[FFmpeg-devel-irc] IRC log for 2011-02-27

[irc at mansr.com]

[00:38:16] <j0sh> merbanan: ah alright
[07:22:26] <ohsix> what are the odds of something like this uses mplayer or ffmpeg? http://www.dealextreme.com/p/1080p-full-hd-2-5-sata-hdd-media-player-with-hdmi-usb-host-optical-av-power-sd-mmc-slot-53806
[07:22:51] <kierank> maybe for demuxing
[07:23:41] <ohsix> i wanna get one but theres never any information on what they actually play, aside from a list of formats and containers
[07:24:27] <ohsix> need vobsub too D:
[07:24:51] <kierank> there's never any information probably because the manufacturer doesn't even know themselves
[07:26:32] <ohsix> i have a drive free thatneeds a case anyways, figured i'd do both
[10:05:53] <mmu> gee, the fate run I started before going to bed isn't finished
[10:30:08] <lu_zero> mmu how so?
[10:31:01] <mmu> athlon xp ...
[10:31:16] <mmu> could use a progress indicator on stdout
[10:31:46] <mmu> it's at fate-musep...
[11:00:51] <lu_zero> uhm
[11:59:07] <DonDiego> BBB: still queueing patches for ruggles? why?
[12:21:40] <mmu>  real    411m27.047s
[12:21:41] <mmu>  user    33m57.142s
[12:21:42] <mmu>  sys     304m18.420s
[12:21:59] <mru> os not so good
[12:22:44] <uau> lu_zero: why "pkg-config sdl --modversion > /dev/null 2>&1" instead of "pkg-config --exists sdl"?
[12:23:13] <mmu> building vbox along
[12:23:25] <mmu> http://fate.ffmpeg.org/x86_32-haiku-gcc-4.4.4
[12:39:42] <lu_zero> uau: no reason at all, your proposal makes more sense
[12:47:15] <DonDiego> btw, i'm (finally) fixing the snapshot generation script
[12:47:30] <lu_zero> =)
[12:47:31] <lu_zero> nice
[12:47:58] <DonDiego> should the tarball unpack to a directory called plain 'ffmpeg' or 'ffmpeg-something' where 'something' is the git hash for HEAD or so?
[12:48:19] <DonDiego> i've got a better idea - i will go with the date..
[12:48:23] <lu_zero> =)
[12:48:30] <lu_zero> that would be nicer indeed
[12:50:07] <DonDiego> confessional style of debugging strikes again :)
[12:50:27] <DonDiego> it's funny how often you get better ideas on your own when you explain a problem... :)
[12:50:44] <andoma> yeah :)
[12:51:03] <mru> sometimes it's only when you try to explain it that you actually fully formulate the problem
[12:57:23] <DonDiego> yes
[12:57:48] <DonDiego> what should i use as name for the snapshot tarballs?
[12:58:10] <DonDiego> i'm creating one with full git history and another one with the .git subdir deleted
[12:58:38] <DonDiego> previously they were called 'checkout' and 'export' to mirror the respective svn commands used to generate them
[12:58:58] <DonDiego> 'history' and 'bare' maybe?
[12:59:04] <mru> what's the point of creating snapshots with .git included?
[12:59:17] <mru> people who want that are better off using git directly
[12:59:34] <mru> see also the "git archive" command
[12:59:51] <DonDiego> maybe git is not allowed through the company firewall or similar
[13:00:04] <mru> then clone over http
[13:00:08] <mru> we support that
[13:00:15] <DonDiego> and probably less bandwidth used and less load for our server
[13:01:00] <mru> I doubt a few http cloners will be a problem
[13:07:36] <DonDiego> i don't think so either, but a git tarball cannot hurt
[13:07:57] <DonDiego> it could be useful for some people
[13:08:31] <DonDiego> it's just a few more commands in the shell script...
[13:54:18] <CIA-15> ffmpeg: Mans Rullgard <mans at mansr.com> master * r0b32da90f8 ffmpeg/libavcodec/arm/vp8_armv6.S:
[13:54:18] <CIA-15> ffmpeg: ARM: VP8: fix build on systems with global symbol prefix
[13:54:18] <CIA-15> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[13:54:20] <CIA-15> ffmpeg: Mans Rullgard <mans at mansr.com> master * r8b454c352f ffmpeg/libavcodec/arm/vp8dsp_neon.S:
[13:54:20] <CIA-15> ffmpeg: ARM: fix vp8 neon with pic enabled
[13:54:20] <CIA-15> ffmpeg: The assembler emits literal pools too far from the load instructions,
[13:54:20] <CIA-15> ffmpeg: so we must do it explicitly at a suitable location.
[13:54:21] <CIA-15> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[13:55:00] <lu_zero> mru: is there a specific reason for that?
[13:55:12] <mru> which?
[13:57:05] <lu_zero> The assembler emits literal pools too far from the load  instructions
[13:57:13] <mru> buggy assembler
[13:57:16] <mru> what else?
[13:57:32] <lu_zero> that's what I'm wondering
[14:21:59] <BBB> lu_zero: can you please re-queue it? not sure how but it got lost
[14:22:05] <BBB> although I'm pretty sure I did queue and push it
[14:22:07] <BBB> odd
[14:25:25] <lu_zero> just noticed now it was missing ^^;
[14:25:35] <lu_zero> wine-tested it though
[14:25:39] <lu_zero> seems working fine
[14:25:46] <lu_zero> queuing up myself then =)
[14:27:01] <lu_zero> curious it doesn't apply anymore to master...
[14:27:19] <lu_zero> uhm
[14:27:25] <lu_zero> got in already
[14:47:19] <BBB> where is kshishkov
[14:51:19] <siretart> DonDiego: ffmpeg 0.5.4?
[14:56:16] <lu_zero> BBB: btw
[14:56:22] <BBB> ?
[14:56:34] <lu_zero> the applehttpproto does solve a number of issues
[14:56:40] <BBB> ?
[14:56:56] <lu_zero> (some are solved removing the timestamp guessing code but I'm digressing)
[14:57:04] <BBB> :-p
[14:58:33] <lu_zero> BBB: still I'd push martin's alternative as well
[14:59:08] <BBB> I'm lost
[14:59:16] <BBB> which lternative?
[15:00:24] <lu_zero> we have two implementations from martin
[15:01:23] <BBB> I probably missed them :-p
[15:01:27] <BBB> looking into some major vc1 issues
[15:01:28] <lu_zero> as protocol -> more robust towards non-standard implementation, relies on the mpegts demuxer
[15:02:12] <lu_zero> as chain demuxer -> gets the best of the timestamp guessing code breaking remuxing and can miss segments in certain situations
[15:08:02] <lu_zero> btw I'm thinking about having the tcp protocol have a listen option
[15:09:06] * lu_zero found useful do stuff like ffmpeg -i tcp://localhost:1234 on an host and  ffmpeg -i stuff tcp://thathost:1234 on another
[15:10:02] <siretart> neat
[15:27:45] <lu_zero> uhmm
[15:27:58] <lu_zero> do we really need that rentry code on tcp.c?
[15:37:03] <Sean_McG> if I'm fixing a patch using git, do I have to unstage it, fix it and recommit or is there a way to edit it?
[15:38:03] <jannau> Sean_McG: if it's the last commit: edit; git add -p; git commit --amend
[15:38:56] <Sean_McG> OK
[15:46:22] <mru> fate is looking very green today
[15:46:43] <mmu> including for Haiku :)
[15:49:17] <mru> all the failures are actually compiler bugs
[15:49:53] <mmu> eh
[15:56:47] <spaam> mru: did you report them?
[15:57:13] <mru> some of them
[16:10:30] <siretart> mru: can we keep it green until we bump and branch 0.7? ;-)
[16:23:27] <mru> siretart: wouldn't that be great?
[16:26:34] <siretart> mru: absolutely!
[17:41:23] <Kovensky> supergreen!
[17:42:07] * mru adds some more broken compilers
[17:42:26] * Kovensky suggests tcc
[17:42:44] <mru> I have no interest in tcc
[17:47:23] <mmu> sdcc maybe ?
[17:47:45] <BBB> mycc!
[17:48:35] <BBB> mru: http://software.intel.com/en-us/forums/showthread.php?t=81056&o=a&s=lr
[17:48:38] <mmu> PacifiC but I don't think it knows mmx in 16bit
[17:48:40] <BBB> mru: we should submit problem reports to icc
[17:48:41] <mmu> :p
[17:48:52] <mru> go right ahead if you like
[17:49:45] <BBB> who's interested in icc here?
[17:49:48] <mru> there, another clang bug filed
[17:49:49] * BBB scans for victims
[17:49:57] <BBB> used to be carl eugen, but he's not on irc
[17:51:00] <BBB> also I have the vc1 reference decoder running
[17:51:10] <BBB> now I can see if weird crap output by our decoder is actually sane
[17:51:24] <mmu> FCC ?
[17:51:33] <mmu> ah no that's a stupid US admin
[18:04:47] <kshishkov> BBB: I'll look at VC-1 in half an hour or so
[18:04:57] <BBB> awesome
[18:05:08] <BBB> I'm gonna run the testsuite with reference decoder and then compare
[18:09:55] <Kovensky> 14:42.45 mru: I have no interest in tcc <-- I know, it's just to see it crying like a little girl in the corner when faced with ffmpeg
[18:10:12] <mru> it's more fun doing that with "serious" compilers
[18:14:02] <BBB> kshishkov: reference decoder picture order is as screwed up as with ours - I'm pretty sure this is not intended but maybe it's a bitstreambug or so? looks really weird
[18:27:07] * mru finds a /* WTF?? */ comment in binutils
[18:27:09] <mru> scary stuff
[18:59:45] <kshishkov> BBB: I think they didn't care
[19:01:33] <kshishkov> BBB: so neither should I :P
[19:02:17] <Sean_McG> what's tcc?
[19:02:27] <iive> tiny c compiler
[19:02:30] <kshishkov> C compiler from FFmpeg creator
[19:02:35] <kshishkov> (QEMU too)
[19:02:38] <Sean_McG> ah
[19:02:55] <iive> qemu uses it to compile code in realtime.. (kind of)
[19:03:35] <kshishkov> mru: have you tried PCC release or the fact BSD loves it is enough to shun it?
[19:03:54] <mru> ICE on ffmpeg
[19:03:58] <mru> several
[19:04:28] <mru> and it generally sucks too
[19:04:32] <kshishkov> well, even if its current devs are crazy they may listen to you
[19:32:46] <mru> hmm, maybe I should file another clang bug
[19:33:44] <kshishkov> got hooked on compiler bug reporting?
[19:34:42] <Sean_McG> I'm admittedly not smart enough to fix the bugs I've reported on GCC, but I like to know that I at least helped
[19:44:20] <lu_zero> kshishkov: the current devs are more crazy than gcc ones?
[19:45:01] <kshishkov> lu_zero: who knows - but tey are crazy by definition. Ask mru what he thinks about devs from Umeå
[19:45:32] <mru> luleå
[19:45:56] <kshishkov> ah yes
[19:46:04] <kshishkov> same thing for you though
[19:46:53] <lu_zero> ^^;
[19:49:36] <lu_zero> the fact they use cvs tells everything...
[19:49:43] <Sean_McG> lol
[19:49:50] <kshishkov> that they are still in 70s?
[19:49:58] * lu_zero ponders to suggest them to move to git in their mailing list
[19:49:59] <Sean_McG> I shouldn't laugh, I have to use cvs at work (government standards SUCK)
[19:50:09] <lu_zero> Sean_McG: O_o?
[19:50:24] <lu_zero> which government is that retro`?
[19:50:35] <Sean_McG> yeah I tried to get subversion and they said we haven't approved it for use yet
[19:50:41] <Sean_McG> I'm Canadian
[19:51:16] <kshishkov> lu_zero: nowadays I find git-svn extremely useful
[19:51:17] <lu_zero> ah
[19:51:25] <lu_zero> kshishkov: ^^
[19:51:40] <kshishkov> I think I'm the second guy who used it at work
[19:52:06] <mru> Sean_McG: I'd suggest running some kind of git frontend on your workstation
[19:52:20] <mru> and obviously don't tell IT dept about it
[19:52:28] <Sean_McG> they have inventory scanners.
[19:53:14] <Sean_McG> so I just put up with it... and Rational Developer doesn't have integration with other SCMs anyways
[19:53:19] <Sean_McG> other than CVS & ClearCase
[19:53:33] <Sean_McG> and we don't have ClearCase licenses
[19:54:28] * kshishkov always thought that IBM software was just a subliminal way to promote IBM hardware
[19:54:45] <Sean_McG> in the case of WebSphere I'm not so sure
[19:55:00] <Sean_McG> most of our server are HP blades
[19:56:14] <mru> what, you can't install your own sw?
[19:56:23] <Sean_McG> nope
[19:56:35] <mru> do yourself a favour and quit
[19:56:37] <mru> NOW
[19:56:38] <Sean_McG> I've actually had an IT Sec. guy come to my desk when I installed the Active Directory toolkit
[19:57:00] <Sean_McG> mru: I'm looking to leave, but there isn't much here for my skillset (I'm a Tivoli guy)
[19:57:17] <mru> where's "here"?
[19:57:29] <mru> and how far are you willing to move?
[19:57:37] <Sean_McG> I'm with Canada Revenue, and I'm not really willing to move
[19:58:09] <mru> I was expect a city as answer...
[19:58:11] <mru> +ing
[19:58:17] <Sean_McG> oh.... Ottawa, Ontario
[19:58:26] <Sean_McG> Canada's capital
[19:58:51] * thresh shudders when he hears IBM
[19:58:54] <kshishkov> not for all Canada though :P
[19:59:02] <mru> well, if you don't want to move...
[19:59:03] <kshishkov> thresh: HP C compiler is better?
[19:59:08] * thresh shudders more
[19:59:19] <mru> IBM's xlc compiler also fails on ffmpeg
[19:59:26] <Sean_McG> really? I'd figure vendor compilers blew away gcc
[19:59:35] <Sean_McG> mru: are there a lot of gcc-isms in ffmpeg?
[19:59:39] <mru> none
[19:59:42] <mru> only inline asm
[19:59:52] <mru> and some attributes
[19:59:57] <mru> but they're all optional
[20:00:08] <mru> a pure c99 compiler should work
[20:00:34] <kshishkov> Sean_McG: we actually learned about lots of IBM products at uni, our department is even IBM Business Partner but I ended not working by speciality and proud of it!
[20:01:01] <Sean_McG> kshishkov: *nodnod*
[20:01:05] <Sean_McG> kshishkov: where did you study?
[20:01:41] <mru> it was fun playing with an ibm (ex) supercomputer at uni
[20:01:49] <kshishkov> Sean_McG: National Technical University "Kharkiv Polytechnical Institute" (Ukraine)
[20:02:13] <Sean_McG> I guess the beefiest machine we have at work is our zSeries
[20:02:29] <Sean_McG> I've done mainframe stuff before, it's kinda boring
[20:02:42] <kshishkov> mru: at our uni we had documentation for one ordered to be thrown away short before we got server
[20:03:48] <mru> they let us play on an sp/2 system
[20:03:57] <Sean_McG> processing tax data is more I/O intensive than compute
[20:04:16] <Sean_McG> they proved that when they tried to switch to Sun and the machines keeled over with the load
[20:04:40] <kshishkov> mru: we had Intenet access at some classes but it being ~2 KB/sec (for whole class) was not very appealing
[20:05:06] <mru> kshishkov: heard of sunet?
[20:05:17] <kshishkov> nope
[20:05:22] <mru> swedish university network
[20:05:36] <mru> very fast
[20:05:55] <kshishkov> we had an access to Ukrainian universities network but nobody actually knew where it is and how to use it
[20:06:41] <spaam> sunet is awesomre
[20:06:47] <spaam> awesome
[20:07:02] <kshishkov> Swedish Internet in general too
[20:07:16] <kshishkov> even ordinary grannies can get fast enough access
[20:07:54] <mru> does clang use a builtin assembler?
[20:08:12] <kshishkov> yep
[20:08:21] <mru> that explains it
[20:08:37] <mru> time for a bug report, I think
[20:09:21] <ruggles> Tjoppen: ping
[20:09:23] <Sean_McG> mmm fiber to the home
[20:09:30] <Sean_McG> so jealous
[20:09:58] <mru> it would be nice, but adsl is usually good enough
[20:10:18] <mru> the only time I've really suffered was when I got slashdotted
[20:10:31] <mru> had to quickly mirror the page elsewhere
[20:10:34] <kshishkov> or when you have to upload something somewhere else
[20:11:10] <Sean_McG> I get 10mbit/512kbit cable here, and pay ~$50 a month for the priviledge!
[20:11:25] <Sean_McG> (it's well known that Canadians pay WAAAAAY too much for Internet)
[20:12:04] <mru> I pay about £30 for 24M/2M and 8 static IP
[20:12:25] <kshishkov> I pay ~30EUR for 32/1M
[20:12:41] <spaam> kshishkov: http://www.sunet.se/download/18.6d7c8917128274d3dd080006061/optosunetbroschyr_eng.pdf about sunet :)
[20:13:41] <kshishkov> spaam: not useful for me
[20:15:29] <spaam> ok :)
[20:17:15] <kshishkov> here even the only decent lingonsylt is the one made from Scandinavian berries
[20:17:47] <Sean_McG> mmm lingonberry juice... now I wanna go to Ikea!
[20:18:06] * kshishkov wants to go to Sweden but that's no news
[20:19:21] <kshishkov> the best sausage I've ever tried is called Hjörtkorv too
[20:20:02] <mru> sure that's not hjort?
[20:20:20] <kshishkov> no
[20:20:35] <kshishkov> could be renkorv though too
[20:20:39] <mru> google gets exactly one hit for Hjörtkorv
[20:21:35] <mru> and that one seems to be a typo
[20:21:45] <kshishkov> indeed
[20:26:25] <Sean_McG> damn, I never wrote down how I didn my 64-bit openssl builds
[20:26:38] * mru does emerge openssl :)
[20:26:42] <spaam> :)
[20:26:51] <Sean_McG> I'm on Solaris 10 SPARC in this case :P~
[20:27:00] <mru> you can still use portage
[20:27:00] <Sean_McG> SunFreeware's package is only 32-bit
[20:27:20] <lu_zero> Sean_McG: gentoo-prefix is your friend
[20:27:54] <Sean_McG> I'm not sure I want to install a whole new packaging system
[20:28:33] <mru> portage is awesome
[20:28:35] <lu_zero> http://www.gentoo.org/proj/en/gentoo-alt/prefix/bootstrap-solaris.xml
[20:28:57] <spaam> kan you use portage on osx?
[20:29:03] <lu_zero> sure
[20:29:10] <mru> I've seen it running on windows even
[20:29:15] <kshishkov> I used itthere
[20:29:18] <spaam> haha cool
[20:29:24] <mru> does it work on irix?
[20:29:25] <lu_zero> http://www.gentoo.org/proj/en/gentoo-alt/prefix/bootstrap-macos.xml
[20:29:34] <spaam> cool
[20:29:35] <lu_zero> same guide as solaris
[20:29:44] <spaam> then i can remove macports :)
[20:30:03] <lu_zero> spaam: help and contributions welcome =)
[20:32:12] <spaam> lu_zero: i will try it tomorrow  at work :)
[20:33:38] <lu_zero> =)
[20:36:56] <Tjoppen> ruggles: you rang?
[20:38:00] <Dark_Shikari> http://www.theonion.com/articles/nation-shudders-at-large-block-of-uninterrupted-te,16932/
[20:39:00] <mru> Dark_Shikari: they should try reading german
[20:39:32] <mru> I once saw a single sentence span an entire page
[20:39:50] <mru> if I read german more often, I'd probably see more of those
[20:44:19] <kshishkov> mru: I wrote such sentences in Russian too!
[21:53:03] <DonDiego> siretart: still online?
[21:53:30] <DonDiego> siretart: i gave you a paste with changes for 0.5.4, did you apply them already?
[21:55:51] <siretart> DonDiego: no, I didn't see the past
[21:55:56] <siretart> maybe my client crashed
[21:56:00] <siretart> and I'm about to crash as well
[21:56:51] <DonDiego> beastd: hi, long time no see
[21:57:15] <mru> peloverde: nice link
[21:57:29] <beastd> Hi Diego
[21:57:54] <Sean_McG> watching this poisonous people clip that Alex posted (I've never seen it before)
[21:59:42] <siretart> DonDiego: what changes do you have?
[22:02:04] <Sean_McG> 5PM and I am nowhere near done the chores I need to do, plus I was gonna make a pot of chili
[22:02:07] <Sean_McG> argh
[22:21:11] <ruggles> Tjoppen: do you have any lxf 20-bit pcm samples? we have the decoder but i can't find a sample to test with anywhere.
[22:21:58] <spaam> ruggles: cant you create 20bit pcm stuff with rox ?
[22:22:01] <spaam> sox.
[22:22:05] <merbanan> ruggles: I'll remind him tomorrow
[22:22:24] <saintd3v> peloverde: did you get a chance to glance at that patch?
[22:22:33] <ruggles> spaam: it's a special format, not normal 20-bit pcm
[22:22:47] <ruggles> CODEC_ID_PCM_LXF
[22:22:56] <spaam> ruggles: ah ok :)
[22:25:27] <BBB> ruggles: is aften merge done?
[22:26:57] <ruggles> BBB: not yet. basically setting ac3 metadata via avoption is the only part left that is currently in aften and is worth porting.
[22:27:49] <ruggles> but i took a small break to work on audio decoding API. just going back through all decoders to double-check and test things not covered by fate.
[22:28:41] <ruggles> which is why i need a PCM_LXF sample. :)
[22:28:49] <saintd3v> ruggles: did you see the FFPsyContext patch?
[22:29:07] <ruggles> saintd3v: no...
[22:29:19] <ruggles> on ffmpeg-devel?
[22:31:39] <saintd3v> no
[22:32:02] <saintd3v> swear i posted a link here, but can't find it in my logs...
[22:32:13] <saintd3v> ruggles: http://pastebin.com/j4gNgj4e
[22:34:20] <Tjoppen> ruggles: I do. I can take a look at work tomorrow
[22:34:25] <ruggles> saintd3v: ac3 needs 7 channels. there is a separate coupling channel that goes through the psy model & bit allocation.
[22:34:28] <ruggles> Tjoppen: thanks
[22:34:48] <Tjoppen> 4-cba  e
[22:34:54] <Tjoppen> *4-channel 20-bit IIRC
[22:35:44] <ruggles> saintd3v: i like the patch though
[22:42:11] <saintd3v> ruggles: sure, 6 was just an arbitrary number, it could be 20 for all it really matters.
[23:08:40] <DonDiego> gnite
[23:32:46] <peloverde> saintd3v: the patch doesn't seem to play nice with the spreading changes
[23:37:11] <saintd3v> oh, forgot i had comitted the energy spreading changeset already
[23:41:38] <saintd3v> peloverde: this is vs. ffmpeg.org http://pastebin.com/MwPKkzyj
[23:44:14] <mru> anyone here on friendly terms with clang?
[23:44:31] <mru> error: unable to interface with target machine
[23:44:41] <mru> ^^ any ideas what that means?
[23:44:57] <mru> Yuvi: ^^
[23:47:54] <Sean_McG> brb
[23:48:09] <roxfan> haha
[23:48:32] <roxfan> at fosdem they bragged about how their error messages are so much better than gcc
[23:48:55] <mru> some of their code error messages are better
[23:49:04] <mru> but their command line syntax is a complete mess
[23:49:25] <mru> there are at least 3 layers of option handling
[23:49:33] <peloverde> saintd3v: what is aacpsy.c using from float.h?
[23:50:48] <saintd3v> nothing any more. i was using FLT_MIN somewhere.
[23:51:19] <roxfan> looks like it happens if TM->addPassesToEmitFile fails
[23:51:19] <saintd3v> oh, for pe_max
[23:51:22] <roxfan> "This method should return true if emission of this file type is not supported, or false on success."
[23:52:14] <saintd3v> but i just recently changed that to use shoehorn values similar to what 3gpp uses