[FFmpeg-devel-irc] IRC log for 2010-07-09

[irc at mansr.com]

[00:00:45] <Vitor1001> But I still insist in the fate2 target. I'd hate to wait "not-too-long" to test more codecs...
[00:00:52] <mru> ok, ok
[00:02:00] <mru> reload git
[00:05:54] <mru> right the, lets make some tests
[00:05:55] <Vitor1001> mru: Perfect for me.
[00:06:03] <mru> +n
[00:07:21] <Vitor1001> mru: Well, now I'm going to sleep. But tomorrow I'll have a look at it.
[00:07:22] <Vitor1001> Thanks!
[00:07:44] <mru> I've been wanting to get this done forever
[00:07:49] <mru> thanks for the kick in the rear
[00:08:59] <Vitor1001> ;)
[00:09:04] <Vitor1001> Good night!
[00:09:25] <mru> it's troll driven development :-)
[00:41:38] <CIA-99> ffmpeg: mru * r24132 /trunk/tests/ (91 files in 4 dirs):
[00:41:39] <CIA-99> ffmpeg: tiny_psnr: print max absolute difference between files
[00:41:39] <CIA-99> ffmpeg: Regression test reference updates are due to the extra output
[00:41:39] <CIA-99> ffmpeg: from tiny_psnr.
[00:41:39] <CIA-99> ffmpeg: Patch by Vitor Sessak
[00:41:39] <CIA-99> ffmpeg: mru * r24133 /trunk/ (Makefile tests/fate-run.sh):
[00:41:40] <CIA-99> ffmpeg: fate: pass additional variables to test script
[00:41:40] <CIA-99> ffmpeg: This passes the following additional variables to the test script:
[00:41:41] <CIA-99> ffmpeg: - CMP method to use comparing output with reference, default 'diff'
[00:41:41] <CIA-99> ffmpeg: - REF reference file, default 'tests/ref/fate/${test}'
[00:41:42] <CIA-99> ffmpeg: - FUZZ maximum fuzz factor in comparison, meaning depends on CMP
[00:41:42] <CIA-99> ffmpeg: mru * r24134 /trunk/ (Makefile tests/fate-run.sh):
[00:41:43] <CIA-99> ffmpeg: fate: add oneoff comparison method
[00:41:43] <CIA-99> ffmpeg: This adds the 'oneoff' comparison method. It compares two s16le pcm
[00:41:44] <CIA-99> ffmpeg: files with a max difference <= the fuzz factor treated as success.
[00:41:44] <CIA-99> ffmpeg: Default fuzz is 1. On failure, the full tiny_psnr output is printed.
[01:06:33] <bcoudurier> hey guys
[01:06:44] <Dark_Shikari> hey bcoudurier's onjoin script
[01:06:54] <bcoudurier> not today
[01:11:46] <Compn> Dark_Shikari : i think thats 'hi guys'
[01:12:04] * Compn is sad when mru , diego and michaelni fight :(
[01:16:57] * hyc just deletes most of those emails without reading
[01:17:05] * hyc remains blissfully unaware
[01:17:10] <Dark_Shikari> the funniest part is that in all this time
[01:17:14] <Dark_Shikari> they could have just fixed it correctly
[01:17:33] <CIA-99> ffmpeg: siretart * r24136 /branches/ (0.6/libavcodec/ac3_parser.c 0.6):
[01:17:33] <CIA-99> ffmpeg: ac3: make the value of codec_id during (E-)AC-3 parsing stay CODEC_ID_EAC3
[01:17:33] <CIA-99> ffmpeg: if any E-AC-3 frames have been detected instead of switching back and forth for
[01:17:33] <CIA-99> ffmpeg: AC-3 core + dependent E-AC-3 substream(s).
[01:17:33] <CIA-99> ffmpeg: Fixes Issue 2022.
[01:17:33] <CIA-99> ffmpeg: backport r24103 by jbr
[01:19:25] <hyc> yes and no. they can't commit a fix until everyone agrees, otherwise it's just a commit war
[01:19:45] <Dark_Shikari> Not quite.
[01:19:53] <Dark_Shikari> Everyone agrees the current one is insufficient.
[01:19:56] <Dark_Shikari> Everyone agrees on how to fix it.
[01:20:05] <Dark_Shikari> Look at what they're arguing about...
[01:20:11] <Dark_Shikari> they're arguing about whether the partial fix was a step forward or back.
[01:20:16] <hyc> but that seems to be par for the course here. all of the patches I've submitted take only a few minutes to write and test. they all take forever to get positive review...
[01:20:18] <Dark_Shikari> DESPITE THE FACT THAT THEY ALL AGREE ON WHAT A COMPLETE FIX WOULD BE.
[01:20:45] <mru> that's michael in a nutshelll
[01:21:14] <Dark_Shikari> s/michael/michael and mru and diego/
[01:21:20] <hyc> lol
[01:21:44] <Dark_Shikari> If any of you were not that, you would have committed a full fix already.
[01:21:48] <Dark_Shikari> But a full fix isn't committed.
[01:21:52] <Dark_Shikari> Therefore, you are all responsible.
[01:22:11] <mru> look, I've been doing tons of other work
[01:22:31] <mru> I had nothing to do with that commit
[01:22:40] <Dark_Shikari> Then why are you involved in the whine-a-thon?
[01:23:06] <mru> because I can't stand by and watch michael slaughter diego
[01:23:25] <Dark_Shikari> Then commit a fix.
[01:23:40] <mru> why the fuck should _I_ do it?
[01:23:46] <Dark_Shikari> 09:23 <@mru> because I can't stand by and watch michael slaughter diego
[01:24:13] <mru> just fixing would still let michael believe he was right
[01:24:18] <mru> and that nobody opposed him
[01:24:22] <Dark_Shikari> It would stop him from "slaughtering diego"
[01:24:26] <Dark_Shikari> which is what you wanted.
[01:25:01] <hyc> eh, I see mru's point. you have to get the point across too, not just badnage up the immediate situation.
[01:25:02] <siretart> probably yes on this particular case
[01:25:13] <hyc> bandage
[01:25:20] <Dark_Shikari> hyc: "getting the point across" in an argument between mru and michael
[01:25:35] <Dark_Shikari> is about as likely as a communist party election win in Texas
[01:28:24] <hyc> I guess. I'm still a relative newbie here. I see mru's interactions on irc all the time, michael never seems to come around here.
[01:29:17] <drv> michael reads the archived logs and sometimes responds on the mailing list
[01:30:18] * beandog waves to michael
[01:36:27] <Compn> ya i agree with Dark_Shikari , a fix could be done pretty quickly
[01:36:38] <Compn> you guys just like to argue and point at each other and say its his fault! he argued!
[01:37:00] <Compn> yes there are faults, but damn you guys argue a lot
[01:37:07] <Compn> not important
[01:38:17] <Compn> hyc : yes, patch acceptance in ffmpeg is a slow process. but you could be a reviewer and do it faster if you wanted :P
[01:40:16] <hyc> eh? be a reviewer for my own patches? that doesn't sound right...
[01:40:28] <Dark_Shikari> that's why so much code in ffmpeg is shit
[01:40:39] <Dark_Shikari> nothing gets done unless you're your own reviewer
[01:40:44] <Dark_Shikari> and if you're your own reviewer, shit gets in.
[01:42:43] <siretart> do you think that better tool support for reviewing patches (like some webapp) would encourage more and better reviews?
[01:42:49] <Dark_Shikari> NO.
[01:47:03] <Compn> hyc : a reviewer for all patches
[01:47:05] <Compn> lol siretart
[01:47:13] <Compn> the thing is
[01:47:23] <Compn> ffmpeg devs review a patch and let patch submitters fix them
[01:47:35] <Compn> what you want are ffmpeg devs that review + fix + commit a patch themselves
[01:47:38] <hyc> Compn: well, I have only studied small parts of the code
[01:48:35] <siretart> Compn: no I don't. I was thinking about somethink like https://code.edge.launchpad.net/bundlebuggy, which recognizes patches to the mailing list and tracks their status, like needs work, or merged in
[01:48:50] <siretart> bah, laggy ssh here in .ca
[01:49:04] <Compn> ffmpeg has used roundup for that a few times
[01:49:19] <mru> roundup is shit
[01:49:21] <siretart> roundup has a mailinglist tracking feature for picking up patches?
[01:49:25] <mru> all bug trackers are shit
[01:49:35] <Dark_Shikari> +1000000000000000000000000000000000000000000
[01:50:01] <Compn> siretart : no, also i didnt say it was meant for that, just that we've used that in the past. it could be used for that purpose if we set roundup to watch ffmpeg-patches and sent patches there of course...
[01:50:29] <siretart> sounds like a lot of overhead to me
[01:50:32] <mru> you can mail patches to roundup and it'll open new issues for them
[01:50:49] <Compn> yeah ffmpeg-issues i think
[01:50:54] <mru> but that wouln't help getting them reviewed
[01:51:16] <Compn> i think siretart is talking about patches that get lost
[01:51:21] <siretart> exactly
[01:51:55] <Dark_Shikari> Patches get lost on bugtrackers too.
[01:52:17] <Compn> we have a lack of bugtracker maintainers (like carl) as well
[01:54:10] <hyc> I think it's worth trying, easier to find entries in a tracker than dig thru email archives
[01:54:55] <mru> a patch tracker would only work if a few conditions were met:
[01:55:01] <mru> 1. all patches were entered
[01:55:24] <mru> 2. the tracker is updated when patches are applied
[01:55:29] <mru> 3. it doesn't suck ass
[01:56:18] <hyc> the other problem is that trackers tend to be separate from the mailing list where discussions take place, things get awkward
[01:56:49] <hyc> but once you have a tracker, it's easy to set a commit template that requires the tracker item #
[01:56:56] <drv> i think wine uses something like that, although that works better as there is a single committer
[01:57:01] <hyc> so that commit logs that are forwarded to the tracker can automatically update it
[01:58:01] <Compn> siretart : while it wont work for ffmpeg, i think a patch tracker would do wonders in mplayer :)
[01:58:24] <mru> it would have to be something specially designed for patches and nothing else
[01:59:09] <Compn> siretart : also, no one is saying a 3rd party cant try it and see how it works. it might be a good way to show the devs how useful it is :P
[01:59:13] <Compn> assuming it works well
[01:59:32] <mru> sure, feel free to set up as many trackers as you like
[01:59:48] <Compn> since they are all quite boneheaded :)
[01:59:53] <Compn> hehe
[01:59:58] <drv> this is the wine one: http://wiki.winehq.org/PatchWatcher
[02:00:07] <drv> automatically picks up patches from a mailing list
[02:00:42] <Compn> siretart : if you are interesting in running one temporarily, i'm sure the devs will at least take a quick look at it
[02:00:50] <mru> it takes me more than 10 seconds to work out what to do => it sucks
[02:02:44] <drv> actually i think that is the wrong thing, that is some kind of automated patch tester, but i'm pretty sure they have an automated tracker somewhere...
[02:03:08] <drv> http://source.winehq.org/patches/
[02:03:58] <Compn> automated patch tester ala fate would be badass too
[02:04:03] <mru> I already see a weakness
[02:04:21] <mru> look at items 2 and 3 on that list
[02:04:28] <mru> they are different versions of the same patch
[02:04:49] <mru> automated patch tester is dangerous
[02:05:07] <mru> anyone could send a patch full of malicious code to the ml
[02:05:19] <mru> it would of course never be applied
[02:05:25] <drv> yeah, they require a specific format for the subject with different tries of the same patch
[02:05:53] <Compn> mru : the patch tester would just be checking for regressions, binary bloat, etc. not testing for ... security
[02:05:59] <hyc> yeah, pi should just be 3.
[02:06:32] <mru> Compn: ok, so I send a patch which replaces the configure script with rm -rf /
[02:06:50] <drv> there's no need to actually run the patch if you're just tracking them
[02:07:11] <mru> 03:03 < Compn> automated patch tester ala fate would be badass too
[02:07:13] <hyc> right. you don't need an automated patch tester, or certainly not as a frontend.
[02:07:24] <Compn> mru : and? the patch tracker would say that the reg tests failed ...
[02:07:32] <Compn> tester*
[02:07:41] <Compn> and compilation failed
[02:07:42] <mru> no, there'd be nothing left of the patch tester after that patch
[02:07:42] <Compn> etc
[02:07:45] <hyc> can run it once any reviewer has given it a once over
[02:08:01] <drv> you could do some kind of sandbox thing, but that quickly devolves into madness (or at least a lot of work)
[02:08:05] <mru> now that one is of course obvious and easy to protect against
[02:08:06] <Compn> so linux is the most secure os in the world but it cant sandbox ? :P
[02:08:15] <mru> but what about one that turns the machine into a spambot?
[02:08:36] <Compn> alright forget it
[02:08:39] <hyc> notice that the linux-next tree is merged manually by a dedicated maintainer. nobody autocommits emailed patches into their tree.
[02:08:51] <mru> there is no sandbox I'd trust enough to run random code off a mailing list in
[02:08:55] <Compn> i'm not really interested in arguing this, if ffmpeg devs dont want it, nevermind :)
[02:09:10] <mru> it would be a neat thing to have
[02:09:15] <hyc> tracking is different from autocommit
[02:09:15] <drv> anyway, the point is not really testing patches, just showing a list of ones that need attention
[02:09:15] <mru> unfortunately it can't be done safely
[02:09:56] <mru> well, michael refuses to use a web browser so there's no point discussing it
[02:10:53] <Dark_Shikari> everyone vote 6
[02:10:58] <hyc> (sheesh you guys probably spent more time arguing over docs than anybody ever spent actually reading them...)
[02:11:26] * mru never reads them
[02:11:26] <Dark_Shikari> \6/
[02:11:26] <mru> and for the record, I'm not arguing about the docs
[02:11:34] <mru> I'm arguing about the way michael is treating diego in particular
[02:11:35] * Compn hasnt read ffmpeg docs
[02:12:15] <hyc> the vote is already flawed because the change wasn't 1st or 3rd person. it was present vs future tense. at least the changes that I saw go by...
[02:12:21] <Dark_Shikari> Hence why I propose 6.
[02:13:03] <mru> I don't know what the proper name for that tense is
[02:13:14] <mru> could be seen as imperative too
[02:13:26] <mru> or a degenerate future
[02:13:43] <mru> whatever it is, it's what dictionaries use
[02:13:51] <mru> so it makes sense for documentation
[02:14:01] <drv> when i do read the docs, i never think about what tense or person they are in
[02:14:09] <beandog> yah me neither
[02:14:24] <drv> but some people like to make everything into an engineering problem
[02:14:39] <mru> I mostly think about what kind of person could have written it
[02:14:44] <hyc> for the record, I prefer imperative. "Calculate X". Not "Calculates X"
[02:15:11] <mru> yes, so do all good english-speakers
[02:15:12] <hyc> the latter is actually incomplete, should be "This calculates X" but that's just a waste of words.
[02:17:40] <hyc> and I certainly didn't join a software community to discuss grammar...
[02:17:49] * Compn laughs at getting hyc stuck arguing about docs
[02:18:30] <Dark_Shikari> oh god
[02:18:34] <Dark_Shikari> now we have a bikeshed about grammar and pi
[02:19:17] <hyc> btw, where does "bikeshed" originate?
[02:19:21] <mru> there is nothing more stupid than attempting to vote on something that has an objectively correct answer
[02:19:35] <mru> bikeshed.org
[02:19:43] <mru> try reloading a few times
[02:21:11] <saintdev> i vote pi = 2*C/D
[02:21:47] <mru> oh you're one of those...
[02:22:01] <saintdev> mru: no, just trolling :P
[02:22:36] * drv hails from the great state that once attempted to square the circle
[02:31:03] <hyc> so not only do lots of people toss in their 2cents on small issues, lots of them are wrong. :P
[02:31:25] <mru> the majority is always wrong
[02:31:31] <mru> that's a safe assumption
[02:33:18] <beandog> heh
[02:37:24] <mru> a lot of people have great difficulty differentiating between what actually is true and what they would like to be true
[02:40:59] <mru> "Klingon programs don't have parameters.  They have arguments and win them." -- Walter Bushell
[02:42:47] <Compn> so uh
[02:43:00] <Compn> did they ever show a klingon programming all that spaceship stuff ?
[02:43:06] <Compn> or what
[02:43:42] <mru> star trek is quite unrealistic at times
[02:44:02] <mru> like there's no way many of those "people" they meet would ever manage to build a spaceship
[02:45:01] <j0sh> well, at least some things are realistic about star trek
[02:45:09] <j0sh> like warp speed, time travel and wormholes
[02:45:22] <mru> eh, warp speed is realistic?
[02:45:29] <hyc> :P
[02:45:50] <mru> and what about "beaming" down to some hostile planet dressed in pyjamas?
[02:46:54] <Compn> armor is for the weak i guess
[02:47:29] <j0sh> yes, dermal regenerators will heal all
[02:47:37] <mru> the klingons seem to be wearing some kind of armour
[02:47:43] <mru> except worf of course
[02:47:59] <hyc> ah, it was a 5 year mission and you only saw it once a week. you never saw them exploring the atmosphere-less planets in environmental suits.
[02:48:06] <j0sh> well, that's because dermal regenerators work with energy weapons
[02:48:13] <j0sh> and klingons like hand-to-hand combat
[02:48:17] <j0sh> like, with their bat-leths
[02:48:18] <j0sh> and stuff
[02:48:29] <j0sh> thats why they need armor
[02:48:57] <mru> phasers set to kill seem to, well, kill
[02:49:03] <j0sh> true
[02:49:25] <hyc> except when they don't. how the hell does a creature resist disintegration?
[02:49:25] <mru> modern scifi shows are a bit more realistic in that regard
[02:49:39] <mru> like in the stargate series, they first send a robot, then they send marines
[02:49:57] <mru> _then_ they send a cute scientist who gets kidnapped by aliens
[02:50:04] <j0sh> hehe
[02:50:22] <hyc> yeah well, the aliens are smart enough to recognize soldiers and lay low at first
[02:51:31] <hyc> I doubt anyone here has much good to say about any scifi tv or film. those writers are either lower IQ, or writing for an audience with lower IQ.
[02:51:48] <Compn> hyc : you seen 'odyssey 5'?
[02:51:55] <Compn> now theres some good scifi tv :)
[02:51:55] <hyc> nope
[02:52:13] <Compn> i highly recommend it
[02:52:23] <j0sh> battlestar galalctica is pretty good scifi tv
[02:52:28] <j0sh> the new one, at least
[02:52:34] <Compn> too soap opera ish for me
[02:52:36] <mru> only the first 3 seasons
[02:52:39] <Compn> whos a robot this week durr
[02:52:49] <Dark_Shikari> I never got BSG.
[02:52:54] <Dark_Shikari> You want good scifi?
[02:52:59] <Dark_Shikari> Ghost in the Shell: Standalone Complex
[02:53:02] <Dark_Shikari> no fucking spaceships
[02:53:04] <mru> and starbuck returning as some kind of angel... wtf?
[02:53:15] <Dark_Shikari> no religion, no stupid kids (I'm looking at you wesley)
[02:53:38] <Dark_Shikari> no episodic this-series-will-go-on-for-9-seasons-until-we-run-out-of-money-and-cancel-it
[02:53:44] <Dark_Shikari> like Star Trek does
[02:53:44] <mru> someone should make a new star trek series where they go back in time and kill wesley
[02:53:47] <Compn> gits:sac had a lot of non-plot advancing episodes tho
[02:53:54] <Dark_Shikari> Compn: those are fine
[02:53:59] <Dark_Shikari> What isn't fine is when there IS no plot
[02:54:01] <Compn> yeha, some are great, some not so much
[02:54:02] <Dark_Shikari> i.e. it's written as they go
[02:54:07] <Compn> right
[02:54:07] <Dark_Shikari> and then they decide one day to end it
[02:54:10] <Dark_Shikari> so they tell the writers to do that.
[02:54:12] <hyc> yeah, gits was kind of annoying sometimes. meandered too much.
[02:54:21] <j0sh> i enjoyed sarah connor chronicles while it was on hulu
[02:54:29] <Dark_Shikari> I thought the non-main-plot episodes were great.  sometimes better than the plot ones.
[02:54:39] <Compn> terminator not directed by cameron? heh
[02:54:43] <Dark_Shikari> It avoided what I'd call "Gundam Syndrome"
[02:54:52] <Dark_Shikari> i.e. every episode starts exactly where the last ended
[02:54:58] <Compn> the tachikoma episode(s) were crazy
[02:55:01] <Dark_Shikari> so it might as well be one long movie with a lot of ad breaks.
[02:55:28] <Dark_Shikari> Compn: Those were possibly the _best_ episodes.  And an excuse for the writer to philosophize about machine intelligence, of course.
[02:55:36] <Compn> oh i agree :)
[02:55:38] <Compn> crazy good
[02:55:43] <Dark_Shikari> The thing I liked most is how they intentionally fucked with the relationship between humans and machines
[02:55:48] <Dark_Shikari> normally in every stereotypical scifi ever
[02:55:51] <Dark_Shikari> machines are cold and unfeeling
[02:55:53] <mru> it bugs me when they change the special effects from one season to the next
[02:55:54] <Dark_Shikari> humans are, well, humans
[02:55:59] <Dark_Shikari> GITS:SAC said -- fuck that
[02:56:03] <Dark_Shikari> humans are cold and unfeeling, for the most part
[02:56:13] <Dark_Shikari> the machines are the counterpoint to humans
[02:57:39] <Compn> i did see some good scifi, 'twelve months of the summer' (swedish) and moebius (argentinian)
[02:57:44] <Compn> scifi movies
[02:58:34] <mru> good swedish film??? cannot be
[02:59:22] <Compn> ehe
[02:59:31] <Compn> it was a tv movie
[03:01:09] <mru> even worse
[03:01:45] <mru> everybody knows all the best swedes go elsewhere
[03:01:59] <mru> there are quite a few swedish actors working in hollywood
[04:17:10] <_av500_> what stopped you?
[04:45:35] <pengvado> If I don't document any of my code, does that make me exempt from flamewars about documentation style?
[05:09:49] <Compn> well i hope you didnt use pi at all
[05:10:34] <thresh> moroning
[05:12:08] <wbs> pengvado: of course not, then you'll be flamed for not documenting your code :-)
[05:52:16] <hyc> I'm going to re-indent all of my code in Petrarchan sonnet form
[06:54:20] <j0sh> hyc: any idea why ffmpeg's native rtmp won't handshake with rtmpsrv?
[08:32:33] <markuman> i've got a question to the "missing picture in access unit" message. does parse_nal_units search for a pair of frame? http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2009-July/073538.html
[08:53:49] <Tjoppen> kshishkov: any luck with that vc-1 bug?
[08:54:31] <kshishkov> Tjoppen: not yet and I think the best way to fix it is to introduce interlaced decoding support
[08:55:52] <Tjoppen> ah, ok. not a big issue for me since that was just a test. my proper files work fine in most players
[08:56:03] <Tjoppen> although as said, crashes are important
[08:56:08] <kshishkov> indeed
[08:56:22] <Tjoppen> sadly, I have yet to manage to make a file that WMP will play :(
[08:56:53] <Tjoppen> except remuxing that 1440x576 sample
[09:44:21] <mru> http://varnish-cache.org/docs/phk/autocrap/
[09:44:40] <av500> yeah
[09:45:07] <kshishkov> "so copy&paste became the order of the day" <- sounds like an understatement
[09:47:38] <av500> ...The fourth school is the “DoxyGen” school, which lets a program collect a mindless list of hyperlinked variable, procedure, class and filenames, and call that “documentation”....
[09:47:44] <av500> :)
[09:48:06] <kshishkov> well, have you ever heard of WEB?
[09:48:25] <av500> a mindless list of hyperlinked pages....
[09:48:39] <mru> doxygen-type tools, if used responsibly, can serve to generate an API reference
[09:48:47] <mru> API reference != documentation
[09:49:04] <kshishkov> av500: no, that's really documenting language from DEK
[09:49:09] <Dark_Shikari> what is a web?  a miserable little pile of html!
[09:49:15] <Dark_Shikari> but enough links, have at you!
[09:50:03] <Tjoppen> regarding autocrap: is CMake the thing to use nowadays? I poked a bit at it and it wasn't too hard to get a library+executable style project going
[09:50:43] <thresh> cmake, blergh
[09:51:10] <kshishkov> scons competitor?
[09:51:26] <mru> whatever happened to jam?
[09:51:33] <mru> scones with jam, mmmm...
[09:51:40] <av500> +1
[09:51:53] <av500> and clotted cream
[09:51:55] <Honoome> kshishkov: nothing can compete with scons regarding idiocy
[09:51:58] <mru> that's another problem with these newfangled tools
[09:52:01] <Honoome> even though some things try
[09:52:06] <mru> there's no guarantee they'll stick around
[09:52:21] * kshishkov looks at M4
[09:52:26] <av500> otoh autotool hell will be there forever...
[09:52:30] <mru> and M4 looks back at kshishkov
[09:52:45] <ohsix> the second you need something cmake doesn't do, its untenable; and theres no corpus of standard do-ers to draw on, its all ad-hoc
[09:52:45] <Honoome> well, the copy-paste of idiotic code is just the same between autoconf and cmake, to be honest
[09:52:46] <mru> av500: I was implying that 'make' isn't going away
[09:52:57] <Honoome> what ohsix just said
[09:53:15] <Tjoppen> ohsix: that's the impression I got too
[09:53:24] <mru> that's where all these fancy systems fall short
[09:53:33] <Honoome> so cmake sounded cool and clean when just a few of mostly-identical things were to be done
[09:53:39] <Tjoppen> if it doesn't support library X you need to add locator scripts for it anyway
[09:53:45] <mru> as long as you stick to exactly the limited scenarios the authors had in mind, they sort of get the job done
[09:53:47] <Honoome> give it a couple of years (even less in the hands of the kde guys) and it'll be just as bad as autohell
[09:53:48] <mru> after much cursing
[09:53:55] <Honoome> and the .cmake files are just the same as the .m4 ones
[09:54:05] <av500> ic
[09:54:25] <Honoome> (if you have never seen the kde 3 build system you never saw REAL autohell
[09:54:29] <ohsix> i'd sooner ad-hoc it all in make, a target list is about what you need for simple-for-cmake stuff
[09:54:43] <astrange> was kde 3 the one where all the apps were built at .so with a shell to launch them?
[09:54:47] <kshishkov> Honoome: are you implying that even KDE is not worthy to be left in hands of KDE guys?
[09:54:48] <astrange> *as
[09:55:03] <mru> astrange: sounds familiar
[09:55:17] <Honoome> they hacked so much around automake, libtool and even autoconf that you changed a single version and hell broke loose... we had to replace their admin/ directory IN EACH FUCKING PROJECT to use newer tools
[09:55:22] <mru> starting any app would also start a gazillion rpc servers of various kinds
[09:55:36] <Honoome> astrange: sorta, that's not the problem with the build system as much as with their trying to implement prelink-without-prelink
[09:55:52] <Honoome> kshishkov: I imply that whoever does buildsystem for kde should be denied access to anything buildsystem-related
[09:56:00] <mru> why do people care so much about app startup time?
[09:56:09] <mru> apps I use frequently are always running
[09:56:45] <Honoome> mru: there is one reason for them to hope for prelink & similar
[09:57:12] <Honoome> their kdelibs package (used to, probably still) has HUGE vtables
[09:57:18] <astrange> if your app quits when the last document closes it's a problem
[09:57:39] <Honoome> which meant you could easily have over 100MB used in relocated .data.rel sections
[09:57:43] <mru> close documents?
[09:57:53] <Honoome> and if you are non-PIC... good luck with the .text that cannot be shared
[09:58:03] * mru uses emacs and rarely kills buffers
[09:58:30] * Honoome goes one step further and uses emacs-daemon
[09:59:01] <mru> is that the same thing as gnuserv?
[09:59:32] <kshishkov> no, it's their missing link to GNU Hurd+EMACS
[09:59:51] <Honoome> mru: don't think so
[10:00:05] <mru> then please explain
[10:00:08] <astrange> is there anything left for the os to do when you have emacs? maybe that's why they didn't finish hurd
[10:00:14] <Honoome> mru: it's like running emacs-23 with server-mode and then using emacsclient to open new frames, in TTY and X
[10:00:24] <mru> astrange: emacs doesn't have any builtin device drivers
[10:00:34] <mru> Honoome: that's gnuserv
[10:00:45] <mru> and gnuclient
[10:01:40] <Honoome> mru: I thought that only gave you a subset of features?
[10:02:09] <mru> with xemacs at least it can open a frame anywhere you please
[10:02:40] <Honoome> ah right you use the genetically-mutated version :)
[10:02:57] <mru> xemacs used to be miles better than gnu emacs
[10:03:03] <mru> and it's not called gnu
[10:03:15] <Honoome> well with emacs-daemon I just have an init script to load with the system, and then I have emacs loaded until I shut it down (or it crashes :|)
[10:03:21] <kshishkov> everything seems to be better without gnu (except zoos)
[10:03:21] <mru> I haven't tried gnu emacs for ages
[10:03:37] <mru> all its defaults are different though, so I get confused
[10:03:43] <Honoome> mru: then RMS left the maintainership to people who actually do stuff, and emacs 23 regained popularity :D
[10:03:54] <mru> oh, that's what happened
[10:04:10] <mru> I wonder if rms _ever_ did anything useful
[10:04:33] <Honoome> I could re-use a joke from BBC4's News Quiz
[10:05:07] <Honoome> "he's like God, not believed in by the vast majority of people, hasn't done anything useful in ages, ..."
[10:05:49] <kshishkov> what were original 1000 answers fitting that description?
[10:06:09] <Honoome> kshishkov: it was a longer joke and was referring to politicians in general
[10:06:44] <Honoome> anyway as I keep repeating, all kudos for starting the project/idea to rms... but he has definitely taken his "church of emacs" and "purity" problems to religious levels... and that makes him despisable in my view
[10:06:45] <kshishkov> Honoome: well, GNU is a political system anyway
[10:06:56] <Honoome> kshishkov: worse, it's a religion
[10:07:27] <kshishkov> what's the difference?
[10:07:32] <iive> it's worse.... it is based on reality.
[10:08:01] <Honoome> you can actually look at a politician who compromises or double-talks to get his program passed as somebody that can make a difference
[10:08:13] <Honoome> but gnu does nothing of that, so can't accomplish anything
[10:08:30] <kshishkov> Honoome: reminds me of an old joke "There are three organizations run by a bunch of senile people - Vatican, USSR and IBM"
[10:08:43] <Honoome> like idv in italy, libdems in uk, nader in usa
[10:09:07] <mru> nader lol
[10:09:09] <Honoome> kshishkov: well, vatican and ussr are not really "political" systems... aren't both one-sided anyway? :P
[10:09:18] <mru> libdems are actually in government now
[10:09:31] <Honoome> mru: and that's the first sign of the apocalypse...
[10:09:45] <mru> I think it's a good thing
[10:10:03] <mru> nick clegg is the only one of the leaders who's officially stated he doesn't believe in god
[10:10:09] <mru> the others claim they do
[10:10:13] <Honoome> yes, just so unlikely that I still haven't reformed my jokes to remind me of that fact
[10:10:29] <kshishkov> Honoome: so? Actually USSR was democratical as it's said in its constitution. But political system != presence of political parties
[10:10:31] <Honoome> mru: didn't he also add that his sons go to church because his _wife_ believes in god? :P
[10:10:41] <mru> his wife is catholic, yes
[10:10:44] <iive> Honoome: compromises with your principles always comes with a prise, that you pay much later.
[10:11:09] <Honoome> iive: and you can see how well that's working out with gnu/hurd, no?
[10:11:12] <iive> Honoome: my country during communist times also had 2 parties.
[10:11:16] <mru> either way, I hope they can temper cameron a bit
[10:12:11] <mru> Honoome: yeah, you have rms using a computer that's worse than the gdium and accesses the web by email request
[10:12:43] <iive> Honoome: hurd is bad by technical design, and not because of fsf/gnu freedom principles.
[10:13:04] <mru> the problem with hurd is they can't seem to settle on a design and go with it
[10:13:14] <mru> they already changed microkernels something like 4 times
[10:13:16] <Honoome> iive: beg you to differ, in that they rejected alternative microkernel options on the base of principles.
[10:13:27] <mru> and spent the next year porting everything
[10:14:32] <Honoome> iive: btw I sure hope you're not compromising by using a computer with a proprietary bios.. or a washing machine with proprietary firmware...
[10:14:41] <iive> Honoome: i'm not going to argue with you on a matter I have no interest or knowledge of.
[10:15:07] <iive> Honoome: i've RE the firmware of all my washing machines
[10:15:24] <kshishkov> iive: me too. All zero of them
[10:16:02] <Honoome> no cellphone or tv set either
[10:17:09] <Honoome> [yes it's an argument for absurdity.. just shows how "no compromises" approaches only work on paper, and once you actually try to get stuff done you _have_ to compromise.. and if you do it might help (libdems docet, at this point)]
[10:17:31] <iive> Honoome: the police in my country stores location report on all cellphones for 6 months.
[10:18:18] <iive> but I hear that android is getting some steam.
[10:18:36] <Honoome> and? I'm pretty sure my provider can tell what I have been doing for much more than that, just so you know...
[10:18:59] <Honoome> and android is neither entirely free nor a "no-compromises" situation, all the opposite, rather
[10:19:00] <iive> but your provider can't put you in jail.
[10:19:15] <hyc> android phones home to google all the time
[10:19:26] <Honoome> iive: corrupted police or generally bad laws have nothing to do with software
[10:19:32] <av500> iive: isp can do worse than jail, it can cut you from ze interwebs :)
[10:19:54] <hyc> unless, like me, you identify the IP addresses it tries to contact, and route them to space
[10:20:05] <iive> i still keep by 56kbps modem :P
[10:20:19] <mru> I don't care if the phone talks to google once in a while
[10:20:22] <mru> they provide a good service
[10:20:50] <hyc> when I am roaming and I need to use the data connection, I want it carrying my traffic and only my traffic.
[10:20:53] <mru> I don't trust them blindly, I'd never put my credit card details on their systems
[10:20:56] <hyc> not some background spy traffic...
[10:21:40] <hyc> at $15/MB for data roaming, it's criminal for them to use the network without your permission
[10:21:44] <mru> I do trust google to provide the service they say they do
[10:21:54] <mru> that's steep
[10:22:13] <mru> here I was moaning over vodafone raising it to £1/MB
[10:22:17] <hyc> that's what T-Mobile charges if I use their sim outside the US
[10:22:19] <mru> used to be 20p/MB
[10:22:30] <mru> within europe at least
[10:22:31] <iive> well, google can install and remove applications from phones... it happened already.
[10:22:37] <mru> oh, shock!
[10:22:51] <mru> with my old phone _I_ couldn't install or remove apps
[10:23:01] <astrange> google is probably one of the safest places to give your cc to
[10:23:03] <mru> so it's a net improvement
[10:23:16] <Honoome> maximum betrayal, they removed _A MALWARE_
[10:23:31] <mru> astrange: I wouldn't just store it in google docs or something
[10:23:37] <Honoome> who dares trying to take issue with malicious software they let through in the first place?!
[10:24:09] <hyc> I suppose in some ways it's not about trust. sure, they provide a good service. But still, they shouldn't be making expensive connections to anything without first getting explicit user permission
[10:24:21] <mru> astrange: I would of course not mind buying something from google using a cc
[10:24:48] <mru> hyc: what if you turn off background data?
[10:24:53] <hyc> that's still one of my big gripes with android. They have separate UIDs per app. they could easily implement uid-based iptables rules to control network access, but they don't.
[10:25:27] <hyc> without other blocks in place, it always calls back to the home servers looking for firmware updates, every 12 hours and every time you turn the phone on
[10:25:54] <hyc> even with all background/sync settings disabled
[10:26:30] <hyc> s/phone on/data network on/
[10:26:59] <mru> but with data off, it doesn't, I hope...
[10:27:06] <hyc> I would usually leave data roaming disabled, but if I needed to check email or use ssh I would toggle data roaming on
[10:27:17] <hyc> and then see a burst of traffic before I had even done anything
[10:27:39] * Honoome enabled data roaming when at FOSDEM... costed less than the sleepwell's wifi
[10:27:57] <hyc> drop into shell, run netstat, see a bunch of traffic to various subnets registered to google.com
[10:28:54] <hyc> so then I just put a bunch of route add net xx.xx.xx.xx gw 127.0.0.2
[10:28:58] <hyc> into a shell script
[10:29:53] <hyc> dunno how much it cost me before I finally figured out what the phone was doing...
[10:42:53] <DonDiego> how would i go about extracting all the commits done to subdir X into patches with git?
[10:43:19] <mru> git format-patch since..until subdir
[10:43:22] <mru> or something like that
[10:45:49] * DonDiego tries
[10:47:34] <DonDiego> git is not at all easy to get the hang of..
[10:48:05] <DonDiego> though i think i can see the valleys of glory behind the learning curve mountains..
[10:48:15] <elenril> meh, you're just too used to svn
[10:48:56] <DonDiego> nope
[10:49:00] <elenril> git is very easy once you read a few tutorials
[10:49:05] <DonDiego> git is just weird in places
[10:49:36] <DonDiego> and there are gazillions of commands, plus the lack of revision numbers makes some operations inherently difficult
[10:49:46] <DonDiego> i want
[10:50:02] <DonDiego> git format-patch r1 - HEAD subdir
[10:50:13] <DonDiego> now i have to figure out how to specify those revisions..
[10:50:38] <DonDiego> both mercurial and bazaar use revision numbers alongside hashes without trouble..
[10:50:38] <twice11> hmm. git log goes down to beginning.
[10:50:48] <twice11> I think you want "git log subdir"
[10:50:57] <twice11> err, "git log -p subdir"
[10:51:15] <twice11> -p shows diffs for all the displayed commits.
[10:52:09] <DonDiego> that does not give me separate diffs for importing..
[10:52:23] <DonDiego> and it likely won't apply as a monolithic diff
[10:52:38] <twice11> Ah, I get it.
[10:53:30] <elenril> err how do you use revision number with nonlinear development again?
[10:53:33] <elenril> +s
[10:53:44] <twice11> The conceptual problem of git with your use case is that there is no continous stream of revisions.
[10:54:20] <CIA-99> ffmpeg: mru * r24137 /trunk/libavcodec/ (Makefile libxvidff.c): Allow using libxvid RC without enabling encoding wrapper
[10:54:20] <CIA-99> ffmpeg: mru * r24138 /trunk/libavcodec/ (libxvid_internal.h libxvidff.c): libxvid: add some const fixing a warning
[10:54:24] <twice11> If a development model has a continous revision stream (for example order of commits on a master server), a tag like "r1" can be added on each commit to that tree.
[10:54:27] <elenril> and i don't really see the difference between searching the log for a rev number vs searching the log for a hash
[10:55:26] <mru> the one thing hashes lose are immediately obvious temporal ordering
[10:56:10] <mru> but with non-linear development rev numbers make no sense
[10:57:46] <DonDiego> i repeat: hg and bzr cope with no problem
[10:58:01] <DonDiego> you can just add (local) number aliases to hashes
[10:58:09] <DonDiego> no problem at all and a big usability win
[10:58:10] * mru recommends "the emperor's new mind" by roger penrose
[10:58:35] <mru> or for a lighter treatment, "a brief histor of time" by hawking
[10:58:52] <DonDiego> i have a perfectly linear dev history here..
[10:59:16] <mru> the relativistic view of time and causality is a good analogue for branched development
[10:59:35] <ohsix> it is indeed
[10:59:42] <mru> bitkeeper tried to use rev numbers in a distributed system
[10:59:46] <ohsix> your inertial frame or mine?
[10:59:47] <mru> it got very, very confusing
[11:00:25] <ohsix> if you want you can tag everything as you commit too, or at least between series of edits
[11:01:42] <DonDiego> apparently none of you used either hg or bzr
[11:01:51] <mru> I don't care what they do
[11:01:59] <ohsix> git isn't hg or bzr
[11:02:04] <DonDiego> it's very simple: each commit has an identifying hash
[11:02:13] <DonDiego> each hash is given a local rev number
[11:02:13] <mru> the fact is that in branched development, any kind of linear numbering is unstable
[11:02:21] <mru> my r1234 isn't the same as your r1234
[11:02:29] <DonDiego> yes, but that's not a problem
[11:02:41] <mru> what good does a volatile number like that do me?
[11:02:44] <DonDiego> when i talk to you we can speak of hashes
[11:02:46] <mru> and what about multiple branches?
[11:02:54] <DonDiego> but locally i can use rev numbers, which is handy
[11:03:12] <mru> but your r1234 today is not the same as your r1234 tomorrow
[11:03:13] <elenril> just steal x264's script if you want rev number so much
[11:03:18] <mru> time and space... all the same
[11:03:27] <ohsix> i wouln'd be surprised if theres a widget you could plug in to add rev numbers
[11:03:30] <elenril> most people apparrently don't need them
[11:03:34] <DonDiego> my rev numbers are stable, no problem
[11:03:41] <mru> elenril: the x264 script counts the revs one by one every time
[11:03:41] <elenril> until you rebase
[11:03:46] <mru> it clearly doesn't scale well
[11:04:14] <mru> DonDiego: you still live in a classical world...
[11:04:16] <mru> poor guy
[11:04:19] <av500> why not use the atomic clock UTC time as rev number :)
[11:04:34] <mru> av500: that doesn't exactly help
[11:04:38] <DonDiego> you guys are not getting my point
[11:04:42] <mru> sure we are
[11:04:45] <mru> we're just not accepting it
[11:04:49] <av500> there is no point :)
[11:04:56] <mru> spoons are not pointy
[11:04:57] <DonDiego> you are disputing that rev numbers can be useful
[11:05:06] <DonDiego> which is nonsense
[11:05:09] <mru> they are useful only in a strictly linear environment
[11:05:11] <mru> git is not linear
[11:05:19] <DonDiego> neither is hg nor bazaar
[11:05:23] <DonDiego> the world is not a git
[11:05:28] <mru> hashes are stable, sequence numbers are not
[11:05:35] <DonDiego> i know
[11:05:35] <ohsix> a rev number is less useful than a list of revisions & descriptions of them, no?
[11:05:36] <av500> DonDiego: cutting git back to one central master repo, yes, it could have a rev number
[11:05:45] <av500> back to svn
[11:05:46] <DonDiego> NOOOOOOOOOOO
[11:05:48] <mru> when the meaning of a rev number can change from one moment to the next, why should I use it?
[11:05:58] <DonDiego> look guys
[11:06:03] <DonDiego> I UNDERSTAND THAT
[11:06:05] * av500 looks
[11:06:08] <av500> ok
[11:06:11] <ohsix> you could use it in your tree
[11:06:15] <DonDiego> THIS IS NOOOOOOTTTT WHAT I WAS TALKING ABOUT
[11:06:31] <av500> ohsix: I have a swing in my tree :)
[11:06:38] <ohsix> if you're rolling patches from that instead of pushing that stuff would be left behind anyways
[11:06:55] <DonDiego> the meaning of "what time is it?" changes from one second to the next, why do you ever pose the question?
[11:07:07] <DonDiego> but *at that moment* it has uses
[11:07:13] <DonDiego> funny isn't it
[11:07:23] <DonDiego> you ask and every time you get a different answer
[11:07:26] <av500> well, year 666 is always the same
[11:07:27] <DonDiego> yet it is useful information
[11:08:07] <Tjoppen> can't you use HEAD~1 and the like?
[11:08:22] <mru> DonDiego: asking the time is only meaningful if you have a stable calendar
[11:08:36] <Tjoppen> git rebase -i HEAD~100  for instance
[11:08:37] <ohsix> even in a series of patches they're kind of silly, cherrypick the changes into a new branch and send patches from there
[11:08:53] <av500> mru: guess why my Xmas is 13 days late :)
[11:08:55] <DonDiego> nope
[11:09:04] <DonDiego> it is useful if you have a stable point of referenc
[11:09:05] <DonDiego> e
[11:09:15] <mru> which you don't
[11:09:23] <DonDiego> which i have for long enough moments in time for rev numbers to be useful
[11:09:24] <mru> where is the centre of the universe?
[11:09:32] <mru> answer: right here
[11:09:43] <DonDiego> the center of the universe is wherever i define it
[11:09:51] <DonDiego> it's like coordinate systems in physics
[11:09:53] <av500> no, it is wherever we VOTE it is!
[11:09:55] <av500> dammit
[11:09:57] <mru> every point is the centre when viewed from that point
[11:09:59] <DonDiego> physicists change them all the time
[11:10:03] <DonDiego> and THEN use them
[11:10:12] <ohsix> um
[11:10:19] <DonDiego> you tell me coordinate systems have no uses because they are arbitrary
[11:10:24] <DonDiego> *you* are missing the point
[11:10:28] <ohsix> thats not a good digression in that particular discussion
[11:10:33] * mru maps the ffmpeg commit history in spherical coordinates
[11:10:50] <av500> git -> relativity discussion -> .... -> anime
[11:10:58] <mru> profit?
[11:10:58] <elenril> DonDiego: so what's the problem, nobody's stopping you from using rev numbers
[11:11:11] <av500> just tag each commit with N+1
[11:11:17] <elenril> yeah, this
[11:11:25] <av500> in post commit hook or so
[11:11:36] <av500> that could be useful on e.g. ffmpeg master git
[11:11:42] <mru> then create some branches and watch all hell break loose
[11:12:00] <mru> that said, I'm intending to keep the master repo strictly linear
[11:12:08] <ohsix> at the moment he isn't doing it the easy way with a branch anyways
[11:12:10] <mru> so tagging each revision there could be done
[11:12:13] <av500> yes
[11:12:22] <DonDiego> ohsix: what would that way be?
[11:12:29] <av500> and ppl can do ffgit_fetch_rev xyz
[11:13:06] <ohsix> new branch, pick revs from log, possibly with browser tool like giggle; then format patch there
[11:13:52] <elenril> how do i find a host on lan if i don't know its ip address
[11:13:59] <mru> use dns
[11:14:03] <av500> ask 8.8.8.8
[11:14:17] <ohsix> ping 255.255.255.255 & arp -a
[11:14:17] <elenril> it's a dumb ap
[11:14:38] <mru> that only works if it responds to bcast pings
[11:14:54] <mru> nmap with the right flags should find it
[11:15:07] <ohsix> smurf u
[11:15:08] <kshishkov> elenril: it's _l_an, just follow the cable
[11:15:27] <mru> kshishkov: could be wireless
[11:15:37] <elenril> mru: i know, i did it once
[11:15:41] <elenril> just forgot what flags
[11:16:08] <mru> nmap -PN -T5 192.168.0.0/24
[11:16:14] <kshishkov> mru: it's called wlan then and you can use sniffer then
[11:16:23] <mru> adjust for your local network address
[11:16:38] <av500> kshishkov: or metal plate in back of head
[11:22:10] <DonDiego> \o/
[11:22:19] <DonDiego> i managed to do it with the help of sed...
[11:22:43] <av500> my favorite obj file manipulator?
[11:22:46] <mru> DonDiego: http://xkcd.com/763/
[11:23:16] <elenril> managed to do what?
[11:23:49] <mru> elenril: see link
[11:24:00] <DonDiego> that reminds me of the day i ventured to teach my mom how to play mp3s on her computer
[11:24:16] <av500> with sed?
[11:24:32] <mru> hmm, is it possible to write a dct in sed?
[11:24:33] <twnqx> someone did that?
[11:24:35] <DonDiego> when she started the endeavour by opening word (it has a file picker after all) i immediately gave up
[11:24:36] <av500> cat x.mp4 | sed -e s.... > /dev/dsp
[11:24:47] <twnqx> i started a player in awk once...
[11:24:51] <mru> maybe the sedtris girl can do it
[11:25:08] <av500> i guess sed it turning complete...
[11:25:20] <av500> is
[11:25:25] <twnqx> tuning? turing?
[11:25:35] <av500> turnip
[11:25:52] <kshishkov> twnqx: you can play Tetris there with rotating elements
[11:26:10] <twnqx> i know sedtris...
[11:26:54] <twice11> if we may put one or two a strategic xxd's in the mp4 to dev/dsp pipe, it should be possible...
[12:03:37] <lu_zero> $good $morning
[12:03:46] <lu_zero> for any improper definition of good and morning
[12:11:47] <CIA-99> ffmpeg: janne * r24139 /trunk/libavformat/nutenc.c:
[12:11:48] <CIA-99> ffmpeg: rename get_length to ff_get_v_length and put_v to ff_put_v
[12:11:48] <CIA-99> ffmpeg: patch by Michael Chinen < mchinen gmail >
[12:15:23] <CIA-99> ffmpeg: janne * r24140 /trunk/libavformat/ (aviobuf.c nutenc.c internal.h):
[12:15:23] <CIA-99> ffmpeg: move ff_get_v_length and ff_put_v from nutenc.c to internal.h/aviobuf.c
[12:15:23] <CIA-99> ffmpeg: patch by Micheal Chinen < mchinen gmail >
[12:16:47] <Tjoppen> grr.. either WMP sucks ass, or mainconcept does, or both
[12:16:54] <kshishkov> both
[12:17:20] <Tjoppen> I creave a working .vc1 file using their sdk, remux that to wmv. plays fine in ffplay and vlc, but of course not in wmp
[12:17:29] <mru> mainconcept is a frankencorp
[12:17:46] <mru> the elecard guys are smart
[12:18:05] <mru> the german part not so much
[12:18:10] <kshishkov> mru: smart - yes, but they are Russians
[12:18:10] <Tjoppen> their api is even more ass. took me the better part of two days to figure out how to get the extradata out. doesn't even seem to be possible for the h264 encoder
[12:18:29] <av500> Tjoppen: the lavf asf muxer produces files that wmp refuses to play
[12:18:35] <mru> kshishkov: peter gubanov is a generally nice guy
[12:18:39] <mru> even though he's russian
[12:18:56] <Tjoppen> av500: that's.. interesting. I tried .avi though, which didn't work either
[12:18:58] * thresh ahems
[12:19:12] <mru> thresh: just checking if you're awake
[12:19:28] <thresh> I was on a meeting playing 'bullshit bingo'
[12:19:39] <thresh> won three times, then got bored
[12:19:53] <Honoome> rotfl
[12:22:46] * Honoome found a bug in ruby ...
[12:22:57] <mru> hardly surprising
[12:23:47] <kshishkov> Honoome: besides ruby itself?
[12:24:21] <lu_zero> kshishkov: it was a a pythonism
[12:24:50] <Honoome> now of course the problem was that I had no connectivity from within the container to the rest of the world (stupid modem)
[12:25:12] <Honoome> but the fact that instead of telling me "I cannot connect", it exploded for an "invalid method" doesn't really float my boat
[12:25:51] <av500> container?
[12:26:03] <av500> you are trying to leave europe illegaly?
[12:27:02] <kierank> has this been posted yet: http://varnish-cache.org/docs/phk/autocrap/
[12:27:25] <av500> yes
[12:27:44] <av500> [11:44] <mru> http://varnish-cache.org/docs/phk/autocrap/
[12:27:46] <Honoome> av500: lxc: linux-containers
[12:27:56] <mru> some discussion: http://news.ycombinator.com/item?id=1499738
[12:28:06] <Honoome> I'll reply with http://www.flameeyes.eu/autotools-mythbuster/ :P
[12:28:19] <kshishkov> Honoome: sounds like Java 2 EE - e.g. not the thing I'd like to touch
[12:28:35] <av500> Honoome: I tried to read that once....
[12:29:23] <kshishkov> tl;dr
[12:29:30] <mru> no amount of mythbusting can alter the hard fact that every version of each autotool is incompatible with every other version and often with itself
[12:29:57] <Honoome> mru: true.. that can only be fixed with hard-bashing the authors in the head with a clue-by-four
[12:30:17] <Honoome> otoh people who diss autotools end up suggesting cmake and scons.. both of which have the same or worse problems
[12:30:45] <wbs> ffmpeg configure has to be the best build system I've seen this far
[12:30:50] <av500> I diss autotools for checking whether I have a fortran compiler
[12:30:52] <DonDiego> i'll go on record to say that Honoome's autotools-mythbuster is excellent documentation
[12:31:00] <DonDiego> it helped me a lot at work
[12:31:02] <av500> so its not a mythbuster
[12:31:26] <wbs> if it was packaged for easy reuse with other projects, it would be even better. :-)
[12:31:34] <av500> ffconf
[12:31:37] <DonDiego> Honoome: since you're not into beer, remind me to buy you dinner next time we meet :)
[12:31:49] <av500> i'll drink the beer
[12:31:58] <spaam> DonDiego: date with Honoome ? :)
[12:32:15] <Honoome> gha now the situation is becoming embarassing :P
[12:32:15] <ohsix> if only projects wouldn't resign themselves to dead eras when they play around with the implementation language and the build system
[12:32:52] <Honoome> av500: "Autotools Mythbuster" was the title I used for a few blog posts bashing broken autotools... I reused the name for the guide
[12:33:17] <Honoome> to be honest, it's quite lacunous as it is, but I work on it on spare tim
[12:33:24] <twice11> av500: autotools checks for fortran if you include libtool in your build process. You can limit the languages libtool should support and remove the fortran test by that.
[12:33:35] <av500> me? why me?
[12:33:40] <Honoome> twice11: done already on lt2
[12:33:55] <Honoome> twice11: libtool-2 moved to NOT test for C++ or Fortran by default
[12:34:02] <av500> all I do is to untar random.tgz and do ./configure
[12:34:18] <Honoome> actually broke a few projects because they relied on the C++ compiler to be tested for by libtool
[12:34:31] <twice11> Sometimes even these bloat-o-matic projects seem to do something sensible...
[12:34:48] <Honoome> twice11: that's why I went with the name "Mythbuster"
[12:35:05] <Honoome> because a lot of their bad name is for veeeery old and crappy stuff
[12:35:22] <twice11> OTOH, if AC_PROG_CXX is used, it seems sensible to assume that I want C++...
[12:35:38] <Honoome> or because the writers of configure.ac or Makefile.am used something that worked only out of sheer luck
[12:36:10] <twice11> In my oppinion, all the automatic configuring scripts suck the one or the other way.
[12:36:12] <Honoome> *cough* kde *cough*
[12:36:23] * kshishkov checks whether he has Fortran 77 compiler to read that document by Honoome
[12:36:42] <Honoome> just to say, autoconf 2.64 removed an internal, never-documented, now-unused macro...
[12:36:50] <twice11> Didn't KDE switch to cmake because they were unhappy with auto*?
[12:36:53] <Honoome> 2.65 had to add it back _because it totally fucked up the kde 3 build system that releid on that_
[12:37:09] <Honoome> twice11: they switched because they had no clue about auto* and made the worst mess with it I have seen in a very long time
[12:37:38] <mru> isn't it rather telling when _nobody_ is able to use something properly?
[12:37:43] <ohsix> no clue is a running theme \m/
[12:37:56] <Honoome> mru: yes, that the documentation for auto* sucks
[12:38:07] <Honoome> that the autobook has been left alone for so many years that it's nowhere near covering what it should
[12:38:09] <mru> when using the tool takes more skill and effort than doing without it, one should reconsider the choice of tools
[12:38:15] <Honoome> that the official info pages are badly written
[12:38:30] <mru> and they keep changing the bloody syntax every version
[12:38:34] <Honoome> I have no idea about the nostarch press book sincerely, I haven't read that yet, I started a-m before it was published
[12:38:42] <mru> anyone ever tried to build newlib?
[12:38:48] <mru> it has the worst autohell I've ever seen
[12:38:56] <Honoome> mru: newlib, gcc and binutils are bad.. not as bad as kde3, but bad
[12:39:03] <mru> it can't work with a single autohell version
[12:39:28] <mru> you have to fuck around and build some scripts with one version, some others with another, then rebuild the first ones with a third
[12:39:32] <mru> or something like that
[12:40:17] <ohsix> i crossed for arm w/newlib on ubunto; worked fine
[12:40:42] <mru> I never managed to get the autohell to run even once
[12:41:24] <ohsix> they have a wrapper that people can request given versions with
[12:41:36] <mru> so does gentoo
[12:41:40] <Honoome> I'm not saying that autotools are properly done
[12:41:47] <mru> the problem with newlib is that _no single version_ works
[12:42:07] <Honoome> but I still have to see something that solves all the problems without adding worse
[12:42:30] <Honoome> beside custom-writing the scripts... and even that is not going to solve problems without adding worse unles who writes it has a clue
[12:42:40] <Honoome> and trust me, that would limit the number of people A LOT
[12:42:43] <mru> the solution is in not attempting to solve everything
[12:42:54] <ohsix> all i had is 2.13, it worked
[12:43:23] <Honoome> mru: *shrug* to me the solution is solving the worst problem with autotools: documentation that is crappy
[12:43:39] <mru> you know my solution
[12:44:10] <Honoome> you can afford to write your own build system.. most people don't..
[12:44:10] <mru> use a custom build system solving exactly what it needs to solve for each app
[12:44:14] <Honoome> or at least they should learn NOT to...
[12:45:04] <ohsix> if i wanted to use host triplets and automatic replacement stuff i wouldn't write my own D:
[12:45:32] <ohsix> you end up supporting each target instead of just making large swaths of them look same enough
[12:45:40] <mru> I've always wondered how something with _four_ parts can be called a _triplet_
[12:45:56] <mru> and it's all so gnu-centric anyway
[12:45:59] <ohsix> one part is optional
[12:46:24] <ohsix> but its replaced by unknown if it isn't there, huhuhuhu
[12:46:36] <mru> what's latin for 3.5?
[12:47:21] <Honoome> ohsix: uh? no
[12:47:34] <Honoome> it's only where gnu is an option that it's four
[12:47:38] <ohsix> the name is grandfathered from when there was only 3 or something; like gecos
[12:48:10] <Honoome> x86_64-unknown-linux-gnu x86_64-unknown-linux-uclibc x86_64-unknown-freebsd8.2 x86_64-unknown-kfreebsd-gnu
[12:48:14] <Honoome> see the pattern?
[12:48:22] <ohsix> that sounds squishy, is there an rfc for it
[12:48:23] <mru> I see the pattern
[12:48:29] <mru> but it only works with gnu tools
[12:55:57] <lu_zero> mru: I had to play with newlib
[12:56:04] <lu_zero> and binutils was worse
[12:56:47] <mru> binutils, gcc, newlib, and glibc are all horrible
[12:56:55] <mru> each in their own twisted way
[12:57:14] <lu_zero> still I'm a _tad_ afraid about llvm
[12:58:06] <lu_zero> btw ffconf might be interesting
[13:21:14] <lu_zero> hi BBB
[13:21:20] <BBB> hello
[13:50:12] <mru> Vitor1001: ping
[13:50:27] <Vitor1001> pong
[13:50:55] <mru> since you're working on tests, do you have permissions to put the samples in place?
[13:53:10] <Vitor1001> mru: Yes, but since mike worked hard to keep fate-suite/ small, I'm not cluttering it with samples for tests not yet ok-ed.
[13:54:02] <mru> just checking if there's anything else you need to make your work smooth
[13:54:48] <Vitor1001> thanks
[14:07:44] <DonDiego> how do i set up global ignore patterns with git?
[14:07:53] <av500> .gitignore
[14:07:54] <DonDiego> the manual is not helpful..
[14:08:01] <DonDiego> *global*, not local
[14:08:13] <av500> for every git in the world?
[14:08:18] <elenril> what do you mean by global?
[14:08:32] <DonDiego> for every checkout/repo that my git ever touches
[14:08:52] <mru> why would you want to do that?
[14:09:08] <DonDiego> i was just going to say you would now tell me i don't need that feature..
[14:09:11] <elenril> Patterns read from $GIT_DIR/info/exclude.
[14:09:15] <DonDiego> git fanboys..
[14:09:17] <elenril>  Patterns read from the file specified by the configuration variable core.excludesfile.
[14:09:21] <elenril> ^tfm
[14:09:30] <DonDiego> been there, done that
[14:09:50] <av500> DonDiego: gitaards
[14:10:17] <DonDiego> dbiurrun at spielwiese ~ $ cat .gitconfig
[14:10:18] <DonDiego> [core]
[14:10:18] <DonDiego> excludesfile = ~/.gitignore
[14:10:18] <DonDiego> dbiurrun at spielwiese ~ $ cat .gitignore
[14:10:18] <DonDiego> *.o
[14:10:52] <av500> err
[14:10:56] <DonDiego> dbiurrun at spielwiese ~/src/LIBRARYUPDATEINDICATOR $ git status
[14:10:56] <DonDiego> # On branch master
[14:10:57] <DonDiego> # Untracked files:
[14:10:57] <DonDiego> #   (use "git add <file>..." to include in what will be committed)
[14:10:57] <DonDiego> #
[14:10:57] <DonDiego> #       client/client.o
[14:11:05] <DonDiego> that didn't work all that well...
[14:11:18] <cartman> DonDiego: You might need to replace ~ with absolute path
[14:12:22] <DonDiego> core.excludesfile
[14:12:22] <DonDiego>     In addition to .gitignore (per-directory) and .git/info/exclude, git looks into this file for patterns of files which are not meant to be tracked. "~/" is expanded to the value of $HOME and "~user/" to the specified user's home directory. See gitignore(5).
[14:12:31] <DonDiego> this is from git-config(5)
[14:12:39] <cartman> ~ didn't work for me
[14:12:43] <cartman> just saying...
[14:12:44] <DonDiego> gitards indeed..
[14:13:03] <mru> DonDiego must be consciously _trying_ to make git hard
[14:13:58] <DonDiego> full path works, the manual is lying
[14:14:07] <elenril> send a bugreport?
[14:14:29] <mru> maybe the expansion is supposed to be done by some system function that's broken on your machine
[14:14:32] <mru> you run debian, right?
[14:15:17] <cartman> DonDiego: didn't work on MacOSX either
[14:15:18] <cartman> git bug
[14:15:37] <DonDiego> this is a centos machine at work in fact
[14:15:44] <mru> lol centos
[14:16:11] <Tjoppen> <vacation>
[14:16:40] <mru> all the crap of redhat without the paid entirprisey stuff
[14:16:46] <mru> absolutely no reason to use
[14:16:54] <cartman> if you can't expand ~, well...
[14:17:30] <mru> whoa there, where the heck did cartman come from?
[14:17:36] <cartman> *magic*
[14:17:40] <mru> cartman: you haven't been here for an eternity
[14:17:45] <cartman> mru: true
[14:17:48] <mru> welcome back
[14:17:51] <cartman> turns out that I am still alive
[14:17:55] <cartman> thank you mru !
[14:27:54] <DonDiego> yo gitards!
[14:28:03] <DonDiego> git-svn is failing to remove directories here..
[14:28:24] <mru> never seen that
[14:29:51] <lu_zero> DonDiego: it hates you
[14:30:06] * mru suspects the feeling is mutual
[14:30:27] <lu_zero> it feels your fear and behaves accordingly
[14:30:48] * lu_zero would expect to read the news about DonDiego being mauled by git
[14:31:09] <lu_zero> jokes aside, what were you trying to do?
[14:31:59] <elenril> >git svn failing
[14:32:02] <elenril> sounds about right
[14:37:24] <markuman> i've got a question to the "missing picture in access unit" message. does parse_nal_units search for a pair of frame? http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2009-July/073538.html
[15:13:41] <BBB> mru: libavfi depend on libavco ok with you? sounds about right to me to be honest
[15:14:16] <mru> I think another lib is an execise in work creation
[15:15:04] <twice11> libavfilter and libavcodec both alreasy exist, or did I get something wrong?
[15:20:51] <siretart> TBH, I wouldn't mind too much if libavutil was merged back into libavcodec. but I guess there were good reasons for splitting lavu off.
[15:21:21] <mru> we had our reasons
[15:21:30] <mru> I don't quite remember what they were
[15:22:53] <av500> were colours involved?
[15:24:11] <mru> a full spectrum, I'm sure
[15:24:29] <twice11> In america, only colors could have been involved. Better ask for bikesheds, I hope they are spelled the same everywhere.
[15:25:09] <av500> twice11: fringe usage of english dialects is irrelevant :)
[15:25:38] <av500> but, let's not argue, we can vote over it....
[15:25:48] <mru> reminds me of the time I called windows a "fringe os"
[15:25:55] <mru> apparently that annoyed some people
[15:26:55] <av500> the truth is always annoying to some
[15:31:23] <BBB> mru: I share your opinion, I understand it was split off, I respect adn appreciate it, but I wouldn't mind at all if it was merged back
[15:31:52] <BBB> I don't think there's a wide demand for a "general utility lib", where "general utility" is defined as "random stuff used in libavcodec/libavformat"
[15:32:14] <BBB> which is pretty much libavutil
[15:32:47] <mru> it's reasonable to avoid libav* depending the huge libavcodec unless necessary
[15:32:59] <mru> that's pretty much why lavu was created
[15:33:27] <mru> it wasn't even michael's idea
[15:34:35] <siretart> that would make sense if there was some (important) lib that depends on lavu but not on lavc. what's that?
[15:34:45] <av500> lavf?
[15:34:49] <mru> libsws
[15:35:01] <mru> lavfi
[15:35:01] <av500> no, lavf depedns on lavc
[15:35:09] <av500> important?
[15:35:13] <mru> libpostproc
[15:35:25] <Honoome> libxine depends on libavutil only
[15:35:36] <siretart> ok
[15:35:42] <mru> important?
[15:35:43] <Honoome> [the ffmpleg plugin uses libavutil, but the main lib in 1.2 branch uses libavutil "standalone"]
[15:35:48] * av500 discovers libpostproc in his svn checkout
[15:36:09] <Vitor1001> mru: What do you think of this: http://ffmpeg.pastebin.com/vbYQaLTi ?
[15:36:26] <siretart> Honoome: what's the status of 1.2 branch? is it eventually going to get released?
[15:36:37] <Vitor1001> mru: There is a very simple way of getting MAXDIFF == 1: outputting 0 samples.
[15:36:46] <mru> I noticed
[15:40:27] <Vitor1001> mru: BTW, even without your patches, I'm getting on your PPC box:
[15:40:29] <Vitor1001> TEST FATE   twinvq
[15:40:29] <Vitor1001> Could not open input files.
[15:40:30] <Vitor1001> /home/vitor/ffmpeg/tests/fate-run.sh: line 21: [: -gt: unary operator expected
[15:40:35] <Vitor1001> s/your patches/my patch/
[15:41:18] <Honoome> siretart: you got to ask ds for that :/ I really am totally out of the loop
[15:42:50] <siretart> same to me
[15:44:17] <mru> Vitor1001: do you have the ref file?
[15:44:47] <Vitor1001> No, it is a oneoff
[15:45:00] <mru> where are you pointing SAMPLES?
[15:45:16] <Vitor1001> Oh, sorry, my bad.
[15:45:49] <Vitor1001> I was pointing to samples, but no the good directory.
[16:06:46] <CIA-99> ffmpeg: mru * r24141 /trunk/libavcodec/resample2.c: resample: check memory allocations in av_resample_init()
[16:06:46] <CIA-99> ffmpeg: mru * r24142 /trunk/libavcodec/resample2.c: resample: replace VLA with malloc/free
[16:06:49] <CIA-99> ffmpeg: mru * r24143 /trunk/tests/ (13 files in 2 dirs):
[16:06:49] <CIA-99> ffmpeg: tiny_psnr: skip wav headers on input files
[16:06:49] <CIA-99> ffmpeg: The byte count printed excludes the header, and offsets are applied
[16:06:49] <CIA-99> ffmpeg: after the the headers are skipped.
[16:06:50] <CIA-99> ffmpeg: Reference files updated to reflect new output. Some stddev/psnr values
[16:06:50] <CIA-99> ffmpeg: have changed slightly due to headers no longer being compared.
[17:11:08] <CIA-99> ffmpeg: victor * r24144 /trunk/libavcodec/libxvidff.c: Current Cygwin has mkstemp().
[17:20:37] <CIA-99> ffmpeg: vitor * r24145 /trunk/tests/fate2.mak: SIPR tests for "make fate2"
[17:26:22] <mru> Vitor1001: don't forget to upload the ref files to mphq
[17:27:41] <Vitor1001> mru: Already done
[17:29:04] <Vitor1001> bbl
[18:12:08] <dgt84> The mailing list flamewar is heating up...
[18:12:35] <dgt84> This is getting a bit ridiculous
[18:14:01] <mru> you say?
[18:15:10] * mru wonders what he did to anger baptiste so
[18:15:28] <mru> I was always defending him when he and michael were fighting
[18:15:40] <mru> I'm starting to think that was a mistake
[18:15:49] <jai> regarding libavutil, it would have been awesome if the bitstream read/write stuff was in libavutil
[18:16:23] <mru> there is one technical reason for having it in lavc: it's faster there
[18:16:50] <mru> function calls across shared libs are slower than intra-lib calls
[18:17:20] <mru> I agree otherwise
[18:17:25] <Compn> mru : does baptiste still harbor any anger because of the changes he has to keep forked ?
[18:17:34] * Compn rememebers someone is upset and has his own fork
[18:17:37] <mru> none of that was my fault
[18:17:47] <mru> as I said, I WAS ON HIS SIDE
[18:19:23] <jai> mru: ah, didnt think about that
[18:19:29] <jai> thx for the explanation
[18:19:39] <jai> right now i just copypaste parts of it
[18:23:06] <neerfri> hi guys ! I get no answer in the fmpeg channel, mind if I throw in a question ?
[18:23:21] <mru> why would we answer here?
[18:23:27] <neerfri> avlib programming related
[18:23:52] <neerfri> have been banging my head on this for couple a days...
[18:24:03] * mru hands neerfri a hammer
[18:24:21] <neerfri> all is fine, but av_encode_video always returns 0
[18:24:47] <neerfri> and audio is working fine, but when using interleaved writing nothing gets written...
[18:25:29] <neerfri> mru: because you have a heart of gold ? :-)
[18:26:03] <mru> not anymore, I sold it on ebay
[18:26:08] <neerfri> :-D
[18:26:45] <neerfri> know how much is it for a golden tooth ?
[18:27:12] <mru> quite some pain at the dentist I'd imagine
[18:27:36] <kshishkov> put some anesthesy on your wallet first
[18:28:13] <neerfri> so, how about helping me with the root treatment... ? I'm so sure it's tiny...
[18:28:32] <neerfri> I just lost every creativity I might have had trying to fight this...
[18:28:58] <neerfri> I'm out of ideas, have been comparing my code to ffmpeg.c's code for like a whole day...
[18:30:01] <lu_zero> neerfri: try to simplify/hardwire ffmpeg.c till you get it doing what you want
[18:30:15] <neerfri> I sort of did.
[18:30:16] <kshishkov> or look at output_example.c first
[18:30:23] <neerfri> I have a 97% done app
[18:30:30] * lu_zero assumed that was done already
[18:30:40] <neerfri> I'm way after output_example.c... :-)
[18:30:41] <mru> neerfri: here, take these: %%%
[18:30:48] <neerfri> :-D
[18:31:46] <neerfri> mru: you're shooting them in chains are you.... :-)
[18:51:51] <mru> can someone please kill baptiste for me?
[18:52:23] <kshishkov> why?
[18:52:31] <mru> he's being a total jerk
[18:52:42] <kshishkov> just take a vacation off the ffmpeg-devel ML
[18:52:49] <hjb> what's new?
[18:53:07] <hjb> I'm waiting for baptiste to start answering questions with monosyllabic grunts
[18:57:43] <mru> mini-poll: have I said anything offensive to baptiste in the last few days other than maybe in the last hour?
[19:02:04] <Dark_Shikari> 6: stop with the polls
[19:02:48] <mru> I was asking for people's opinion
[19:02:59] <mru> not asking for a vote over facts
[19:03:45] <mru> I honestly don't understand how baptiste can claim to be offended by something I said
[19:04:03] <mru> none of the "discussions" have been about his areas of code
[19:04:21] <mru> he has barely participated in them
[19:04:36] <mru> I think he said something silly in one of the voting threads, that's all
[19:04:38] <hjb> I'd say calling somebody out in IRC would probably offend them
[19:04:58] <mru> hjb: what did I do?
[19:05:29] <hjb> I don't want to get involved, man
[19:05:49] <hjb> (so I probably should have left out the snarky comment in the first place)
[19:05:52] <hjb> my bad
[19:05:52] <mru> just tell me what I said that was offensive
[19:05:57] <mru> seriously, I want to know
[19:05:58] <hjb> no farkin clue
[19:06:12] <mru> so don't fuckin' go implying I did then
[19:06:22] <jai> mru: maybe he took offense at the "jealousy" part
[19:06:37] <mru> he claims to have been offended prior to that
[19:06:49] <jai> ah
[19:06:54] * jai shrugs :)
[19:06:57] <mru> and he says I'm generally a bad person or something to that effect
[19:08:19] <mru> oh, I did reply in slightly sarcastic tone when he declared my and Dark_Shikari's "votes" void
[19:09:51] <markuman> i've got a question to the "missing picture in access unit" message. does parse_nal_units search for a pair of frame? http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2009-July/073538.html
[19:27:57] <CIA-99> ffmpeg: mru * r24146 /trunk/tests/tiny_psnr.c: tiny_psnr: use abs()
[19:28:26] <Dark_Shikari> I love the argument over the hypothetical system without abs
[19:30:21] <mru> there is no argument
[19:30:30] <mru> funman found one though
[19:30:34] <Dark_Shikari> lol
[19:30:52] <mru> on some particular version of hp-ux, if you include math.h before stdlib.h, you get no abs()
[19:34:00] <mru> but we applied the patch with abs()
[19:34:06] <mru> and everybody is happy
[20:03:10] <_av500_> mru: i dont think it makes sense to start digging who said what when
[20:03:26] <_av500_> over irc and the ML
[20:04:00] <mru> I'm just curious whether anyone else interpreted anything I said as offensive
[20:04:32] <mru> I'm not afraid to offend, but I want to choose when I do it
[20:22:17] <lu_zero> BBB: I broke rtpdec_h264 =P
[20:25:37] <lu_zero> actually av_base64_decode seems broken
[20:25:48] <capitao_azeitona> Hi people! someone here knows how to install ffmpeg-php ?
[20:26:30] <mru> no
[20:44:15] <peloverde> Is anyone wanting on (specific) things from me? I've had a hectic week but think I'm al caught up
[20:45:05] * elenril wants aacenc to not suck ;)
[20:45:17] <mru> peloverde: what is the best way to go about testing aacdec?
[20:46:05] <peloverde> An off-by-one or SNR based approach to a refernce rendering is the only way I can think of
[20:46:30] <mru> well, we have an off-by-one testing facility now
[20:46:38] <peloverde> we do?
[20:46:42] <mru> since yesterday
[20:48:10] <_av500_> janneg: quick, upload our BBB: http://youtube-global.blogspot.com/2010/07/whats-bigger-than-1080p-4k-video-comes.html
[20:48:32] <peloverde> ok, I will look at vqf and propse something based on that
[20:49:06] <lu_zero> wonderful
[20:49:25] * lu_zero found 3 bugs with a single test...
[20:49:32] <mru> that's efficiency
[20:49:38] <mru> can you fix them with a single patch?
[20:50:01] <peloverde> I think for now I will try to pick a subset of streams taht cover a wide range of bitstream features
[20:50:26] <mru> yeah, do that
[20:50:31] <lu_zero> mru: 2 bugs are in ffmpeg one in feng
[20:57:17] <capitao_azeitona> someone here uses ffmpeg-php
[20:57:18] <capitao_azeitona> ?
[20:57:24] <mru> no
[20:57:40] <capitao_azeitona> thanks..
[21:00:21] * lu_zero is puzzled
[21:00:28] <lu_zero> idx>=FF_ARRAY_ELEMS(map2)
[21:00:38] <lu_zero> idx being (unsigned)-3
[21:01:28] <lu_zero> how's possible it's valued false?
[21:02:25] <lu_zero> FF_ARRAY_ELEMS(map2) should be 80
[21:03:24] <BBB> lu_zero: fix it ;)
[21:05:41] <lu_zero> BBB: done
[21:05:56] <BBB> h264rtp?
[21:05:58] <BBB> that was quick
[21:18:45] <Dark_Shikari> BBB: got the normal loopfilter done yet? =p
[21:24:55] <mt> hi all
[21:28:41] <lu_zero> BBB: pretty much
[21:29:05] <lu_zero> once I managed to figure out why print index-something didn't do what I expected in gdb...
[21:29:53] <mt> I'm working on converting wma pro to fixed point (for rockbox), I'm almost done, but run_level_decode seems to be producing an error (Not 100% sure though, still investigating), is there something obviously wrong with this : http://www.pastie.org/1036828 ? (I'm saving the samples in Q.16 format and the level table is just the same as the float one but of type int32_t)
[21:29:57] <lu_zero> now I should find enough will to reread the h264 packetizer in feng and find out what's wrong with it for non avc1 packets
[21:33:49] <lu_zero> apparently I didn't get the startcode correctly -_-
[21:38:03] <mru> mt: don't you know the val = abs^sign - sign trick?
[21:39:01] <mru> mt: I would suggest some good old-fashioned printf debugging
[21:39:16] <mru> print the values read in both float and int versions
[21:39:17] <mru> compare
[21:40:30] <BBB> Dark_Shikari: running it in a debugger, there's a (few?) bug(s) that make it not bitexact, working on it
[21:42:10] <mt> mru:  Did so but couldn't find differences, at least not in the first few samples. Will try to go further. Thanks.
[21:42:36] <mru> that's how I'd do it...
[21:42:54] <mru> if the values are the same, the problem must be elsewhere
[21:45:28] <Dark_Shikari> I fucking love xiph.  It's 2010 and they're telling people to do VFR by adding null packets.
[21:45:44] <mru> lol
[21:45:50] <mru> in what container?
[21:45:52] <Dark_Shikari> Ogg.
[21:45:59] <mru> wtf
[21:46:02] <Dark_Shikari> The Ogg Theora mapping is CFR-only
[21:46:02] <mru> ogg has timestamps
[21:46:05] <mru> oh
[21:46:07] <Dark_Shikari> I AM NOT MAKING THIS UP
[21:46:09] <mru> *facepalm*
[21:46:20] <Dark_Shikari> They are seriously telling people to implement it with null packets
[21:46:24] <Dark_Shikari> of course....
[21:46:26] <Dark_Shikari> the null packets break
[21:46:30] <Dark_Shikari> MPC (so probably ffmpeg)
[21:46:31] <Dark_Shikari> and VLC
[21:46:46] <Dark_Shikari> so two separate independent muxers are both broken because of them
[21:46:50] <Dark_Shikari> er, demuxers
[21:47:02] <mru> what do the null packets look like?
[21:47:12] <Dark_Shikari> I think they're 0 byte packets
[21:47:15] <Dark_Shikari> Just empty
[21:47:28] <mru> packets or pages?
[21:47:31] <Dark_Shikari> Dunno
[21:47:42] <Dark_Shikari> I don't know how libtheora does it.
[21:48:04] <mru> a null _packet_ in ogg would be simply a 0 in the roman-style packet size list
[21:48:44] <mru> don't know if that's allowed by the spec
[21:49:24] <Yuvi> yes, that's how it's done, for dropped frames too
[21:49:26] <Dark_Shikari> they say whatever they're doing is allowed
[21:49:56] <Yuvi> ffmpeg works since it drops size 0 packets
[21:50:09] <mru> how does our demuxer handle them?
[21:50:13] <mru> I can't remember
[21:50:23] <Yuvi> returns them, then they're discarded in utils.c iirc
[21:50:25] <mru> I don't remember thinking about it
[21:50:42] <mru> the way it's made, it ought to simply return them
[21:51:27] <Yuvi> I've been meaning to to drop them in the demuxer (unless whatever flag is related to them is set) but that means you have to figure out the timestamps of the actual packets
[21:52:00] <mru> if it works, I wouldn't bother
[21:52:35] <mru> anyone stupid enough to mux like that deserves whatever tiny performance hit it might give
[21:53:00] <Yuvi> one fun bit is that it's the same way they use to drop frames, and one complaint against ffmpeg is that we don't create duplicate frames for that case
[21:53:15] <mru> wtf
[21:53:16] <Yuvi> even though it's impossible to determine whether the 0 byte packet means VFR or dropped frame
[21:53:32] <Dark_Shikari> it's because they don't believe in vfr
[21:53:33] <mru> for display purposes there is no difference
[21:53:36] <Dark_Shikari> so obviously it means a dropped frame
[21:57:11] <Yuvi> mru: yep, and at the same time they complain about players that don't drop frames for the vfr case (everything that doesn't use lavf I believe ;)
[22:00:38] <mru> geez, he's still cranky over those unprototyped functions
[22:00:46] <Compn> oh carl meant denial of service, not dos dos
[22:00:47] <Compn> ;\
[23:17:10] <peloverde> how do I set SAMPLES for fate?
[23:18:01] <mru> configure --samples=/path/to/fate-suite
[23:18:08] <mru> or make SAMPLES=/path/...
[23:18:30] <peloverde> thanks
[23:18:38] <mru> anyone want to make a few bucks telling someone how to use ffmpeg?
[23:18:51] <mru> encoding for iphone
[23:18:53] <Dark_Shikari> sure
[23:19:49] <mru> they probably just need someone to tell them what settings to use
[23:19:53] <Dark_Shikari> that's fine with me
[23:19:59] <mru> ok, I'll forward it
[23:26:56] <peloverde> mru: The .pcm files are interleaved?
[23:27:14] <mru> they are whatever ffmpeg outputs
[23:27:22] <mru> so yes, interleaved
[23:27:23] <CIA-99> ffmpeg: mru * r24147 /trunk/tests/fate-run.sh: fate: exit with error immediately if test command fails
[23:27:25] <CIA-99> ffmpeg: mru * r24148 /trunk/tests/tiny_psnr.c: tiny_psnr: check fread() return values
[23:27:26] <CIA-99> ffmpeg: mru * r24149 /trunk/tests/tiny_psnr.c: tiny_psnr: change error exit code from -1 to 1
[23:36:56] <CIA-99> ffmpeg: mru * r24150 /trunk/tests/fate-run.sh: fate: delete output files from successful tests
[23:38:07] <CIA-99> ffmpeg: mru * r24151 /trunk/Makefile: Remove dead target from .PHONY list
[23:42:38] <peloverde> is fate2.mak supposed to be in some sort of order?
[23:42:45] <mru> not really
[23:42:49] <mru> group as you see fit
[23:42:54] <peloverde> ok
[23:43:04] <mru> I suggest keeping related tests together
[23:44:09] <peloverde> agreed, I just wasn't sure if it was supposed to be alphabetical but needed to be fixed or something like that
[23:49:06] <peloverde> Is fate supposed to spam weird shell errors on top of its own errors?
[23:49:47] <mru> no
[23:50:00] <mru> what are you getting?
[23:50:05] <peloverde> "[: 34: -gt: unexpected operator"
[23:50:20] <mru> that means your ref file is missing
[23:51:34] <peloverde> yes
[23:51:44] <mru> maybe I should fix that...
[23:55:22] <mru> done
[23:56:05] <CIA-99> ffmpeg: mru * r24152 /trunk/tests/fate-run.sh: fate: fail with a sensible error message if reference file is missing
[23:56:49] <peloverde> thanks
[23:58:19] <peloverde> Does anywhere in this string tell me where the max error falls? "stddev:    0.57 PSNR:101.15 MAXDIFF:    2 bytes:   727040/   727040"
[23:58:31] <mru> MAXDIFF
[23:58:39] <mru> value is 2
[23:58:46] <mru> so you need FUZZ = 2
[23:59:32] <mru> see commit message for r24133
[23:59:44] <mru> perhaps not the best place for documentation