitor1001: well, if you are goingto undertake that task, maybe why not? Maybe the same devs will improve our MP2 encoder too ;)
[18:15:11] <kshishkov> yes
[18:15:27] <kshishkov> but it's only in autor posession
[18:15:55] <Vitor1001> lame is a project that is in a pretty mature stage, also in the sense of not evolving very fast
[18:16:13] <Vitor1001> but I don't have time for that :(
[18:16:25] <Vitor1001> and I don't think it would need to be hostile...
[18:19:29] <kshishkov> still, FFmpeg can benefit from good lossy audio encoder or two
[18:25:10] <Vitor1001> is a good psychoacoustic model all ffaac is missing?
[18:25:23] <kshishkov> no
[18:25:27] <Vitor1001> :(
[18:25:42] <kshishkov> it's mostly good rate control and good bit allocator
[18:28:26] <peloverde> I think I know how to make the coef coder not suck
[18:28:26] <Compn> someone take out amr from wiki yet ?
[18:28:41] <peloverde> I still have no intention of touching the psymodel
[18:28:46] * kshishkov looks at Compn
[18:29:25] <kshishkov> peloverde: can you say technical reasons why coef coder suck?
[18:30:17] <Vitor1001> Compn: yes
[18:30:25] <peloverde> Well there are three of them and it's slow
[18:31:29] <kshishkov> and none of them works as supposed I fear
[18:31:52] <peloverde> In general it makes more sense to build a codeword than find it in the table
[18:32:32] <kshishkov> ah
[18:32:44] <peloverde> All the shit in the spec when you are supposed to divide and subtract that makes writing a decoder suck (that we do with tables in our decoder) can be done as adds and multiplies in the encoder
[18:33:39] <peloverde> Also we can pick up quite a bit of speed by templating
[18:34:14] <CIA-90> ffmpeg: stefano * r21944 /trunk/ (doc/ffprobe-doc.texi ffprobe.c):
[18:34:14] <CIA-90> ffmpeg: Add support to an option -f which forces the format to use for opening
[18:34:14] <CIA-90> ffmpeg: the probed multimedia resource.
[18:34:35] <peloverde> we waste a lot of time checking if things are unsigned and what size they are and if escapes are possible
[18:35:25] <peloverde> I have a patch to inline the coder into 6 cases: 0, 2U, 2S, 4U, 4S, and ESC
[18:35:52] <peloverde> where 2U is unsigned pair for instance
[18:37:34] <Compn> ah thats what still needs to be done
[18:37:43] <Compn> add ffamr to mplayer codecs.conf :)
[18:38:28] <kshishkov> ffamrnb for now
[18:38:34] <Compn> right
[18:40:39] <peloverde> Also money would help, PS has money attached, if google managed to find $100 million for on2 and wants an AAC encoder they are going to have to find a few dollars somewhere
[18:41:01] <kshishkov> :)
[18:41:48] <peloverde> I don't want $100 million but I'd like to be able to pay my rent, buy a new PC, and afford a few meals out
[18:42:11] <kshishkov> a thousand bucks then
[18:42:13] <iive> peloverde: submit your CV to google.
[18:43:11] <Compn> kshishkov : btw have you thought about moving to usa/canada? the northern parts are cold ...
[18:43:16] <peloverde> I learned when I applied to M$ that if you have a specific skillset and are applying to a big company you get lost in the shuffle
[18:43:18] <Compn> moving/finding job there
[18:43:50] <kshishkov> Compn: too far away from Sweden
[18:44:02] <iive> peloverde: you have taste for evil!
[18:44:03] <Compn> peloverde : yeah, but with google you can ask skal for help in the shuffle
[18:44:14] <Honoome> peloverde: you'd end up getting asked a lot about python during the interview, for google development positions
[18:45:19] <kshishkov> and probably VB for M$
[18:46:08] <janneg> probably .net now
[18:46:18] <kshishkov> vb#.net
[18:46:35] <peloverde> microsoft asked a little about .net, they mostly asked about datastructures and whatnot
[18:46:54] <peloverde> and not a single question about signal processing
[18:46:57] <kshishkov> and you were overqualified
[18:47:24] <Compn> microsoft just wants to throw more developers at IE or windows api and hope someone can work on either without going insane :P
[18:47:26] <kshishkov> peloverde: look at me, I have not studied signal processing at all
[18:48:11] <peloverde> I was in thoery interving for a position in the Windows Media group but I think all I got was a screen interview for comp scis
[18:48:53] <kshishkov> do you think their WM division knows _anything_ about signal processing?
[18:49:09] <peloverde> Rico malvar wrote a decent book
[18:49:12] <kierank> presumably gary sullivan does
[18:49:18] <kshishkov> IIRC their audio codecs used Intel implementation of FFT
[18:49:34] <kshishkov> Malvar - yes, known transforms guy
[18:49:39] * elenril wonders if GET_UTF8 is broken
[18:50:22] <iive> peloverde: did you got some tick puzzle questions?
[18:50:32] <peloverde> no puzzle questions
[18:50:46] <peloverde> puzzle questions are pretty out of vogue these days
[18:52:38] <peloverde> I worry that I would get a generic screening interview at google and get tossed aside, I also worry I would get stuck on something internal (like when was the last time skal submitted any code)
[18:53:14] <kshishkov> not that he contributed much before that
[18:53:59] <DonDiego> geez
[18:54:02] <DonDiego> amr-nb...
[18:54:02] <iive> were you soc student?
[18:55:14] <iive> you can write you've completed a number of tasks in soc, it should be enough to skip the first round.
[18:55:25] <peloverde> I was never SoC
[18:55:50] <iive> even as ... teacher?
[18:56:21] <peloverde> never, this will be my first year as a mentor if i cna find a student
[18:57:39] <elenril> anybody has a working knowledge of utf8?
[18:57:45] <peloverde> FFmpeg started participating in SoC when I graduated, and then I worked for a a semiconductor firm for two years
[18:58:55] <peloverde> Pretty much I picked up all my signal processing theory in school and all my multimedia knowledge at work
[18:58:56] <kshishkov> elenril: what exactly?
[18:59:04] <iive> elenril: it would be much easier if you ask specific questions, instead of asking for asking.
[18:59:31] <kshishkov> peloverde: you're lucky. Neither school nor university gave me any real education
[18:59:34] <elenril> kshishkov: GET_UTF8 seems to return zeros for non-ascii characters
[18:59:49] <elenril> i'd like someone to check it for signs of obvious brokenness
[19:00:10] <kshishkov> I'll look right now
[19:00:18] <iive> be sure you read chars as unsigned. is that ffmpeg code?
[19:00:48] <kshishkov> elenril: how you use it?
[19:01:45] <elenril> http://pastebin.ca/1804892
[19:03:16] <elenril> works fine with latin strings
[19:05:19] <iive> elenril: what is the type of q?
[19:05:22] <kshishkov> hmm, even for non-ascii strings it should set ch to the first character
[19:05:47] <elenril> iive: char*
[19:05:48] <kshishkov> can't you print q[0] and q[1] to verify input?
[19:06:16] <iive> elenril: try with uint8_t
[19:06:39] <kshishkov> indeed
[19:09:39] <elenril> heh, seems to work
[19:09:46] <elenril> why would it break just because of this?
[19:09:54] <kshishkov> of course
[19:10:04] <kshishkov> input value range differs
[19:10:49] <iive> 0x80 is extended to 0xffffff80
[19:11:37] * elenril should really learn C soon
[19:11:54] <kshishkov> or go in management
[19:12:41] <elenril> =p
[19:12:54] <elenril> i'm a physicist, not a programmer
[19:13:49] <kshishkov> yes, I saw what people like you can do. Hasn't got through all Half Life game though.
[19:14:25] <elenril> lol
[19:14:26] <iive> i bet you haven't got to the better half yet.
[19:33:37] <CIA-90> ffmpeg: cehoyos * r21945 /trunk/libavcodec/h264.c:
[19:33:37] <CIA-90> ffmpeg: Remove unused variable mb_xy.
[19:33:37] <CIA-90> ffmpeg: Patch by avcoder, ffmpeg gmail
[19:42:17] <ramiro> elenril: is there any reason to use the external amr libraries instead of ffamr-nb for decoding? as in is there any feature missing from ffamr-nb?
[19:42:27] <ramiro> elenril: oops, that wasn't meant for you
[19:42:48] <kshishkov> probably not
[19:42:54] * elenril agrees ;)
[19:42:59] <kshishkov> and IIRC it had not special features
[19:43:26] <jai> hello ramiro
[19:43:34] <ramiro> jai: hi, I just got your reply =)
[19:43:55] <ramiro> I just skipped over that thread, let me check.
[19:44:42] <jai> ramiro: sure, np :)
[19:45:35] <elenril> kshishkov: what about reverse conversion utf16->utf8. how much space would that need?
[19:45:55] <kshishkov> twice as well ;)
[19:45:59] <Kovensky> ohi ramiro, still dodging teh floods? =p
[19:46:11] <kshishkov> in char size - 2 bytes, out char size 1-4 bytes
[19:48:20] <elenril> thanks, thought as much
[19:50:14] <ramiro> Kovensky: what floods? I'm not in sao paulo.
[19:50:15] <elenril> everybody rejoice, now asfenc/dec can read/write metadata correctly!
[19:50:28] <elenril> (not that anybody cares about asf)
[19:50:32] <twnqx> what about mov/mp4?
[19:50:33] <kshishkov> *bored* yaaaay
[19:50:50] <Kovensky> ramiro: I meant the state
[19:50:58] <elenril> mov is :effort:
[19:51:10] <Kovensky> ramiro: where are you anyway
[19:51:28] <ramiro> jai: I'm no parser expert (in fact I still don't understand most of it), but it seems that if ff_combine_frame is called, buffer is allocated, and so it must be freed at some point. so I guess it's ok if it works...
[19:51:45] <jai> ramiro: indeed
[19:51:53] <ramiro> Kovensky: florianopolis
[19:52:05] <Kovensky> ramiro: for some reason I thought you were in SP
[19:52:06] <Kovensky> lol
[19:52:44] <ramiro> Kovensky: no, I don't like that place. too many people. I only go there for concerts =)
[19:53:08] <Kovensky> heh
[19:53:18] <Kovensky> btw ramiro, I added a paypal button to my page
[19:53:26] <Kovensky> two weeks and $0 ._.
[19:53:47] <kshishkov> swap it randomly with download link ;)
[19:53:51] <elenril> lol
[19:53:53] <ramiro> lol
[19:53:59] <jai> heh
[19:54:09] <Kovensky> lol
[19:54:18] * kshishkov received much more in donations
[19:54:22] <elenril> that's because you didn't make new builds for so long
[19:54:39] <ramiro> Kovensky: well, it's kind of hidden
[19:55:47] <ramiro> I once had a script that generated a random amount of money (from 2 to 20$) and put it in the donation link, with a "refresh" button. people thought it was funny and I got a bunch of 2.something donations...
[19:57:33] <ramiro> Kovensky: did you add dxva2api.h to build ffmpeg?
[19:58:00] <Kovensky> ramiro: yes, ffmpeg is built with fxva support; but mplayer can't use it :(
[19:58:03] <Kovensky> dxva*
[19:58:27] <ramiro> hmm, oh well, I don't know about mplayer...
[19:58:42] <peloverde> I added a donate link and got nothing too
[20:00:18] <kshishkov> Kovensky: another reason to dith MPlayer for FFplay ;)
[20:00:51] <kshishkov> peloverde: I found that having popular blog is better than "Donate" on less popular one ;)
[20:00:55] <elenril> ffplay - sdl -meh
[20:01:10] <kshishkov> elenril: libvo support is welcome ;)
[20:01:10] <Vitor1001> Who
[20:01:20] <ramiro> kshishkov: after michael's rdft output, ffplay beats the hell out of WMP's visualization =)
[20:01:35] <elenril> kshishkov: :effort:
[20:01:38] <peloverde> I just put it directly on my git page. I haven't blogged in years
[20:01:45] <Vitor1001> err, who said valgrind devs never fix bugs?
[20:01:46] <Vitor1001> ;)
[20:01:47] <Vitor1001>  https://bugs.kde.org/show_bug.cgi?id=210264
[20:01:51] <ramiro> that definitely got us a couple of steps closer to world domination.
[20:01:53] <peloverde> ramiro, now all we need is an sse rdft
[20:02:11] <Honoome> given I freed myself from Gentoo today I might be finally able to look into getting a Planet Multimedia to aggregate our blogs :P
[20:02:22] <kshishkov> and altivec
[20:02:28] <Kovensky> did SDL ever fix their static linking bug?
[20:02:40] <ramiro> Vitor1001: elenril said he's a physicist. didn't you graduate in physics too?
[20:02:48] <Honoome> Kovensky: did SDL fix _anything_ in the past five years? :P
[20:03:01] <ramiro> Honoome: no, but they indent to charge for 1.3
[20:03:13] <ohsix> freed from gentoo again? :O
[20:03:27] <Vitor1001> ramiro: Trying to finish my phd in physics
[20:03:28] <Kovensky> 1.3 was WIP back when I first heard of SDL
[20:03:31] <Kovensky> that was almost 10 years ago
[20:03:55] <kshishkov> Kovensky: same as FFmpeg 0.5, I presume?
[20:03:58] <Honoome> they are trying to reuse part of the Duke Nukem Forever code
[20:05:03] <Kovensky> explains everything
[20:05:04] <ramiro> Kovensky: another idea is to make your builds slightly broken, and suggest that people may sponsor better builds...
[20:05:06] * jai has a local ffplay hack to output to libvo gl2
[20:10:30] <ohsix> hm, cant ffmpeg do virtualdub stype dubbing when one track is cbr and one is vbr?
[20:12:49] <elenril> btw anybody wants to apply a few patches for me?
[20:13:18] <Kovensky> more meta stuff? =p
[20:13:24] <elenril> ofc
[20:13:50] <elenril> metadata is important!
[20:14:31] <Kovensky> yes it is
[20:14:49] <Kovensky> more importantly, I want a way to be able to export / import metadata
[20:15:03] <Kovensky> it seems that I may end up having to write my own thing for doing that >_>
[20:15:26] <twnqx> just define a metadata XML
[20:15:49] <elenril> Kovensky: what to you mean by that?
[20:16:10] <Kovensky> dump metadata from a file to a text file
[20:16:15] <twnqx> he wants to be able to import/export metadata standalone
[20:16:18] <Kovensky> load the dumped metadata into another file
[20:16:27] <elenril> write a muxer/demuxer
[20:16:30] <elenril> should be easy
[20:16:46] <Kovensky> o_O
[20:16:54] <Kovensky> how would that work
[20:16:55] * elenril slaps twnqx for using the X-word here
[20:17:13] <Honoome> could be worse
[20:17:34] <Kovensky> how worse
[20:17:37] <mru> could be xiph-specified xml
[20:17:47] <mru> double-x
[20:17:52] <Honoome> haha :)
[20:18:00] <Honoome> I was thinking of aXML though
[20:18:19] <Yuvi> mru: http://wiki.xiph.org/XMLEmbedding ?
[20:19:09] <elenril> Kovensky: a muxer consisting of only write_header()
[20:19:29] <Kovensky> hmm
[20:19:30] <elenril> i supposed even i could write it in less than one afternoon
[20:19:40] <Kovensky> lol
[20:20:01] <Kovensky> I was thinking of using YAML for the export ._.
[20:20:30] <elenril> why do you need anything like that
[20:20:48] <Kovensky> because uploading tags is much faster than uploading .flac
[20:21:13] <elenril> just use a text file with tag=value\n
[20:21:40] <Honoome> ini-compatible key-value files! I love that :D
[20:21:48] <Kovensky> lol
[20:22:02] <mru> but without [this]
[20:22:06] <Kovensky> key: value\n is better :D
[20:22:13] <Honoome> mru: that's true… but still… :)
[20:22:28] <Honoome> you can use the same code to parse both in most cases and there are quite a few libraries in almost every language that support it
[20:22:29] <mru> key $arbitrary_separator value
[20:23:04] <Honoome> [and it actually says a lot, about Microsoft (or whoever used them first) having had a not half-bad idea at the time…]
[20:23:29] <jai> json perhaps
[20:23:43] * mru prefers frddy
[20:23:44] * Kovensky prefers YAML to JSON
[20:23:54] <jai> hmm
[20:23:58] * Kovensky pats Dark_Shikari
[20:24:18] <mru> jaison
[20:24:46] <elenril> lol
[20:25:10] <jai> :)
[20:25:52] * elenril should learn how to use git send-email
[20:26:20] <jai> man page isn't useful?
[20:26:32] <elenril> :effort:
[20:26:43] <mru> less than sending patches manually
[20:27:06] <elenril> btw is there anything better than mutt?
[20:27:11] <Honoome> git send-email --to=ffmpeg-devel at .. origin/master..
[20:27:35] <mru> elenril: matter of taste
[20:27:38] <Honoome> or if it's just the one last patch use -1
[20:27:39] <elenril> i'm starting to hate it
[20:27:40] * mru prefers cat
[20:27:59] <elenril> http://en.wikipedia.org/wiki/Cat ?
[20:28:03] <mru> but if you're a dog person...
[20:28:04] <Kovensky> what about tac | tac
[20:28:16] <Honoome> mru: “emerge dog”
[20:28:35] <Honoome> and I'm not even kidding :
[20:28:35] <Honoome> :P
[20:28:41] <mru> I know
[20:28:55] <twnqx> anyone here with an idea for a decent spice simulator for windows?
[20:29:10] <Kovensky> to simulate what
[20:29:16] <Honoome> I use Evolution anyway… and mutt for sending automated mails
[20:29:18] <elenril> http://notmuchmail.org/ this looks pretty interesting
[20:29:21] <elenril> anybody tried it?
[20:29:24] <Honoome> Kovensky: peppermint, sage, basil…
[20:29:41] <mru> twnqx: doesn't spice work in cygwin?
[20:29:53] <twnqx> i want a gui to click the pieces
[20:29:55] <twnqx> :P
[20:30:04] <twnqx> otherwise i'd just use spice-ng, fast and nice
[20:30:06] <jai> elenril: give sylpheed a try if you prefer GTK UIs
[20:30:27] * mru used to run some bizarre patchwork of spice and matlab
[20:30:40] <elenril> jai: i prefer something that runs in screen ;)
[20:30:54] <Honoome> elenril: emacs and vm :D
[20:30:58] <peloverde> twnqx, LTSpice has a nice gui
[20:30:59] <mru> emacs and gnus
[20:31:00] <elenril> >emacs
[20:31:00] <Honoome> [it's actually not half bad]
[20:31:06] <mru> gnus >>>> vm
[20:31:13] <Honoome> mru: gnus is not much for mail as much for mailing lists
[20:31:15] * elenril uses vim
[20:31:23] <mru> gnus is for everything
[20:31:34] <Kovensky> gnus are*
[20:31:36] * Kovensky ducks
[20:31:44] <twnqx> peloverde: will give it a shot, thanks
[20:31:55] * Honoome goes to the espresso machine
[20:32:02] * mru suddenly has flashbacks of pspice
[20:32:13] * Kovensky still wonders what spice is
[20:32:28] <Honoome> Kovensky: cinnamon! :P
[20:32:38] <peloverde> Kovensky, a circuit simulator
[20:32:52] <Honoome> peloverde: oh shush I was having so much fun :(
[20:32:55] <mru> there's spice in the kitchen, spice in the lab, and spice in dune
[20:33:01] <Kovensky> I remember it now D:
[20:33:06] * Kovensky has used it before, a long time ago
[20:33:23] * mru had curry for dinner the other day...
[20:33:34] * Kovensky does not like spicy food
[20:33:43] <mru> don't come to my house then
[20:33:44] * twnqx does sometimes
[20:34:30] <Kovensky> though the last time I tried spicy food was back when I was 10
[20:34:31] <Kovensky> lol
[20:36:38] <mru> now take your electronics simulator, enter schematics for a computer, run a simulator on it, enter schematics for a computer...
[20:37:20] <_av500_> a drug?
[20:38:14] * _av500_ finds his terminal on drugs
[20:38:33] <mru> http://www.youtube.com/watch?v=Rw8gE3lnpLQ
[20:41:09] <elenril> yo dawg
[20:47:59] <CIA-90> ffmpeg: mru * r21946 /trunk/libavutil/mathematics.h: More accurate value for log2(1)
[20:49:15] <elenril> more accurate value of zero?
[20:50:45] <mru> zero is the most important number, got to get it right
[20:51:33] * mru blames 0 and ) being on the same key
[20:52:28] * Kovensky never knew log2(1) == 0
[20:52:49] <mru> log_everything(1) == 0
[20:52:49] * elenril thinks Kovensky needs to repeat his math class
[20:53:00] <kierank> lol
[20:53:25] * Kovensky also thinks so
[20:54:04] <Honoome> mru: beside log_0(1) iirc?
[20:54:23] <mru> everything == positive numbers
[20:55:23] <twnqx> hm
[20:55:33] <twnqx> my circuit approaches -9GV
[20:55:42] * Honoome waits for Baptiste to check his patch … should solve the Rebus
[20:55:42] <twnqx> something.. is quite wrong here
[20:55:55] * elenril throws EUNDEFINED at Honoome 
[20:56:04] <Honoome> elenril: uh? :P
[20:56:05] <mru> ERANGE
[20:56:41] <Honoome> ah log_0 ^^ I thought you referred to Baptiste, since it was the metadata patch
[20:57:19] <elenril> well he's not here either
[20:57:27] * Kovensky throws ENOTIMPLEMENTED at elenril
[20:57:57] * elenril didn't see him here for more than a week
[20:58:11] * Kovensky misses his onjoin script
[20:58:21] <mru> hi guys
[20:58:37] <Kovensky> it's not the same thing :(
[21:00:24] <bcoudurier_> hi guys
[21:00:42] <mru> better?
[21:00:46] <elenril> sounds fake for some reason
[21:00:49] * elenril wonders why
[21:01:02] <elenril> probably the color
[21:01:51] <Honoome> elenril: it was the wrong shade of invisible pink?
[21:02:27] <elenril> no, my nick-coloring script colors real bcoudrier blue
[21:02:33] <elenril> this fake one was yellow
[21:02:51] <Honoome> for me it was green
[21:03:02] <elenril> you should fix it then ;)
[21:03:22] <Kovensky> for me he was yellow too
[21:03:23] <Kovensky> lol
[21:03:39] <elenril> Honoome: see -- your script is broken
[21:03:54] * Honoome has no script, this is Quassel!
[21:04:18] * Kovensky doesn't even try quassel precisely for that reason
[21:04:30] * Kovensky is addicted to perl scripting
[21:04:37] * elenril is addicted to screen
[21:05:09] <Honoome> I'm addicted to Gentoo, but I'm trying to stop that
[21:07:31] <Kovensky> lol
[21:07:42] <Kovensky> I can't even get myself addicted to lunix :(
[21:07:57] <elenril> then you're DoingItWrong
[21:08:08] <Kovensky> only windows has graphic drivers for my thing-that-draws-to-the-LCD
[21:08:19] <elenril> hey, there's a trope for that
[21:08:59] <Honoome> elenril: about drivers?
[21:09:16] <elenril> Honoome: about DoingItWrong
[21:09:23] <Honoome> no wonders
[21:09:25] <elenril> (actually it redirects to EpicFail)
[21:12:05] <Kovensky> the only things that keep pushing me back to windows are driver issues
[21:12:26] <Honoome> change videocard :P
[21:12:46] <Kovensky> Honoome: craptop :P
[21:12:52] <Honoome> change crap :P
[21:13:01] <Kovensky> I can't use fbsd here either, even with vesa, becausethere's no sis191 support <_<
[21:13:04] <Kovensky> not even with ndis
[21:13:09] <Kovensky> +\
[21:13:30] <elenril> 1)learn how to RE
[21:13:34] <elenril> 2) write the drivers
[21:13:38] <elenril> 3) ???
[21:13:51] <elenril> 4) world domination
[21:13:52] <Kovensky> it's to crap to give a damn
[21:13:57] <Kovensky> too*
[21:14:00] <Honoome> ouch sis… that's… bad
[21:14:02] <Kovensky> and no, I don't want to help SiS
[21:14:17] * elenril wonders how Honoome writes those ellipses
[21:14:30] <Kovensky> he has some silly s/// :V
[21:14:32] <Honoome> elenril: super+.
[21:14:36] <Honoome> I remapped the keyboard :D
[21:14:50] * Honoome got a bunch of other unicode charaters in there, including ←→ and ★ :D
[21:15:28] <Kovensky> compose+...
[21:15:42] <Kovensky> ←→ <-- dunno how to write this one lol
[21:15:43] <Honoome> Kovensky: compose is not working in qt4 for me =_=
[21:15:48] <Kovensky> ★ <-- ほし
[21:15:53] <mru> what's the point? 7 bits ought to be enough for anyone
[21:16:04] <Honoome> Kovensky: to compose the arrows just compose <-
[21:16:06] <Honoome> and ->
[21:16:17] <Honoome> but the three dots don't compose to the ellipsis here
[21:16:33] <Kovensky> ←→ <-- works with MS IME
[21:16:36] <elenril> ? ?oo?
[21:16:39] <Kovensky> <- space -> space
[21:47:07] <DonDiego> [sipr @ 0x904aef0]Internal error, IDCT permutation not set
[21:47:50] <mru> not like sipr would need it...
[21:47:59] <mru> what prints that line?
[21:48:27] <DonDiego> http://samples.ffmpeg.org/V-codecs/MSS1/GipsyGuitar.wmv
[21:48:31] <DonDiego> played with mplayer
[21:48:41] <DonDiego> can somebody quickly run this through ffplay?
[21:48:50] <DonDiego> my tree will take a moment to recompile..
[21:49:29] <CIA-90> ffmpeg: mru * r21947 /trunk/libavcodec/mathops.h: Add zero_extend() function
[21:49:29] <CIA-90> ffmpeg: mru * r21948 /trunk/libavcodec/get_bits.h: Deobfuscate LE SHOW_[SU]BITS; these are simple sign/zero-extend
[21:49:44] <mru> no errors
[21:50:22] <DonDiego> k
[21:57:21] * mru likes closing issues as invalid <5 minutes after they're opened :-)
[22:00:59] <DonDiego> SDL_OpenAudio:
[22:01:00] <DonDiego> /tmp/GipsyGuitar.wmv: could not open codecs
[22:01:01] <DonDiego> hmm
[22:13:10] <j-b> Vitor1001: nice
[22:13:24] <Vitor1001> j-b: :D
[22:13:38] <Vitor1001> j-b: But I wasn't the one who did most of the work
[22:13:48] <j-b> you merged :)
[22:13:52] <Vitor1001> It works fine on VLC?
[22:14:09] <j-b> I'll tell you in half an hour
[22:14:10] <Vitor1001> Well, I finished the review. Collin started it.
[22:14:16] <Vitor1001> Cool
[22:14:45] <j-b> but, that's a huge "pain" out of my schedule...
[22:15:00] <j-b> no more "I can't hear this 3gpp" file in VLC...
[22:15:16] <mru> is amr still used in the wild?
[22:15:21] <Vitor1001> Aren't the windows binaries distributed with opencoreamr?
[22:15:29] <j-b> Vitor1001: they are not.
[22:15:35] <Vitor1001> Why?
[22:15:41] <j-b> GPLv3
[22:16:00] <Vitor1001> I see
[22:16:18] <j-b> I had private builds
[22:16:29] <j-b> but the official one weren't with it
[22:16:37] <j-b> mru: no idea, but I see many complaints
[22:17:31] <Honoome> I'm quite sure Symbial still uses AMR
[22:17:50] <j-b> Vitor1001: anyway, many thanks to you... And superdump ain't around
[22:19:41] <Vitor1001> j-b: Do you get also a lot of complains about HE-AAC?
[22:20:11] <j-b> Vitor1001: no. Faad2 works IIRC
[22:20:35] <Vitor1001> VLC uses it by default?
[22:20:46] <j-b> yes sir
[22:24:50] <j-b> Vitor1001: but with DVB-HD Sub, WMAVoice, AMR-NB, WMAPro, Indeo5, Avcodec sub, Atrac1, avformat improvements + DxVA/VAAPI/OpenMax decoding, VLC 1.1 might get quite popular
[22:25:21] <Vitor1001> :D
[22:25:49] <Vitor1001> I think I also remember someone asking also about bink in some vlc ML ;)
[22:25:53] <kierank> "DVB-HD Sub" --> what's that?
[22:26:56] <j-b> Vitor1001: yes, I am waiting for Bink
[22:27:07] <j-b> Vitor1001: but that is less important
[22:27:20] <Vitor1001> j-b: see r21937
[22:27:46] <j-b> http://git.videolan.org/?p=vlc.git;a=commitdiff;h=91b3be07653bbfef1ac666436bf4f134ebcc1b2e
[22:27:49] <Vitor1001> j-b: bink is committed (both audio and video)
[22:28:16] <j-b> video is commited?
[22:28:21] <j-b> I missed that
[22:28:35] <Vitor1001> Since 2 PM today
[22:28:41] <j-b> Kewl
[22:33:53] <mru> we're dominating more than ever
[22:42:43] <kierank> merbzt, ping
[22:47:56] <peloverde> As far as the AAC goes, we are rapidly reaching the point where both faad and ffmpeg can handle nontrivial streams the other can't
[22:48:16] <Dark_Shikari> yup
[22:48:18] <Dark_Shikari> we're already there
[22:49:19] <peloverde> I don't know whay faad is so resistent to implementing whole levels and profiles
[22:50:33] <DonDiego> well, ffmpeg cannot handle sbr yet..
[22:50:44] <DonDiego> so i guess faad2 is still in front
[22:50:52] <DonDiego> what can ffmpeg do that faad2 cannot?
[22:52:44] <peloverde> channel coupling
[22:53:00] <peloverde> decode the conformance files within the required accuracy parameters
[22:53:36] <peloverde> Also SBR works fine, it's just not merged
[22:54:01] <mru> so faad isn't in fact an aac decoder at all ;-)
[22:55:06] <peloverde> indeed :)
[22:55:31] <peloverde> Also a 5.1 channel LC decoder is required to support at least one dependent coupling channel
[22:55:52] <peloverde> (even though nobody seens to use it in the wild)
[22:57:10] <Dark_Shikari> peloverde: another thing ffmpeg does
[22:57:15] <Dark_Shikari> not be horribly slow
[22:57:49] <peloverde> Dark_Shikari, we are horribly slow on mainprofile with the backward predictor on, but nobody uses that
[22:58:46] <iive> peloverde: does faad handle that case better?
[22:59:41] <peloverde> I'm sure if they are then they aren't conformant, you need to round to half float periodiclly or evertrhing goes to hell
[22:59:50] <peloverde> that's where most of the speed loss comes from
[23:01:46] <mru> cortex-a9 can do that in hw
[23:02:11] <_av500_> mru: can it be backported :)
[23:02:28] <mru> maybe with one of those plasma cannons...
[23:08:52] <j-b> Vitor?
[23:08:53] <j-b> too late
[23:08:57] <peloverde> Anyway I don't think anyone really cares about that feature, i just added it because flash and faad support it and i wanted to be checklist compliant
[23:23:40] <CIA-90> ffmpeg: michael * r21949 /trunk/libavcodec/mpeg12.c:
[23:23:40] <CIA-90> ffmpeg: Fix timestamp association for mpeg2 field pictures.
[23:23:40] <CIA-90> ffmpeg: Fixes /MPlayer/incoming/codec_copy_fails_vob_to_mpeg-ts/codec_copy_fails_vob_to_mpeg-ts.vob
[23:29:10] <CIA-90> ffmpeg: mru * r21950 /trunk/configure: Suppress armcc warnings about unknown attributes
[23:29:11] <CIA-90> ffmpeg: mru * r21951 /trunk/libavcodec/get_bits.h:
[23:29:11] <CIA-90> ffmpeg: get/show_bits() can read up to MIN_CACHE_BITS bits
[23:29:11] <CIA-90> ffmpeg: The limit for get/show_bits_long() to use get/show_bits() directly
[23:29:11] <CIA-90> ffmpeg: should be MIN_CACHE_BITS, not 17.