[FFmpeg-devel-irc] IRC log for 2011-01-27

[irc at mansr.com]

[00:45:45] <Dark_Shikari> michaelni: sorry, I think I flamed too hard
[00:50:00] <michaelni> Dark_Shikari, throw as much at me as you like, ive a thick skin after this
[00:50:20] <Dark_Shikari> lol
[00:50:38] <Dark_Shikari> in short: I think it would be better if you offered to give something as well.
[00:50:45] <michaelni> i really am a bit upset after i read the secret logs
[00:50:48] <Dark_Shikari> For example, "I'll resign as BDFL officially, but only if <X> <Y> <Z>."
[00:50:57] <Dark_Shikari> As opposed to "I want <X> <Y> <Z> now!"
[00:51:01] <michaelni> i hoped to find libel from diego and mans but none of that
[00:51:18] <Dark_Shikari> I think you overestimate how much people cared about mans...
[00:51:25] <michaelni> i wrote this mail really quickly ...
[00:51:27] <Dark_Shikari> I didn't even know until after I agreed to get involved that mans was coming back.
[00:51:55] <Dark_Shikari> oh also, I sent you (and the others) an email asking how you'd like your x264 LLC shares paid out, since we just hit profitability.
[00:52:13] * michaelni remembers no mail
[00:52:24] <michaelni> when was that?
[00:52:27] <Dark_Shikari> This minute.
[00:52:35] <michaelni> ok then it makes sense :)
[00:52:39] <Dark_Shikari> lol
[00:52:42] <michaelni> ill look later
[00:52:52] <Dark_Shikari> no problem, no rush.
[00:53:03] <Jumpyshoes> :O x264 LLC is now profitable?
[00:53:12] <Dark_Shikari> Total income so far: $37750
[00:53:16] <Dark_Shikari> Total costs so far: $25835
[00:53:20] <Jumpyshoes> :o
[00:53:21] <Jumpyshoes> nice
[00:53:33] * Dark_Shikari pats his trusty spreadsheet.
[00:54:15] <Dark_Shikari> michaelni: btw, I think you should set yourself a goal of something you want to do.  right now you seem to just be reviewing things and talking about things with no real aim.
[00:54:24] <Dark_Shikari> Even something like "make motionest faster" is a goal, though I don't think a very useful one.
[00:54:40] <Dark_Shikari> For example it could be "modernize ffmpeg's assembly code", since a whole lot of the asm is mmx-only (for mpegvideo, etc).
[00:54:51] <michaelni> about resigning as BDFL, i just dont see how it could work, at the least i need to retain control over the accounts
[00:55:03] <michaelni> alternative is chaos IMHO
[00:55:12] <Dark_Shikari> what do you mean by retain control over accounts?
[00:55:16] <Dark_Shikari> you mean root?
[00:55:55] <michaelni> i mean when lets say ffmpeg is on soureforge someone is in charge to add/remove accountgs
[00:56:24] <michaelni> for the project that is
[00:56:29] <Dark_Shikari> that doesn't mean you need to be BDFL for that.
[00:56:35] <michaelni> you cant have noone
[00:56:36] <Dark_Shikari> For example, a man at a nuke silo has the ability to launch the nuke.
[00:56:42] <Dark_Shikari> That doesn't mean he has the RIGHT to unless he's ordered to.
[00:56:57] <Dark_Shikari> You can have the _ability_ to manage accounts, while still being required to obey the decisions of the majority.
[00:57:20] <michaelni> that i can live with
[00:57:21] <michaelni> but
[00:57:34] <michaelni> sometimes its simply bad to do long votes ...
[00:57:42] <lu_zero> michaelni: by votes
[00:57:44] <Dark_Shikari> I agree that votes are generally bad.
[00:57:45] <lu_zero> and by count
[00:57:49] <lu_zero> you are in minority
[00:57:51] <lu_zero> end.
[00:58:08] <Dark_Shikari> On the other hand, I think that most things don't need votes.
[00:58:11] <michaelni> lu_zero, what votes what count?
[00:58:13] <Dark_Shikari> "Should $newuser get commit access?"
[00:58:23] <lu_zero> even on your quite biased and ugly "love me hate me" call
[00:58:26] <Dark_Shikari> usually such things don't require much debate.
[00:58:28] <lu_zero> I'm sick of that
[00:58:49] <michaelni> lu_zero, what?
[00:58:52] <lu_zero> YOU
[00:59:08] <bcoudurier> hey please calm down
[00:59:36] <bcoudurier> I don't even understand why votes are so bad
[00:59:45] <Dark_Shikari> bcoudurier: they invite politicking
[00:59:48] <Dark_Shikari> I think
[01:02:17] <bcoudurier> I can see why, but the same can happen without
[01:02:29] <bcoudurier> anonymous votes allow people to express their opinion
[01:02:37] <Dark_Shikari> Anonymous votes are good
[01:02:40] <Dark_Shikari> most of the votes we've had -- haven't been.
[01:02:46] <bcoudurier> I agree with this
[01:02:54] <bcoudurier> and I raised that concern on -legal once
[01:02:56] <Dark_Shikari> non-anonymous voting invites massive politicking.
[01:03:00] <Dark_Shikari> I strongly oppose it.
[01:03:07] <iive> Dark_Shikari: you cant vote anonymously over internet.
[01:03:38] <Dark_Shikari> sure you can
[01:03:59] <lu_zero> in gentoo we do every time we need
[01:05:43] <iive> you need independent third parity whom you can trust to count the vote correctly and to keep the secret.
[01:06:31] <uau> iive: i think there actually are cryptographic protocols that enable a secret vote without a third party
[01:06:55] <iive> no, no protocols yet
[01:07:11] <iive> i've read about some math that may allow it, but it is untested field.
[01:07:59] <BBB> votes are bad, don't do votes
[01:08:03] <BBB> blegh
[01:08:52] <BBB> bcoudurier: we've abused votes in the past, as a system of making a decision when we can't get to one as a community
[01:08:55] <uau> iive: a google search returns hits that look like they're talking about such a protocol
[01:09:11] <BBB> bcoudurier: the result has generally been that half developers don't get their way, half do, the votes list two extremes and thus half of the devs are unhappy
[01:09:26] <BBB> bcoudurier: whereas in many cases we could've come to a consensus that met all developers more or less in the middle
[01:09:38] <BBB> bcoudurier: fun vote: foundation name. bad vote: michael stays leader or steps down
[01:09:51] <BBB> first: there's a few literal choices and really, nobody cares that much anyway
[01:09:53] <iive> BBB: votes are used when such consensus cannot be reached
[01:10:07] <BBB> second: there's many ways in between that would satisfy the whole developer body more than those extremes
[01:10:12] <BBB> so the vote was pointless :(
[01:10:13] <iive> as a method to stop arguing.
[01:10:34] <BBB> iive: keep believing that, but clearly as you've noticed in the past week, votes do not end the arguing
[01:10:39] <BBB> they just leave half the developer body unhappy
[01:10:41] <BBB> if not more
[01:10:59] <iive> BBB: i don't see any voting in the past week.
[01:11:15] <BBB> thank god no
[01:11:19] <BBB> that would've made it worse
[01:11:46] <BBB> here's a vote that makes no sense whatsoever: who shall be project leader: michael, and mans leaves ffmpeg for good, or mans, and michael leaves ffmpeg for good
[01:11:51] <BBB> that makes _no sense whatsoever_
[01:11:57] <Dark_Shikari> It's also a false dichotomy.
[01:11:58] <BBB> if you don't see why, then sorry, you're an idiot
[01:12:08] <BBB> there's better choices than either of those two
[01:12:10] <BBB> Dark_Shikari: yes :)
[01:13:21] <iive> BBB: mans and michael have reached stage where they can't work in the same project. what other options do you have?
[01:13:34] <Dark_Shikari> iive: in that case, I pick the third option.
[01:13:36] <Dark_Shikari> Both leave.
[01:13:39] <bcoudurier> sometimes difficult decisions have to be made and consensus cannot be reached
[01:13:39] <Dark_Shikari> Notjoking.jpg
[01:13:58] <bcoudurier> I've had votes against what I wanted in the past and I accepted the outcome
[01:14:23] <BBB> bcoudurier: if nothing else, a vote can help - but we've abused them beyond what they are good for
[01:14:26] <BBB> no more votes for now
[01:14:35] <Dark_Shikari> "If we have to pick between Mans and Michael to leave, I choose that both leave."
[01:14:37] <bcoudurier> if we want to avoid votes, we need reasonable people that agree to a compromise
[01:14:38] <BBB> and michaelni's email just now is an example of how not to do this
[01:14:50] <bcoudurier> many times I didn't see any compromise
[01:15:09] <bcoudurier> and harsh tone
[01:15:51] <michaelni> BBB, by email is poor but the truth should not be hidden
[01:16:03] <Dark_Shikari> michaelni: but it's not the truth, it's just flames
[01:16:11] <iive> Dark_Shikari: indeed, my mistake. there is forth option.
[01:16:20] <Dark_Shikari> fourth?
[01:16:23] <Dark_Shikari> which is that
[01:16:29] <iive> 00,01,10,11
[01:16:31] <Dark_Shikari> michael leaves, mans leaves, both leaves, or...
[01:16:37] <Dark_Shikari> Oh.  I thought you implied 00 wasn't.
[01:19:00] * Compn still not see what michael actually does as 'leader'
[01:19:33] <Compn> i see him rejecting all patches as a maintainer
[01:19:38] <michaelni> iam evil, eat small children, force people to commit seppuku so i can deduct the burial cost
[01:19:40] <Compn> but not him overriding any other maintainers
[01:19:56] <Compn> i havent read every mail...
[01:20:03] <iive> lol
[01:20:16] <Compn> you should eat the large children too, more meat
[01:20:40] * michaelni eats Compn 
[01:21:10] <Compn> but i'm full of transfats!
[01:21:19] * michaelni pukes
[01:21:34] <merbanan> this isn't going to lead anywhere
[01:21:34] <iive> Compn: now you are answering your own questions.
[01:21:51] <BBB> I'm with merbanan
[01:22:17] * Compn guessed if michael came on irc that all people talking shit about him would stop, and everyone would get quiet
[01:22:29] <Dark_Shikari> Talking shit begets talking shit.
[01:22:32] <Dark_Shikari> if nobody talks shit, nobody talks shit.
[01:22:35] <Dark_Shikari> if someone talks shit, everyone talks shit.
[01:22:53] <michaelni> solution to our problems: lets talk shit
[01:23:00] <Dark_Shikari> yaaaaaay talking shit.
[01:23:15] <Compn> michaelni : btw why were you upset about ronald asking carl if he wanted to get paid for fflegal ? because it looked like a bribe ?
[01:23:31] <michaelni> yes, i misunderstood
[01:23:33] <Compn> ah
[01:23:40] <lu_zero> ...
[01:23:46] <Compn> thats why i said not to jump :P
[01:23:51] <Compn> no jumping allowed!
[01:24:03] <lu_zero> something that was already half decided months ago...
[01:24:16] <michaelni> BBB and merbanan leaving isnt solving this either :(
[01:24:29] <Dark_Shikari> michaelni: time to talk shit!
[01:24:34] <michaelni> they should have stayed in the channel
[01:24:35] <Dark_Shikari> Your mother was a hamster and your father smelt of ELDERBERRIES!
[01:24:38] <lu_zero> and still on the ml remains just what you wrote in vitriol
[01:24:38] <Compn> lu_zero : what did michaelni do to make you hate him so much? if you dont mind airing your grievences now ?
[01:24:43] <lu_zero> michaelni: I'm about to sleep
[01:24:48] <bcoudurier> it looked like a bribe considering the situation
[01:24:51] <Compn>  oh too late :P
[01:24:52] <iive> they need to sleep.
[01:24:53] <Dark_Shikari> I empty my nose at you, you pathetic pig-sty cleaner!
[01:24:54] <lu_zero> merbz is on the same time zone
[01:25:02] <lu_zero> BBB has a newborn to care...
[01:25:16] <bcoudurier> lu_zero, "decided" ?
[01:25:16] <iive> there is russian saying that morning is smarter than evening.
[01:25:18] <bcoudurier> how so ?
[01:25:55] <Compn> michaelni : maybe you should apologize on the list for some of them rude mails...
[01:26:07] <lu_zero> bcoudurier: the proposal had been made and there were only positive comments that I recall
[01:26:09] <michaelni> yeah, which ? :)
[01:26:10] <Compn> well thats what i'd do anyways
[01:26:10] <bcoudurier> any foundation matter is decided is when 4 people from the directors vote for
[01:26:22] <michaelni> i mean for which do you think it matters most?
[01:26:40] <lu_zero> bcoudurier: that's why I said _half_
[01:26:50] <lu_zero> and not decided
[01:27:02] <Compn> michaelni : hmm, good question. i am not sure which ones were most upsetting.
[01:27:13] <iive> lu_zero: see, the problem here is that the proposition haven't been official.
[01:27:19] <bcoudurier> besides ronald has no real right to do that
[01:27:30] <michaelni> Anyway if people want we can keep the 2 repos i just dont think this is a good idea
[01:27:30] <iive> so if carl haven't agreed, it may have never been official.
[01:27:38] <Compn> ronald can make an inquiry before making a proposition to the board ?
[01:27:42] <bcoudurier> and as the treasurer it could even more look like a bribe :>
[01:27:42] <iive> been->become
[01:27:52] <Compn> bcoudurier , iive : cant he ?
[01:28:06] <bcoudurier> Im kidding here
[01:28:10] <Compn> its hard to tell.
[01:28:21] <bcoudurier> smiley :>
[01:28:34] <relaxed> that's a bird
[01:28:40] <Compn> is iive kidding ?
[01:28:50] <bcoudurier> anyway the timing was not the best, but the idea is good
[01:29:40] <ubitux> oh guys, you changed again the repository?
[01:29:41] <iive> that's how bribe is done. I do not imply that is what ronald intent.
[01:30:14] <michaelni> ubitux, just wait a minute until diego chnages it back flames me and closes my account
[01:30:23] <ubitux> haha
[01:30:34] <Compn> hopefully mru will come up with a better dload page for all ffmpeg trees
[01:30:34] <ubitux> come on… :p
[01:30:58] <ubitux> you're seriously all insane :)
[01:31:09] <michaelni> I really do NOT want to loose mru
[01:31:12] <Compn> now for some serious questions
[01:31:17] <Compn> michaelni : what are you drinking right now ?
[01:31:21] <iive> I see that ronald then talks about third unrelated thing. Also the mail date is 18'th, and that's the day of the switch, not signing.
[01:31:23] <michaelni> or anyone else
[01:31:32] <michaelni> Compn, water was last
[01:31:59] * Compn drinks sports drink, like gatorade
[01:32:20] * Kovensky goes drink some water
[01:32:25] <lu_zero> Compn: it was started by me and mru was filling the css and the entries since I was tired by that time already...
[01:32:26] <michaelni> iive, the "bribe" really isnt helping us in resolving this
[01:32:51] <Compn> lu_zero : no problem, just trying to help :(
[01:33:07] <michaelni> iive, so lets accept that BBB was fucking stressed by his baby and wrote shit without thinking on the wrong day
[01:33:27] <Compn> michaelni : so how do you propose everyone work together here ?
[01:33:34] <Compn> if you have any ideas...
[01:33:45] <michaelni> Depends on the people
[01:33:57] <michaelni> if people want 2 repos this is possible
[01:34:09] <michaelni> if they want to try 1 this is possible
[01:34:12] <iive> as i just said, the mail is from 18 january. the signing was on 17
[01:34:18] <michaelni> we can also switch it every week
[01:34:30] <Compn> iive : i think everyone is willing to drop it, is that ok with you ?
[01:34:40] <Compn> lol switch it every week
[01:34:50] <iive> yes, I've dropped it.
[01:35:00] <iive> oops, it broke.
[01:36:29] <Compn> michaelni : mru rejected the libmpcodecs stuff, should it stay on a branch or what should be done ?
[01:36:49] <michaelni> Compn, i dont care the faintest about libmpcodecs
[01:37:10] <michaelni> if we can resolve this by doing something with it please people, mru do it with libmpcodecs
[01:37:13] <Compn> i guess no one does :)\
[01:37:41] <Compn> well i cant figure out where you guys differ in code ideas then
[01:38:00] <michaelni> Compn, we dont strongly differ
[01:38:11] <Compn> and by 'you guys' i mean everyone, not just you and mans
[01:38:15] <michaelni> we differ in knowledge that is we have knowledge in different areas
[01:38:26] <michaelni> ahh i meant me and mru
[01:38:52] <michaelni> and that different areas knowledge leads to somewhat different oppinions i think
[01:38:52] <Compn> ehe
[01:39:20] <michaelni> and theres bikesheds ....
[01:39:38] <lu_zero> Compn: some libmpcodecs will be ported in a sprint if I can convince saste or michaelni to support the effort sharing their knowledge
[01:39:39] <michaelni> where things can be either way and down to earth it makes no difference at all
[01:40:30] <michaelni> lu_zero, you can convince me if this is moving us toward a solution but if its not then no
[01:40:33] <Compn> its quite hard to organize and agree and have a committee work on a project together ;\
[01:41:09] <michaelni> i mean i dont want to spend a month on libmp and then find we still are in the same crisis
[01:41:30] <ubitux> btw, this may sound stupid, but why don't you host the main copy of the repository on ffmpeg.org?
[01:41:33] <Compn> lu_zero : well i havent even seen any users use libavfilter yet :P
[01:41:44] <bcoudurier> I crashed :(
[01:41:44] <Dark_Shikari> michaelni: maybe then you should get approval before doing things unilaterally
[01:41:53] <lu_zero> Compn: yadif had been used with satisfaction
[01:41:55] <michaelni> ubitux, trust & paranoia
[01:41:56] <Dark_Shikari> You constantly whine at others for such things, but then you go do them yourself.
[01:42:08] * lu_zero suggested its use in some situations
[01:42:15] <ubitux> michaelni: since you're using git there is not that much to care about
[01:42:30] <ubitux> i mean you can push to somewhere else anytime
[01:42:33] <Compn> ubitux : let the dust settle before drawing the lines :P
[01:42:36] <lu_zero> and still
[01:42:44] <Compn> ubitux : also i dont think michael has commit rights on git.ffmpeg ...
[01:42:59] <michaelni> Dark_Shikari, i know i fucked up with that mail a bit
[01:43:08] <lu_zero> libmpcodecs is something total orthogonal
[01:43:21] <michaelni> I dont think the idea was bad but yes i should have asked and worded it more diplomatically
[01:43:27] <lu_zero> Compn: nobody but those in charge of merging do
[01:43:29] <Dark_Shikari> well, right there I was referring to libmpcodecs
[01:43:37] <Dark_Shikari> I think people have the impression you forced libmpcodecs on ffmpeg
[01:43:39] <ubitux> Compn: well i mean, admitting everything was fine (for example just before the putsch), why was it important to put the main repo elsewhere?
[01:43:40] <Dark_Shikari> even when most disagreed
[01:43:44] <Dark_Shikari> Yet -- you whine at others for doing the same.
[01:44:02] <michaelni> Dark_Shikari, let me explain about libmp
[01:44:11] <ubitux> i mean, at least for users it's disturbing :P
[01:44:13] <Compn> ubitux : its a long story
[01:44:23] <ubitux> Compn: do you have a thread reference?
[01:44:32] <Compn> see git threads in ffmpeg-devel
[01:44:40] <ubitux> 'k thanks.
[01:44:41] <Compn> go back a few months
[01:44:50] * Compn doesnt remember exact threads
[01:44:57] <Compn> but its all on the list
[01:45:00] <michaelni> attila gave me a hint of a possible fork a week or so earlier but no word on who,why or anything, and i thought if iam more active and work on something usefull like libmp i hcould avoid the fork
[01:45:22] <michaelni> thats why i was a bit pushy
[01:45:42] <Compn> Dark_Shikari : i think a few people asked during libavfilter development about porting libmpcodecs
[01:45:48] <Compn> as a wrapper
[01:46:00] <lu_zero> Compn: there was a discussion about it
[01:46:00] <Compn> i think it was even in the soc project description ?
[01:46:24] <lu_zero> started by Stefano I think
[01:46:34] <michaelni> as said if doing something with libmp like droping it resolves things i definityl agree
[01:46:44] <Dark_Shikari> it's not about dropping it, it's about having discussion
[01:47:05] <Compn> michaelni : Dark_Shikari and others want you to ask what to work on before you do it, i guess
[01:47:28] <Dark_Shikari> I think people should be aware of what's going on.
[01:47:29] <Dark_Shikari> Code dumps are bad.
[01:47:32] <Compn> heh
[01:47:38] <michaelni> lords, how can i server thee best?
[01:47:43] <iive> michaelni: i actually joked here that it would be better if you gave ultimatum to commit the wrapper if filters from mp are not ported for x weeks.
[01:48:04] <lu_zero> sigh...
[01:48:08] <Compn> thats what his mail said iive
[01:48:16] <Compn> wrapper would be in until it was ported and faster ...
[01:48:23] <Compn> or someone elses mail
[01:48:41] <iive> maybe somebody quoted me...
[01:48:48] <Compn> lu_zero : yes, stefano didnt want to do it, and michael wanted someone to do it
[01:48:56] <Compn> lu_zero : michael ended up doing it
[01:49:02] <lu_zero> the whole libmpcodecs stuff would be solved in around 3-4 days by collectively porting those 20 filter one by one
[01:49:05] <Compn> i'm not understanding the problem really
[01:49:13] <lu_zero> Compn: have a look at that code
[01:49:15] <lu_zero> and
[01:49:20] <michaelni> stefano tried but he lacked experience with integrating code dumps
[01:49:28] <Compn> ah
[01:49:36] <lu_zero> have fun trying to optimize swscale
[01:50:00] <bcoudurier> well I think having libmpcodecs features is nice
[01:50:07] <michaelni> BBB, hi
[01:50:15] <lu_zero> the features, indeed, the code as is not.
[01:50:19] <bcoudurier> a compromise would have been to not compile them by default
[01:50:28] <bcoudurier> a _compromise_
[01:50:35] <Compn> lu_zero : ah, you are talking about having optimized libavfilter being next to optimizing libmpcodecs ?
[01:50:40] <bcoudurier> saying I don't want this doesn't lead anywhere
[01:50:53] <uau> bcoudurier: i don't think that would have been accepted as a compromise
[01:50:54] <bcoudurier> and it will happen again in the future
[01:50:55] <ubitux> oh, repository url switch again.
[01:50:56] <Compn> lu_zero : i mean, the trouble having the same code so that both can be optimized ?
[01:51:18] <ubitux> (hey, this is starting to be ridiculous)
[01:51:27] <bcoudurier> uau, let the people talk, please
[01:51:34] <lu_zero> Compn: could we drop libmpcodecs?
[01:52:03] <michaelni> lu_zero, as i said if it helps solve the crisis definitly yes
[01:52:18] <lu_zero> having the libmpcodecs interesting filters as avfilter is something 5-6 people would get done in a weekend.
[01:52:26] <Compn> lu_zero : i have no problem with dropping whatever
[01:52:33] <lu_zero> I mean dropping the discussion
[01:52:35] <bcoudurier> lu_zero, let's try to do that then, who is in ?
[01:52:48] <Compn> lu_zero : yes, i agree its not going anywhere
[01:53:07] <Compn> sorry, i brought it up as an example but ... my question got answered quickly
[01:53:11] <lu_zero> I'm in, Benjamin was (shoud re-ask)
[01:53:18] <lu_zero> BBB: you were, you'd be?
[01:53:31] <bcoudurier> that sounds good
[01:53:36] <bcoudurier> I would be in also
[01:53:56] <lu_zero> stefano was debating since the whole idea was told him on the phone that day
[01:54:12] <Dark_Shikari> nice patch by mru.
[01:54:14] <lu_zero> anyway, that is not the problem.
[01:54:39] <bcoudurier> btw, I think cleaning up swscale is needed and that it must be done
[01:55:36] <BBB> lu_zero: sorry, in for what? short summary?
[01:56:22] <Kovensky> BBB: hack weekend to port the most important filters from libmpcodecs to libavfilter
[01:57:53] <lu_zero> bcoudurier: I'm at patch 8, few more needed ^^;
[01:58:04] <bcoudurier> please keep going :)
[01:58:25] <BBB> Kovensky: I can be in, but I want this bs to end first please
[01:58:36] <BBB> and yes, lu_zero, awesome :)
[02:01:16] * michaelni sees links reverted to illegitimate repo and returns to civil disobedience
[02:01:33] <Dark_Shikari> "illegitimate"?
[02:02:02] <michaelni> no public discussion, many details kept secret, no public vote, ...
[02:02:19] <Dark_Shikari> ... says michael, who committed libmpcodes with no public discussion, no public vote...
[02:02:24] <michaelni> people signing actually not knowing it was a takeover and no fork
[02:02:34] <Dark_Shikari> But it was a fork.  There's two repos.
[02:02:46] <Compn> Dark_Shikari : not according to download page ...
[02:03:07] <lu_zero> Compn: the download page has a patch for it now =P
[02:03:08] <michaelni> nor according to suggested new download page
[02:03:22] <michaelni> its still advertising mans tree as more official which it is not
[02:03:27] <Dark_Shikari> "Fork" does not mean "all forks are equal"
[02:03:30] <Dark_Shikari> it means "there are multiple trees"
[02:03:38] <bcoudurier> humm, michaelni please don't
[02:03:57] <Compn> and now we get to bikeshed the tree names in dload page like i guessed we would
[02:04:02] <michaelni> bcoudurier, ?
[02:04:05] <Compn> hooray
[02:04:27] <BBB> anyway, to get back to kovensky's question, yes I will help porting the most important mpfilters if that helps us along, _if_ we can at the same time also figure out a way out of this mess we're currently in
[02:04:34] <BBB> Compn: no, let's vote
[02:04:58] <BBB> I think my tree should be called "The Tree of Corporate Evil"
[02:04:59] <bcoudurier> don't go crazy and into the "civil disobedience" mode
[02:05:14] <michaelni> ahh ok understood
[02:05:42] <Compn> BBB : you should name it after your kid :P
[02:05:53] <uau> BBB: i don't think porting filters will help that much - at least i didn't get the impression that many people would have forked over the filters specifically
[02:06:14] <BBB> uau: that is my impression also, but if it helps I'll do my share
[02:06:26] <lu_zero> the filter are something small
[02:06:36] <michaelni> uau, /BBB mine too
[02:06:54] <BBB> I don't mind doing work to solve this mess, though
[02:07:00] <lu_zero> eh
[02:07:26] <michaelni> BBB i also volunteer but it must be real not waste time for no gain
[02:07:34] <Compn> so both sides should get a list of demands ?
[02:07:43] <uau> michaelni: btw as Dark_Shikari already said, i don't think many people would view it as 'måns's tree'
[02:07:58] <lu_zero> michaelni: having libavfilter used was your wish as well
[02:08:07] <lu_zero> still
[02:08:38] <michaelni> uau, its one of 2 ffmpeg trees, the "mans" one is not more official
[02:08:45] <BBB> Dark_Shikari: btw are you ok with me committing the emu_edge dsp'ification + C speed-up while I keep polishing the asm some more?
[02:08:51] <Dark_Shikari> michaelni: there is one thing I typically disagree with you on that I think is important here
[02:08:54] <Dark_Shikari> Users are the #1 priority
[02:08:56] <Dark_Shikari> Not developers
[02:08:57] <lu_zero> I'd like you wait and see if this way of managing patches for the master is that wrong or not
[02:09:04] <Dark_Shikari> therefore, if putting forked repos up on the ffmpeg download page would confuse people
[02:09:07] <Dark_Shikari> it is better not to do it
[02:09:12] <Dark_Shikari> (i.e. putting up two, separate repos that appear equal)
[02:09:16] <Dark_Shikari> (but aren't the same in content)
[02:09:28] <uau> michaelni: my point was about the name "mans tree" - i don't think other developers view it that way (identifying it with him personally)
[02:09:50] <Dark_Shikari> BBB: yes
[02:09:51] <lu_zero> uau: mans' tree is http://git.mansr.com/?p=ffmpeg
[02:09:52] <michaelni> uau yes i agree with you
[02:09:52] * BBB agrees with uau
[02:10:01] <iive> uau: it is the tree mans created after ffmpeg moved to videolan.
[02:10:03] <Dark_Shikari> asm does not have to be pretty, it has to be fast.
[02:10:09] <Dark_Shikari> "Ugly asm" is better than "no asm".
[02:10:11] <michaelni> i just had to refer to it somehow
[02:10:13] <Dark_Shikari> "pretty asm" is better than "ugly asm".
[02:10:19] <Dark_Shikari> But if there's no pretty asm, ugly is better than nothing.
[02:10:26] <BBB> Dark_Shikari: heh :) I hate to see my name be asosciated with this mess :-p
[02:10:35] <BBB> it's darn fast though
[02:10:52] <bcoudurier> I agree with Dark_Shikari with Users are #1 priority
[02:10:53] <lu_zero> BBB: why it's ugly then?
[02:11:08] <j-b> users are cheap and numerous... Developers aren't
[02:11:10] <bcoudurier> that's the reason I had my own tree in the first place
[02:11:17] <BBB> lu_zero: uh... just look at it, it's a little... convoluted
[02:11:49] <lu_zero> BBB: if you'd have to change it in a week you'd be able?
[02:12:00] <BBB> lu_zero: probably
[02:12:10] <lu_zero> if I'd have to change it you'd be able to tell me, in a month?
[02:12:10] <BBB> unless I have to port mpfilters :-p
[02:12:24] <BBB> two weeks
[02:12:27] <BBB> if not less
[02:12:35] <BBB> probably a week also, depends on how much time you have
[02:12:52] <lu_zero> if I'd have to change/fix it alone I would scream in pain?
[02:12:54] <uau> lu_zero, iive: i don't think michaelni meant that (certainly i don't see how the download page would claim that to be "more official" - the patch says "Personal repository with works in progress" about that one)
[02:13:01] <Compn> i think we can all agree... we need to make kshishkov work on vivo decoders :P
[02:13:08] <BBB> lu_zero: yes
[02:13:28] * lu_zero enumerates the levels of ugliness
[02:13:58] <lu_zero> BBB: if you can make so it would let me or anybody else understand it, it's fine ^^;
[02:14:00] <drv> what's so exciting about vivo? i've never seen one outside the mphq samples
[02:14:24] <BBB> lu_zero: that's the goal :)
[02:14:47] <Compn> drv : it used to be big. like quicktime - vivo - real
[02:15:02] <Compn> - asf
[02:15:48] <Compn> drv : its also just about the last big format i can think of with no support in ffmpeg
[02:16:05] <Dark_Shikari> michaelni: you see, here's the problem again.
[02:16:06] <Compn> btw someone asked about openexr on -users list :)
[02:16:13] <Dark_Shikari> You say "videolan is official tree"
[02:16:20] <Dark_Shikari> You aren't even asking for both to be listed equally
[02:16:24] <Dark_Shikari> you're asking for that to be the only one listed
[02:16:31] <Dark_Shikari> that's not a compromise, that's a dictator decree
[02:16:34] <Compn> Dark_Shikari : you arent asking for both to be listed equally
[02:16:41] <Compn> hurrrr
[02:16:43] <Dark_Shikari> Compn: I'm not, I'm saying Michael isn't proposing that.
[02:16:59] <iive> Dark_Shikari: because they are not equal.
[02:17:06] <Compn> michaelni : try proposing that, please oh please
[02:17:07] <Dark_Shikari> Yes, one has all the developers, one has two devs.
[02:17:34] <lu_zero> the git.ffmpeg.org is what ubuntu and gentoo will use for the time coming....
[02:17:39] <Compn> again with the counting
[02:18:04] <bcoudurier> lu_zero, well ffmpeg.org is not ours
[02:18:29] <lu_zero> while ffmpeg-mt is what chromium and vlc are using for the time coming if -mt isn't merged somewhere else
[02:18:40] <lu_zero> bcoudurier: define ours and define ffmpeg.org
[02:18:48] <Compn> astrange was working on -mt merging
[02:18:49] <j-b> lu_zero: not yet for VLC, but yes, we will
[02:18:52] <Compn> post git switch
[02:19:00] <michaelni> Dark_Shikari, one has 7 commiters from which 4 dont commit, the other has all ffmpeg devels as commiters
[02:19:26] <bcoudurier> lu_zero, fabrice owns ffmpeg.org and the trademark
[02:19:32] <michaelni> and most commits are old bikeshed cosmetics n that 7 devel tree
[02:19:49] <Dark_Shikari> michaelni: er, no
[02:19:53] <uau> lu_zero: there should pretty soon be an mplayer2 release with the default being ffmpeg-mt too (tarball with ffmpeg-mt included to compile the player against)
[02:19:53] <Dark_Shikari> that one has far more than 7 committers
[02:19:59] <Dark_Shikari> "can push" != "can commit"
[02:20:02] <Dark_Shikari> anyone in the entire world can commit
[02:20:32] <iive> Dark_Shikari: please... don't bring semantics
[02:20:34] <bcoudurier> that's the same, pusher == svn commiters
[02:20:35] <michaelni> Dark_Shikari, 7 pushers? sounds odd to be but ok
[02:21:10] <Dark_Shikari> iive: michael brought up the semantics
[02:21:15] <Dark_Shikari> he's complaining about "only a few people are committing"
[02:21:20] <lu_zero> uau: ok, I guess I'll notice from Nikoli when you are ready =)
[02:21:26] <Dark_Shikari> when really he means "dozens of people are having their patches pushed via a few people"
[02:22:35] <bcoudurier> not yet dozens
[02:22:38] <iive> Dark_Shikari: you are confusing x264 development model with ffmpeg one.
[02:22:39] <bcoudurier> but I get the point
[02:22:46] <Dark_Shikari> iive: ffmpeg is the same
[02:22:50] <Dark_Shikari> in fact
[02:22:52] <michaelni> Dark_Shikari, people dont post patches for one tree they can be applied to either
[02:22:56] <Dark_Shikari> we just shifted ffmpeg to x264's model ;)
[02:23:02] <michaelni> i just refuse to work currenzly
[02:23:44] <iive> Dark_Shikari: you can't shift palm to sibir.
[02:23:53] <Dark_Shikari> ?
[02:23:57] <iive> Siberia?
[02:24:11] <Compn> why are we arguing about committers and pushers
[02:24:17] <Compn> we all know how this works and how many there are
[02:24:21] <Compn> this is dumb
[02:24:59] <Compn> if we are going to argue, lets argue about something dumber
[02:25:08] <Dark_Shikari> ok, good idea
[02:25:09] <lu_zero> michaelni: ...
[02:25:12] <Dark_Shikari> x264 has Touhou, ffmpeg needs an equivalent.
[02:25:16] <Dark_Shikari> Let's argue about what ffmpeg should have.
[02:25:24] <lu_zero> harui
[02:25:26] <Compn> you guys watch entirely too much anime
[02:25:34] <Dark_Shikari> "Touhou isn't anime!!!!!1o1oo1no1neo1"
[02:25:36] <Compn> is redline out yet ?
[02:26:10] <lu_zero> haruhi is a book...
[02:26:19] <Dark_Shikari> Haruhi has an anime or three.
[02:26:32] <lu_zero> Dark_Shikari: it's a damn scrambled book
[02:26:32] <iive> Dark_Shikari: you may have missed it, but reimer had some valid complains. He didn't said it, but 3 of his patches are just hanging. not rejected, not pushed, no further reviewed.
[02:26:36] <lu_zero> then an anime
[02:26:39] <bcoudurier> I don't know about this "pushers" thing anyway
[02:26:39] <lu_zero> then a religion
[02:26:50] <Dark_Shikari> iive: which ones?  I can go comment on them
[02:26:58] <bcoudurier> why should specific people be more trusted than some others ?
[02:27:02] <iive> ogm 21 jan
[02:27:09] <lu_zero> iive: we got patchwork for that...
[02:27:28] <Compn> bcoudurier : some idea about those people know more about git and wont fuck git repo up by doing a wrong rebase or something
[02:27:29] <iive> i think they were missing for some reason.
[02:27:34] <lu_zero> bcoudurier: other way round
[02:27:40] <Dark_Shikari> I don't think I'm qualified to comment on ogg demuxer stuff.
[02:27:41] <Dark_Shikari> Who is>?
[02:27:53] <lu_zero> mru wrote one
[02:27:54] <Compn> Dark_Shikari : everyone hates ogg stuff
[02:27:59] <bcoudurier> other way round ?
[02:28:01] <lu_zero> I ported it to ffmpeg
[02:28:01] <iive> mans
[02:28:18] <bcoudurier> it seems to me that maintainers should have pushing rights
[02:28:23] <lu_zero> bcoudurier: the people merging stuff are just doing it
[02:28:31] <lu_zero> so is basically manual labor
[02:28:36] <lu_zero> (and I know I'm slacking)
[02:28:42] <bcoudurier> looks like a maintainer job
[02:28:49] <bcoudurier> + some patch monkeys
[02:29:06] <lu_zero> maintainer job -> reviewing
[02:29:09] <Compn> like svn was ?
[02:29:23] <Compn> just say you want svn rules :p
[02:29:26] <j-b> Well, sorry, but it is ridiculous that the maintainer of FFmpeg is not paid to do that full time
[02:29:27] <bcoudurier> so no maintainers can be pushers ?
[02:29:37] <Dark_Shikari> j-b: I agree
[02:29:55] <j-b> and I mean it
[02:29:56] <lu_zero> bcoudurier: the idea is to have a second person push (and test)
[02:30:01] <bcoudurier> this "pushers" list looks more and more like a way to "control" the project
[02:30:12] <iive> well, now that there is dialog, I think I can get some sleep
[02:30:12] <j-b> FFmpeg is one of the most important project, maybe after the kernel
[02:30:16] <Compn> j-b : no offense, but putting the money thing into the middle of this isnt helping, i dont think
[02:30:21] <j-b> and maybe before
[02:30:27] <Compn> because i dont remember michael ever asking for money
[02:30:36] <j-b> Compn: I don't mean money, I mean time
[02:30:45] <Compn> what
[02:31:03] <bcoudurier> j-b, I agree
[02:31:20] <j-b> the one coordinating the merges and so one should be fulltime 50hours a week on that
[02:31:23] <iive> n8 ppl
[02:31:27] <Compn> oh i see
[02:32:14] <lu_zero> j-b: the idea was to spend one day per week every pusher
[02:32:17] <lu_zero> more or less
[02:32:35] * Compn is exhausted from ffmpeg talks going nowheres
[02:33:18] <uau> j-b: the problem with that is that there's no accepted "the maintainer", _especially_ when it comes to stuff like managing merges
[02:34:00] <j-b> uau: sure, and sorry, but seeing the actual status, it cannot be michaelni, nor mru. It must be someone a bit more neutral
[02:34:01] <michaelni> lu_zero, i can spend 7 per week :) you need more pushers
[02:34:18] <Dark_Shikari> j-b: I agree
[02:34:19] <michaelni> btw, i like to see diego push and review
[02:34:30] <j-b> uau: and probably not the best coder, but the one beeing the best at understanding most parts
[02:34:34] <lu_zero> michaelni: he did already
[02:34:38] <Dark_Shikari> j-b: DEFINITELY not the best coder
[02:34:41] <Dark_Shikari> the time of the best coders is too valuable.
[02:35:09] <Compn> i dont think there is anyone like that of the group
[02:35:16] <j-b> uau: maybe kernel like submaintainer trees and pulling trees
[02:35:17] <lu_zero> Makefile and config bits got reviewed by him
[02:35:21] <Compn> the coders like that either have jobs already due to ffmpeg work
[02:35:37] <Compn> we should hire linus
[02:35:44] <j-b> :)
[02:36:01] <uau> i kind of wondered myself why diego was among the "pushers" as his technical skills don't seem up to par for that (if it's really supposed to be "knowledgeable people")
[02:36:08] <uau> but i guess that's orthogonal to the main discussion
[02:36:14] <j-b> and michaelni's knowledge is way too important to let michaelni go away for stupid bullshit political fight
[02:36:18] <bcoudurier> uau, because this list is weird
[02:36:28] <j-b> (that goes also for mru in the past)
[02:36:40] <bcoudurier> and seems more like "friends" than anything else
[02:36:41] <j-b> Compn: jobs can be changed :D
[02:36:53] <Dark_Shikari> bcoudurier: I'm not sure why I'm on the list.
[02:37:06] <Dark_Shikari> so far my assumption is "because I was put on the list" is the best answer.
[02:37:08] <lu_zero> bcoudurier: people volunteering that had already experience on that kind of tasks
[02:37:16] <lu_zero> diego does that as job now...
[02:37:23] <uau> j-b: i think there isn't enough development in most sub-areas to be worth separate "submaintainer trees"
[02:37:59] <j-b> Compn: I am sorry, but from an external point of view, maybe BBB or lu_zero seem more "calm downed" people
[02:38:07] <j-b> uau: are you sure?
[02:38:22] <bcoudurier> janne and diego seems like a good pair
[02:39:01] <j-b> uau: lavf, lavfi, lavio, libavc/core, libavc-x86, libavc-neon, libavutils/core
[02:39:45] <j-b> uau: I am not telling you guys what you should do, I am just giving ideas, please don't take it in any other way
[02:39:51] <bcoudurier> j-b, yes that would be reasonable
[02:40:46] <uau> j-b: i don't count myself as a real ffmpeg developer (re "you")
[02:40:53] <bcoudurier> http://pastebin.com/ByUCwL5g
[02:41:00] <bcoudurier> can someone fix the compilation quickly ?
[02:41:08] <j-b> uau: nor me :D
[02:41:20] <j-b> uau: it was a general "you"
[02:41:45] <lu_zero> bcoudurier: did you re-run configure?
[02:41:54] <bcoudurier> it's not from me
[02:42:03] <bcoudurier> <pasteeater> looks like d36beb3 is causing some issues with ubuntu users.
[02:42:03] <bcoudurier> <pasteeater> http://pastebin.com/ByUCwL5g
[02:42:03] <bcoudurier> <pasteeater> resolved by --disable-vaapi
[02:42:11] <bcoudurier> that's what you get for not being in #ffmpeg ;)
[02:42:20] <lu_zero> not reading
[02:44:20] <uau> Dark_Shikari: you being on the list at least seems mostly logical as you've done similar things with x264
[02:45:40] <lu_zero> uau: and since it's a pain he has enough with x264 alone ^^;
[02:47:57] <Compn> Dark_Shikari : michaels request to have both repos equal on the page is here > http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2011-January/103580.html
[02:48:20] <Compn> Dark_Shikari : unless you meant something else ?
[02:48:57] <bcoudurier> thanks lu_zero
[02:50:40] <BBB> hey skal :-p
[02:54:06] <skal> hey :)
[02:54:17] <skal> heard some noise, saw some light, came in. :)
[02:55:56] * j-b offers beer to skal 
[02:56:54] <skal> hey, ain't Friday yet
[02:56:59] <skal> but accepted
[02:57:35] <lu_zero> hi skal
[02:57:58] <bcoudurier> I have crown royal on my desk
[02:58:57] <skal> hey lu_zero
[02:59:14] <skal> got a "Corsendonk" here (not opened ;))
[03:02:02] * lu_zero should have a 12 and an enkir somewhere
[03:02:10] <Sean_McG> ....
[03:02:18] * lu_zero right now likes more the water next to him
[03:02:54] <Sean_McG> oh your good friend Al
[03:02:55] <Sean_McG> Cohol
[03:04:52] <saintdev> better than my friend Sal
[03:04:57] <saintdev> Monella
[03:05:20] <Sean_McG> he mafioso?
[03:11:43] <lu_zero> bcoudurier, j-b, somebody with vaapi or dxva could test and report back ^^?
[03:19:01] <skal> heading home, back later
[03:20:50] <Dark_Shikari> uau: I don't do enough ffmpeg stuff though
[03:27:23] <Dark_Shikari> some people here wrote 5 decoders, I didn't even write one
[03:30:48] <Compn> Dark_Shikari : you see that earlier message ?
[03:31:10] <Dark_Shikari> which one?
[03:31:59] <Compn> > Dark_Shikari : michaels request to have both repos equal on the page is here > http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2011-January/103580.html
[03:32:15] <Dark_Shikari> oh that.
[03:32:31] <Compn> well you asked , in an angry tone :P
[03:32:52] <Compn> or maybe frustrated tone
[03:32:53] <Compn> who knows
[03:59:15] <BBB> Dark_Shikari: can you write us a h265 decoder?
[04:01:31] <astrange> ffcelt?
[04:01:43] <Dark_Shikari> ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—ï½—
[04:02:27] <saintdev> ffceltenc just as bad as ffaacenc, but lower latency!
[04:02:52] <saintdev> although that would be kind of hard to do...
[04:02:56] <Dark_Shikari> it's impossible to write a bad celt encoder
[04:03:00] <Dark_Shikari> too much is hardcoded in the bitstream
[04:03:02] <Dark_Shikari> same as with ac3
[04:03:08] <saintdev> i just thought that after i typed it, lol
[04:03:15] <Dark_Shikari> I mean, it's impossible to write a bad one unless you actively sabotage it
[04:03:26] <saintdev> "wait, you would have to REALLY screw things up to manage that"
[04:03:40] <BBB> ffcelt is fun
[04:03:42] <BBB> can you do that?
[04:03:52] <BBB> maybe I should learn new audio tricks by doing that
[04:04:02] <BBB> I guess celt should be in int?
[04:04:08] <Dark_Shikari> it has a fixed and float mode
[04:04:28] <BBB> decoder or bitstream mode?
[04:04:32] <Dark_Shikari> both
[04:04:34] <Dark_Shikari> afaik
[04:04:40] <BBB> so bitstrea
[04:04:42] <Dark_Shikari> I mean, it doesn't affect the format of the bitstream
[04:04:50] <Dark_Shikari> but it's intended to be symmetrical
[04:04:54] <Dark_Shikari> for minimum mismatch error
[04:05:05] <BBB> weird
[04:05:10] <Dark_Shikari> afaik
[04:05:14] <Dark_Shikari> well, I mean it's only natural, right
[04:05:17] <Dark_Shikari> no bitstreams actually use float
[04:05:19] <Dark_Shikari> even, say, mp3
[04:05:35] <BBB> true
[04:05:48] <peloverde> what about that xiph unpack float madness?
[04:08:42] <peloverde> mru: I thought you used to object to removing add_bias?
[04:08:48] <mru> never
[04:09:00] <mru> I want to keep the scale
[04:09:12] <mru> since it's free in most decoders
[04:09:20] <mru> as part of transform or windowing
[04:10:06] <peloverde> is the comment in dsputl.h correct? It seems to imply after this patch they will both use the same scale
[04:10:54] <mru> that's a different thing
[04:11:10] <mru> some people wanted to make all decoders output float in -1..1 range
[04:11:26] <mru> and then convert it to s16 with some horrid c code after
[04:11:49] <mru> I want to keep the option of outputting +-32k range even if in float
[04:11:58] <mru> to make conversion to int easier
[04:12:23] <mru> does that make sense?
[04:13:17] <peloverde> yes
[04:13:45] <mru> and I've already painstakingly written neon code for pre-scaled float to int
[04:14:16] <Dark_Shikari> why is float->int faster if pre-scaled?
[04:14:24] <mru> no need to multiply
[04:14:41] <mru> out=in is faster than out=scale*in
[04:16:55] <mru> 4-cycle latency doesn't make it easier
[04:18:08] <Dark_Shikari> how fast is float -> int16_t (the actual conversion)?
[04:18:38] <mru> almost all neon float operations have 4 cycles latency
[04:19:05] <Dark_Shikari> is there a 4xfloat -> 4x int16_t?
[04:19:12] <Dark_Shikari> or do you have to do -> 4x int32_t and then convert to 16?
[04:19:22] <mru> look at the code
[04:19:35] * Dark_Shikari is lazy
[04:19:38] <mru> it can only issue 2x per cycle
[04:20:07] <mru> I don't quite remember the details, but I did some trickery with 16.16 fixed-point
[04:20:22] <mru> for some of them at least
[04:20:31] <mru> made the interleaving easier
[04:22:03] <mru> then I could use a 32-bit shift and insert instruction
[04:57:17] <Dark_Shikari> what the crap, am I blind or is something weird going on
[04:57:24] <Dark_Shikari> This code works:
[04:57:25] <Dark_Shikari>     shr   edx, cl
[04:57:25] <Dark_Shikari>     add    al, byte [table + rcx]
[04:57:25] <Dark_Shikari>     shr   edx, 1
[04:57:26] <Dark_Shikari>     jne  .loop
[04:57:32] <Dark_Shikari> This code fails the test:
[04:57:33] <Dark_Shikari>     shr   edx, 1
[04:57:33] <Dark_Shikari>     add    al, byte [table + rcx]
[04:57:33] <Dark_Shikari>     shr   edx, cl
[04:57:34] <Dark_Shikari>     jne  .loop
[04:57:46] <Dark_Shikari> ... how can that possibly.... wtf
[04:58:52] <Sean_McG> what's in edx before the first shift?
[04:59:08] <Dark_Shikari> shouldn't matter, should it?
[04:59:15] <Dark_Shikari> It's a bitmask up to 16 bits long.
[04:59:26] <Dark_Shikari> ecx is the result of a "bsf" call (trailing zeroes)
[05:01:03] <pengvado> the effect of shr on flags depends on how much you shift
[05:01:11] <Dark_Shikari> But ne/e only depends on the result.
[05:03:00] <pengvado> >>0 doesn't set flags, not even ZF
[05:03:10] <Dark_Shikari> ahhhhhh
[05:03:22] <Dark_Shikari> that's tricky
[06:12:43] * elenril yawns
[06:12:48] <elenril> wow, :drama;
[06:14:41] <kshishkov> yawn indeed
[06:14:45] <ohsix> o rly
[06:18:33] <thresh> moroning
[06:20:02] <kshishkov> good morning (and a wonderful near-Moscow morning to you)
[06:22:02] <thresh> I'm trying to fight that bad feeling of mornings
[06:22:20] <thresh> Installed that Sleep Analyzer application, it's supposed to wake me up softly
[06:22:28] <kshishkov> did you have that many bad mornings in Andorra?
[06:22:54] <thresh> not really
[06:23:08] <thresh> mountain view somehow fixes that
[06:24:04] <Sean_McG> mornings suck
[06:25:42] <kshishkov> thresh: and probably they didn't end with work :)
[06:26:33] <thresh> kshishkov: very much so
[06:30:41] <elenril> michaelni: you should try http://tvtropes.org/pmwiki/pmwiki.php/Main/EvilLaugh
[06:38:06] <Tjoppen> morrn
[06:38:19] <pJok> god morgon
[06:48:54] * elenril recompiles the whole tree again
[06:48:57] <_av500_> moin
[06:48:58] * elenril glares at Flameeyes 
[06:49:15] <_av500_> Flameeyes: flameeye elenril!
[06:49:48] * _av500_ offers elenril some cheese: http://www.flickr.com/photos/av500/5388262880/
[06:50:09] <thresh> _av500_: well elenril didnt elenril Flameeyes, so why would Flameeyes flameeye elenril?
[06:50:32] * elenril retaliates with a http://tvtropes.org/pmwiki/pmwiki.php/Main/LargeHam
[06:51:57] <_av500_> thresh: because he can?
[06:57:19] <Sean_McG> how very meta
[07:02:40] <siretart>  /build/buildd/ffmpeg-0.7~daily~27417+201101262353~natty1/libavcodec/allcodecs.c:57: undefined reference to `ff_h263_vaapi_hwaccel'
[07:02:51] <siretart> known, transient problem? didn't happen yesterday
[07:05:16] <astrange> patch is in[FFmpeg-devel] [PATCH] Add ff_ to AVHWAccel decoders
[07:05:27] <siretart> ah, thanks
[08:02:36] <cartman> moin
[08:11:17] <superdump> morning
[08:11:38] <spaam> godmorgon
[08:11:43] <elenril> o/
[08:13:30] <cartman> Android 3.0 Preview \o/
[08:19:34] <wbs> cartman: what, where?
[08:19:42] <wbs> ugh, took quite a while to read the backlog from last night ;P
[08:20:11] <wbs> and \o/ for android NDK r5b - my __asm__ patch is one of the "highlights" of the bugfix release :-)
[08:20:45] <cartman> wbs: yup, thanks for that! :)
[08:21:45] <KotH> moin
[08:22:53] <cartman> elow KotH
[08:28:40] <KotH> hoi michaelni
[08:28:50] <KotH> michaelni: was hat dich denn hier her verschlagen? :)
[08:29:24] <Dark_Shikari> sollen wir deutsch sprechen?
[08:29:39] <KotH> nihongo mo ii yo
[08:29:50] <Dark_Shikari> weil michaelni an irc ist, wir sollen immer deutsch sprechen.
[08:29:56] <kshishkov> KotH: nej, man måste pratar på svenska här
[08:30:32] <pJok> KotH, wakarimasen
[08:30:32] <Dark_Shikari> KotH: 日本語
[08:30:34] <kshishkov> Dark_Shikari: バカバカしいやねん
[08:30:35] <Tjoppen> japp, svenska är det official ffspråket
[08:30:46] <Tjoppen> *officiella
[08:30:52] <Dark_Shikari> kshishkov: ________shii__ne__  (only know 30 kana)
[08:30:56] <kshishkov> officiellt?
[08:31:19] <kshishkov> Dark_Shikari: skäms, du kan lära mer
[08:34:22] <kshishkov> Dark_Shikari: och jag har inget tid att använda tyska
[08:34:31] <Dark_Shikari> ich kann das Sprache nicht sprechen
[08:34:38] <Dark_Shikari> oder, nicht lesen
[08:35:11] <siretart> wbs: I guess he's referring to http://android-developers.blogspot.com/2011/01/android-30-platform-preview-and-updated.html
[08:35:12] <pJok> hehe
[08:35:21] <pJok> Dark_Shikari, die Sprache
[08:35:45] <wbs> siretart: ah, thanks
[08:35:52] * pJok needs to kick himself some more and study more japanese
[08:35:56] <siretart> scheiss auf grammar :-)
[08:36:11] <Dark_Shikari> Ich werde die Articlen fast immer vergessen.
[08:36:22] <kshishkov> siretart: und Deutsche Wörter sind zu kurze
[08:36:43] <pJok> kshishkov, biste schwäbish?
[08:36:52] <kshishkov> pJok: absolut inte
[08:37:03] <colde> haha
[08:37:14] <pJok> morn colde
[08:37:20] <colde> morning pJok :)
[08:37:25] <kshishkov> pJok: jag vill gärna gå till Norrland att leva
[08:37:48] <colde> pJok: hows your swedish? :)
[08:38:08] <superdump> att leka
[08:38:53] <royger> does ffmpeg know the number of frames an input video has?
[08:38:55] <kshishkov> superdump: du kan leka men jag vill leva i Norden
[08:39:04] <andoma> kshishkov: "gå" in swedish is more like "walk" rather than "go"
[08:39:04] <pJok> colde, det är faktiskt bra :)
[08:39:06] <Tjoppen> kshishkov: gå till norrland från tyskland? lite väl långt :)
[08:39:11] <colde> Norden er også et fint sted at være :)
[08:39:14] <andoma> walking to norrland is gonna take a while ..
[08:39:16] <superdump> åker
[08:39:34] <av500> royger: depends
[08:39:45] <superdump> andoma: åker would be a better word, right?
[08:39:51] <kshishkov> Tjoppen: jag kan gå genom Danemark eller ta tåget
[08:40:08] <superdump> åka
[08:40:11] <superdump> :)
[08:40:12] <andoma> yes
[08:40:34] <kshishkov> andoma: well, it took me two pendeltåg at most
[08:40:39] <colde> kshishkov: you would need a boat/ship as well though :)
[08:40:57] <kshishkov> colde: ever heard of Öresundsbron?
[08:40:59] <colde> well, actually not, since you could go through jutland
[08:41:01] <colde> yeah
[08:41:03] <royger> av500: the input video is mpeg2, could you please tell me where can I find it?
[08:41:08] <colde> i just kinda remembered :)
[08:41:29] <kshishkov> colde: do they teach anything at Danish schools?
[08:41:38] <colde> Nope
[08:41:51] <colde> This is why we come here afterwards. To learn stuff :)
[08:42:32] <kshishkov> at least my Ukrainian school gave good training as a sweeper/street cleaner
[08:42:35] <spaam> andoma: but far and fara  are better then go :)
[08:43:37] <colde> kshishkov: haha :D
[08:44:29] <kshishkov> colde: I'm serious
[08:45:17] <colde> Really? Thats slightly scary then
[08:46:01] <av500> royger: in what container?
[08:47:20] <pJok> kshishkov, at least you speak english and you're getting somewhere with swedish
[08:48:05] <royger> av500: AVCodecContext or MpegEncContext would be great
[08:48:28] <kshishkov> colde: we had to clean all school territory several times a year for example.
[08:48:52] <kshishkov> pJok: they mostly teach to hate foreign languages but I failed :(
[08:48:59] <av500> royger: no, your mpeg2, in what container is it?
[08:49:15] <pJok> kshishkov, that just means you're a progessive one
[08:49:32] <colde> kshishkov: crazy
[08:49:37] <royger> av500: m2v or mpg usually
[08:50:22] <av500> these must be parsed completely to get the number of video frames
[08:50:44] <royger> av500: no luck then, thanks anyway
[08:59:46] <thresh> siretart: which git do you use for 0.7~ debian packages?
[09:00:53] <siretart> thresh: I use lp:ffmpeg, which is a bzr branch that imports from git.ffmpeg.org
[09:02:07] <thresh> so you chose the jedi side?
[09:02:09] <thresh> :)
[09:02:57] <av500> wasnt it the dark side?
[09:02:59] <superdump> andoma, kshishkov: would 'bo' be a better word for 'live' instead of 'leva' in : jag vill leva i Norden?
[09:03:21] <CIA-38> ffmpeg: Luca Barbato <lu_zero at gentoo.org> master * rd1b6f33bf2 ffmpeg/libavcodec/ (7 files):
[09:03:21] <CIA-38> ffmpeg: Add ff_ to AVHWAccel decoders
[09:03:21] <CIA-38> ffmpeg: That unbreaks compilation of vaapi and dxva2
[09:03:21] <CIA-38> ffmpeg: Signed-off-by: Janne Grunau <janne-ffmpeg at jannau.net>
[09:04:22] <thresh> av500: dark side never had so many followers
[09:04:51] <kshishkov> superdump: http://sv.wikipedia.org/wiki/Du_gamla,_du_fria#Texten
[09:05:21] <peloverde> g.ff.o is probably more amenable to releases
[09:05:40] <kshishkov> thresh: hmm, are you implying there's Darth Michaelius?
[09:05:58] <kshishkov> peloverde: only because it has people caring about releasing
[09:06:10] <superdump> kshishkov: ah, thanks :)
[09:06:23] <peloverde> let's not try to be too hostile with this jedi/sith nonsense
[09:06:41] <superdump> though i expect leva is used there for poetic purposes :)
[09:06:48] * av500 wants the ewok friendly ffmpeg
[09:06:56] <kshishkov> peloverde: it's just in my habit to ridicule everything
[09:07:34] <thresh> kshishkov: haha
[09:08:02] <superdump> it's very true, kshishkov does ridicule everything
[09:08:04] <superdump> :)
[09:08:05] <peloverde> many of these somehow get twisted into serious discussion on the ml
[09:08:12] <kshishkov> superdump: leva is more like general "to live" and bo is "to live somewhere" but I don't think you have such division in English
[09:08:31] <thresh> yeah sorry.
[09:08:41] <superdump> kshishkov: no, it would have be expressed through the way one stresses the word in the sentence
[09:09:02] <superdump> like: i want to _live_ in the north
[09:10:59] <kshishkov> anyway it's completely true
[09:11:20] <superdump> :)
[09:11:31] <superdump> at least we both understand the nuance
[09:12:44] * kshishkov also tried to understand nuances of Julmust tastes, found that julmust from Norrland is the best
[09:14:14] <superdump> any particular brand?
[09:14:38] <superdump> kshishkov: i highly recommend hjortron glögg
[09:15:34] <kshishkov> superdump: somehow I dislike glögg part there
[09:15:45] <kshishkov> Zeunerts and Vasa of course!
[09:16:06] <superdump> you can get non-alcoholic glögg
[09:16:08] * kshishkov remebers he still has a jar of hjortronsylt
[09:24:36] <wbs> elenril: the asfdec patch you posted, could you test it to make sure it doesn't break rtsp-ms/rtpdec_asf decoding? (not sure how much it touches that area, it might possibly)
[09:25:00] <wbs> elenril: you can try one of the ms-rtsp sample urls at http://wiki.multimedia.cx/index.php?title=RTSP#Samples_in_the_wild
[09:40:34] <elenril> wbs: i don't see how it could possibly break anything other than metadata
[09:40:44] <elenril> but tested with two samples, they work fine
[09:40:56] <wbs> ok, good
[09:40:56] <elenril> s/fine/as well as without the patch/
[09:41:01] <wbs> :-)
[09:41:47] <j-b> 'morning
[09:42:02] <wbs> I didn't check what it affected, but it sounded like it possibly could... since asf data is fed a little at a time via a special byteiocontext from the rtpdec layer to the asf demuxer.. and changes in how much is read per packet could screw something up in those cases
[09:50:41] <merbzt> elenril: I think I have an old asf metadata patch some where
[09:59:20] <merbzt> elenril: patch sent
[10:52:33] <j-b> av500: http://magsoft.dinauz.org/videolan/vlc-android/vlc-neon.apk
[10:54:37] <cartman> j-b: supports 2.2?
[10:54:46] <j-b> yes, sir
[10:54:54] <cartman> j-b: you guys are awesome :D
[10:55:23] <j-b> we are not
[10:55:23] <cartman> http://img408.imageshack.us/img408/6885/screenshot20110127at125.png hehe
[10:55:34] <thresh> j-b: does it work?
[10:55:47] <cartman> j-b: link not working though :S
[10:55:57] <j-b> works for me
[10:55:57] <thresh> worksforme
[10:56:05] * cartman blames .tr
[10:56:06] <thresh> well at least downloads, I have no device to test
[10:56:31] <av500> j-b: cool GUI!!!
[10:56:37] * cartman tries from USA
[10:56:39] <j-b> av500: pff :)
[10:56:57] <cartman> av500: screenshot!
[10:57:15] <av500> sec
[10:57:41] <av500> cartman: http://blog.justpulse.com/wp-content/uploads/2009/04/cone-soppera10.png
[10:57:44] <mru> j-b: do you think courmisch will allow vlc on android?
[10:57:47] <av500> it more or less looks like that
[10:57:57] <av500> mru: yes, androuis is "open"
[10:57:59] <cartman> av500: looks awesome
[10:58:02] <j-b> mru: well, well, yes
[10:58:08] <j-b> mru: on the Market, I don't know yet
[10:58:18] <thresh> Failure [INSTALL_FAILED_INVALID_APK]
[10:58:21] <thresh> or htc hero
[10:58:43] <cartman> thresh: truncated zip
[10:58:58] <thresh> hum?
[10:59:02] <thresh> unzip worked fine here
[10:59:24] <thresh> ah
[10:59:27] <thresh> lib/armeabi-v7a/libvlcjni.so
[10:59:31] <cartman> v7
[10:59:32] <cartman> :P
[10:59:33] <thresh> that probably explains as htc hero is v6
[10:59:39] <av500> j-b: it has to be a .mp4?
[11:00:07] <roxfan> "neon" in the name was a hint of sorts
[11:00:24] <thresh> :p
[11:00:41] <j-b> av500: it is just a test for the vout, so yes, we only test with the Mr_MrsSmith sample
[11:00:46] <j-b> av500: but yo can try anyhting
[11:00:57] <av500> url to your sample?
[11:01:05] <av500> i lost it
[11:01:10] <j-b> http://streams.videolan.org/streams-videolan/mp4/Mr_MrsSmith-h264_aac.mp4
[11:02:15] <av500> plays
[11:02:22] <kurosu> thresh: http://magsoft.dinauz.org/videolan/vlc-android/vlc.apk I guess
[11:02:45] <av500> now some 720p...
[11:02:49] <cartman> av500: :D
[11:03:03] <j-b> 720p, I doubt
[11:03:18] <kurosu> tested on my samsung phone, no sound output, no video scaling (broken gl on it though), crash when changing mode and trying to replay test.mp4
[11:03:31] <cartman> pffft phone
[11:03:38] <cartman> will test on Samsung Tab :P
[11:03:43] <j-b> kurosu: expected yes
[11:03:55] <kurosu> j-b: guessed so
[11:04:14] <av500> j-b: hardcoding /sdcard is wrong btw, you have to use Environment.something()
[11:04:22] <av500> getExternalStorage() or something like that
[11:04:26] <kurosu> samsung tab or s are beast regarding hardware
[11:04:41] <cartman> kurosu: its scary fast
[11:04:49] <cartman> 2.x UI sucks for tablet though
[11:04:49] <av500> j-b: http://developer.android.com/reference/android/os/Environment.html#getExternalStorageDirectory()
[11:04:53] <j-b> kurosu: sound is broken and the biggest part those days is to check that the video is actually outputed
[11:04:56] <thresh> kurosu: that's a normal behaviour for desktop version, why would you expect vlc on android to work better? :)
[11:05:41] <kurosu> j-b: all the more since android and many real devices have absurdly large delays
[11:05:56] <av500> j-b: I like your .mp4 debug output :)
[11:06:43] <av500> j-b: it crashes with 720p
[11:06:50] <j-b> kurosu: I mean, it is work in progress :)
[11:06:53] <j-b> av500: sample?
[11:07:04] <av500> some random 720p anime
[11:07:12] <cartman> j-b: did you guys found a fast way to do yuv->rgb conversion?
[11:07:25] <mru> cartman: yes, using hw
[11:07:27] <mru> :)
[11:07:35] <j-b> cartman: I don't think we do that yet
[11:07:37] <cartman> mru: bah :)
[11:07:39] <cartman> j-b: ah ok
[11:08:11] <cartman> mru: I think TI chips do have a hw function to do it?
[11:08:13] <cartman> read somewhere
[11:08:15] <av500> j-b: looks like oom error
[11:08:28] <mru> cartman: _all_ chips have it
[11:08:28] <kurosu> j-b: and I meant that even with a working implementation, audio/video synch might be a tough one
[11:08:41] <cartman> mru: all what chips? :)
[11:08:52] <j-b> cartman: well, before our vout was 2.3 only
[11:08:55] <av500> j-b: http://ffmpeg.pastebin.com/vFDaxRvX
[11:08:59] <mru> cartman: all chips capable of displaying an image
[11:09:01] <j-b> cartman: so now, we are doing 2.2 too
[11:09:03] <cartman> mru: oh
[11:09:13] <cartman> j-b: yup I know, I think 2.3+ will be more fun :)
[11:09:21] <mru> of course a 74LS74 doesn't have one
[11:09:34] <av500> maybe the 74LS565?
[11:09:35] <cartman> mru: but its not exposed for most chips?
[11:09:48] <av500> or the 74LS420
[11:09:54] <mru> android doesn't expose it
[11:10:00] <cartman> bandroid
[11:10:02] <cartman> :P
[11:10:08] <j-b> cartman: GLSL maybe can help
[11:10:12] <mru> lol
[11:10:19] <mru> xbmc guys tried that
[11:10:23] <mru> managed a few fps
[11:10:23] <cartman> j-b: there are some examples for that yes
[11:10:33] <mru> it eats memory bandwidth like crazy
[11:10:39] <j-b> cartman: well, at least, we are not regressing :)
[11:10:42] <cartman> j-b: or we can force mru to donate his neonized magic for that
[11:11:06] <j-b> I have a level+10 in god invocation, if you want
[11:11:08] <mru> hw is still faster
[11:11:14] <cartman> j-b: lol
[11:11:23] * mru is an atheist, won't work
[11:11:32] * cartman is agnostic, who cares
[11:11:50] <kshishkov> mru: what about commandments like "thou shall not memcpy()"?
[11:12:17] <cartman> j-b: anyhow I'll let you know when I test on Galaxy Tab
[11:12:22] <cartman> Boss took it away again
[11:12:32] <av500> clever boss
[11:12:49] <cartman> how clever for not letting me work
[11:12:53] <cartman> and still get paid
[11:12:55] <cartman> looks stupid
[11:13:01] <j-b> av500: I'll try the non-neon version on a 720p then
[11:13:40] <mru> kshishkov: http://26-26-54.hardwarebug.org/0
[11:13:57] <cartman> nice
[11:14:21] <superdump> can't some video output be written to use the android apis? (surface flinger iirc)
[11:14:34] <mru> those are rgb only
[11:14:35] <mru> iiuc
[11:14:36] <cartman> superdump: that needs RGB data
[11:14:44] <superdump> one would have to hook into the hardware for colorspace conversion
[11:14:53] <cartman> Android geniuses forgot about YUV
[11:15:00] <superdump> yeah
[11:15:05] <mru> just mimicking apple
[11:15:23] <av500> j-b: non neon does not crahs
[11:15:26] <cartman> hmm iOS doesn't do it in hw either?
[11:15:32] <superdump> but is there some other way to output YUV without converting?
[11:15:40] <mru> cartman: probably does with their own sw
[11:15:48] <mru> but it's not a public interface
[11:15:50] <superdump> i guess it will be converted somewhere along the way anyway to be composited
[11:15:55] <cartman> mru: ah, expected
[11:16:00] <mru> there is a private ios api that does it
[11:16:01] <superdump> assuming you aren't using fullscreen
[11:16:06] <mru> someone figured out how to do it
[11:16:23] <mru> superdump: hw alpha blending
[11:16:40] <superdump> mru: can accept multiple colorspaces?
[11:16:55] <superdump> why am i writing color and not colour? damnit
[11:17:05] <mru> of course it can
[11:17:11] <mru> the blending happens after conversion
[11:17:23] <superdump> ah, ok
[11:17:27] <lu_zero> mru: do you have more info on it?
[11:17:38] <mru> pick you omap trm
[11:17:40] <mru> your
[11:17:58] <mru> lu_zero: or info on what? ios?
[11:18:00] <superdump> solutions have to be worked out per chip though
[11:18:07] <cartman> ask av500 , his media player is super fast :D
[11:18:21] <lu_zero> yes
[11:19:08] <superdump> as there's no android api to accept buffers in random colourspaces and do the hardware alpha blending
[11:19:11] <superdump> afaik
[11:19:32] <superdump> but then, converting and using the android surface flinger api would have to be done per chip too
[11:19:35] <superdump> pain
[11:19:42] <superdump> omap ftw
[11:20:06] <mru> why would the api have to be per chip?
[11:20:16] <superdump> no the api wouldn't
[11:20:26] <cartman> the conversion is simple math
[11:20:30] <mru> of course drivers are per-chip
[11:20:36] <superdump> but whatever interacts with the chip's drivers to do one thing or another would have to be per chip
[11:20:37] <superdump> right
[11:20:45] <mru> the horror
[11:20:49] <mru> have to do _work_
[11:21:17] <superdump> although at the european celf, it seems v4l work is being done to expose such hardware and capabilities in some kind of graph-able way
[11:21:34] <superdump> and some of that work is going into linux 2.6.38
[11:22:19] <mru> v4l2 already exposes the necessary bits in a hw-independent way
[11:22:31] <mru> (nevermind no two drivers actually behaving alike)
[11:22:46] <kshishkov> and there was GATOS
[11:23:02] <superdump> hmm
[11:23:20] <superdump> when i say v4l, i mean v4l2 (who cares about v4l1 anymore?)
[11:24:04] <superdump> anyway, something like that is the way forward i guess as long as it's done well and the drivers behave (...yeah, i'm crossing my fingers)
[11:24:19] <superdump> or screw everything and use omap
[11:24:20] * pross-au questions the need for so many different hardware drivers. guessing that the driver development profession is unionised.
[11:24:52] <kshishkov> pross-au: because there are so many HW companies
[11:25:08] <superdump> and each hardware does stuff its own way
[11:25:14] <kshishkov> and driver development seems to be in-house
[11:25:19] <superdump> yup
[11:27:45] <mru> kshishkov: nokia wrote the display drivers for omap
[11:28:05] <j-b> mru: http://www.google.com/mobile/android/market-tos.html seems way more open, seeing §3.8
[11:28:15] <superdump> they're still closed because of powervr though, right?
[11:28:33] <mru> superdump: the display drivers are open as can be
[11:28:40] <mru> sgx is mem to mem only
[11:28:40] <superdump> oh
[11:29:14] <superdump> why do i remember people bitching about sgx drivers then?
[11:29:23] <mru> because those are closed
[11:29:37] <mru> but you don't need to touch sgx to display video
[11:29:46] <superdump> oh really?
[11:29:49] <superdump> interesting
[11:30:03] <mru> sgx renders into whatever buffer you tell it
[11:30:12] <mru> you don't need to use it at all
[11:30:20] <superdump> so you can do the same using a general purpose cpu
[11:30:25] <mru> of course
[11:30:26] <superdump> fun
[11:30:31] <mru> it's faster for 3d of course
[11:30:37] <superdump> i saw that the sony NGP has a quad core cortex-A9
[11:30:51] <mru> nice
[11:30:55] <superdump> yeah
[11:30:56] <superdump> :)
[12:06:42] <mru> lyakh: ping
[12:14:35] <cartman> wbs: I am not indian anymore, reported another useful Android bug :P
[12:14:55] <wbs> cartman: good boy *tap on head* ;P
[12:15:07] <cartman> I am probably older than you :P
[12:15:16] <wbs> most certainly, yes :-)
[12:15:38] <cartman> now the toolchain bug that mru's stuff exposes
[12:15:42] <cartman> that'll be harder to report
[12:15:44] <merbzt> I still think you appreciate a good tap
[12:15:48] <cartman> merbzt: yeah
[12:16:23] <mru> age is irrelevant on irc
[12:17:09] <mru> cartman: what bug is that?
[12:17:11] <cartman> http://code.google.com/p/android/issues/detail?id=14348
[12:17:24] <cartman> mru: static bug thing
[12:17:41] <mru> the attribute(((((used)))))?
[12:17:46] <cartman> mru: yup
[12:18:03] <cartman> that looked like LISP btw.
[12:19:00] <mru> no, that would be (attribute '(used . t))
[12:19:29] <kshishkov> cartman: they've been forcing parentheses uage for a reason since ages
[12:19:45] <kshishkov> mru: you can use a macro
[12:19:48] * cartman wonders the reason
[12:20:13] <kshishkov> it was voices somewhere - some GCC dev is LISPer and tried to make it homey
[12:20:39] <cartman> must be Tom Tromey, the emacs man
[12:21:13] <kshishkov> BTW, when Stallman was young (but had those "freedom" ideas already) he was heavy into list
[12:21:16] <kshishkov> *lisp
[12:21:55] * kshishkov wonders if Stallman can also be called "the emacs man"
[12:24:44] <lu_zero> mru: which time you'll arrive and depart from bruxelles?
[12:25:31] <mru> arrive fri 16:55
[12:25:40] <mru> depart mon 13:30
[12:26:06] <thresh> anyone on thursday?
[12:26:54] <kshishkov> mru: are you taking train?
[12:27:00] <mru> plane
[12:27:11] <mru> train from airport to city of course
[12:27:35] <kshishkov> then why do you mention such precise times?
[12:27:54] <kshishkov> planes overall are even less punctual than German trains
[12:28:09] <mru> just quoting the flight booking
[12:28:49] <kshishkov> do you believe it?
[12:29:05] <mru> it's realistic
[12:29:35] <mru> if there's no bad weather they should make it
[12:29:49] <mru> SOU isn't very crowded
[12:30:05] <kshishkov> what about Belgium?
[12:30:24] <mru> not so bad either
[12:30:48] <mru> nothing like LHR or FRA
[12:30:53] <kshishkov> nobody is interested in capital of Europe?
[12:31:09] * kshishkov still likes ARN most
[12:31:17] <mru> AMS is nice
[12:31:48] <mru> ARN is nice when departing, not arriving
[12:31:57] <mru> it's much bigger than it needs to be
[12:32:09] <mru> so security is quick
[12:32:18] <mru> but it takes them forever to get the luggage in from the plane
[12:32:40] <kshishkov> compared to FRA?
[12:33:10] <mru> I don't think I've ever had checked luggage there
[12:33:16] <mru> TXL is quick
[12:33:45] * thresh is having fun reading booking.com traveller comments on st nicholas hotel 
[12:33:48] <colde> CPH is nice both when departing and arriving :)
[12:34:19] <kshishkov> colde: Kastrup? I doubt it
[12:34:30] <colde> Yeah, it is :)
[12:34:34] <kshishkov> it's FRA wannabe
[12:35:02] <kshishkov> and it's also the only place where they've managed to lost my luggage during transfer
[12:35:09] <mru> FRA has the worst signage ever
[12:35:44] <colde> well, i understand why you might not like them, although i doubt they lose more luggage than any other airport :)
[12:35:48] <kshishkov> FRA also has baggage handling in different place than gates
[12:36:07] <kshishkov> mru: I've heard that Italian airports compete with it
[12:38:21] <superdump> NYO is pretty quick but they need more space for passport checking
[12:38:26] <superdump> and staff
[12:38:46] <kshishkov> and security!
[12:39:03] <superdump> their security seems to be pretty quick in my experience
[12:39:06] * thresh shudders
[12:39:18] <superdump> but then i guess i go at times when not so many other people are going
[12:39:22] <lyakh> mru: pong
[12:39:34] <mru> lyakh: about your sh4 optims
[12:40:01] <mru> have you considered writing at least the functions called through pointers in plain asm .S files?
[12:40:05] <kshishkov> thresh: а вы летайте самолетами Аэрофлота
[12:40:33] <lyakh> mru: yeah... but only very lightly;)
[12:40:46] <mru> consider it again, and more profoundly
[12:40:56] <mru> it's much easier to work with than inline
[12:40:57] <lyakh> I looked at sh ABI, the function calling convention is not very... transparent
[12:41:04] <lyakh> at least from the first sight
[12:41:06] <mru> can't be that bad
[12:41:09] <thresh> kshishkov: :S
[12:41:13] <lyakh> no, not _that_ bad
[12:41:19] <mru> first few params in registers, rest on stack
[12:41:35] <lyakh> but bad enough for me to leave it for the first instance;)
[12:41:40] <merbzt> lyakh: why do you work on sh4 ?
[12:42:15] <lyakh> merbzt: well, let's put it this way: it's not just-for-fun
[12:42:17] <lyakh> ;)
[12:42:48] <kshishkov> lyakh: if GCC can generate it why can't you?
[12:42:50] <lyakh> mru: but what are reasons to prefer .S to inline?
[12:43:06] <mru> avoid the clutter of inline asm
[12:43:09] <mru> and avoid gcc
[12:43:15] <mru> also works with non-gcc compilers
[12:43:16] <lyakh> kshishkov: sure I can... just didn't see the point to begin with
[12:43:30] <lyakh> mru: last argument I consider void
[12:43:32] <mru> better register allocation
[12:43:35] <lyakh> for sh
[12:44:33] <lyakh> dunno, so far all optimizations were _inside_ functions
[12:44:47] <mru> where else would they be?
[12:45:09] <mru> all the functions with pointers in e.g. dsputil are easy to write in pure asm
[12:45:12] <lyakh> so, I thought, that overhead that gcc might add at the top and bottom of inline fragments shouldn't be bad enough
[12:47:33] <mru> you'd be surprised
[12:48:28] <lyakh> mru: do you have a specific example, where you think, that converting a certain function X to pure asm would bring a measurable advantage?
[12:48:45] <mru> I didn't read the patches carefully
[12:49:04] <mru> inline asm is bloody hard to read too
[12:53:30] <siretart> lu_zero: no reason to hide this in private. when will you arrive?
[12:54:36] <lu_zero> 17:25
[12:54:37] <siretart> my plane arrives on saturday at 10:35 at Brussels National Airport. I'll then take the cab to fosdem
[12:55:02] <lu_zero> I see
[12:55:07] <siretart> ok. where do I find you all on fosdem then?
[12:55:18] <siretart> I think i have diego's mobile number
[12:55:30] <lu_zero> siretart: I'll be there on Friday
[12:55:34] <kshishkov> siretart: BeagleBoard booth, I presume
[12:55:41] <siretart> kshishkov: excellent. I'll find that
[12:55:42] <lu_zero> so I think we'll meet there
[12:56:11] <mru> my phone number can be found using standard internet services
[12:56:12] <kshishkov> siretart: and I'll arrive there only an hour earlier than you
[12:56:22] <mru> those who are unable to figure it out are not worthy of calling me
[12:56:26] <siretart> kshishkov: but via train, I persume?
[12:56:31] <lu_zero> siretart: you should have already one of mine
[12:56:38] <kshishkov> siretart: of course
[12:57:27] <kshishkov> mru: once you spilled your inofficial Swedish number as well here
[12:57:27] <mru> lu_zero: I have two numbers for you, one ending in 6295, one in 2494
[12:57:43] <mru> that one is harder to find
[12:57:47] <mru> it's not listed anywhere
[12:57:52] <siretart> mru: is the one ending on ..032 up-to-date?
[12:58:06] <mru> +447900001032
[12:58:25] <siretart> that'
[12:58:29] <siretart> s what I've found :-)
[12:58:54] <cartman> everybody call mru now :p
[12:59:13] <mru> cartman: it's your phone bill
[12:59:33] <kshishkov> mru: who knows how Turkish Telecom works, it may bill you instead
[12:59:37] * cartman has 7$ on Skype
[12:59:42] <cartman> kshishkov: LOL
[12:59:50] <cartman> that could happen
[13:09:18] <siretart> our download page contains obsolete pieces of information about svn snapshots. ok to remove? http://pbot.rmdir.de/86f1ce5124c19a4ebdf592343d1bad4a
[13:09:51] <mru> fine with me
[13:10:16] <mru> btw, http://svn.ffmpeg.org/ redirects to git.ff.o
[13:11:14] <superdump> siretart: send a patch to the ML :)
[13:11:26] <siretart> superdump: seriously?
[13:12:04] <mru> it's just removing old cruft from a web page
[13:12:05] <mru> commit it already
[13:12:09] <superdump> well, i don't care personally, but we said that stuff should be reviewed
[13:12:16] <siretart> aaaaand, it is gone
[13:12:18] <mru> code, yes
[13:12:43] <siretart> superdump: it just had been reviewed and ack'ed by mans :-)
[13:12:58] <mru> removing obviously wrong statements from the web page is hardly anything to debate
[13:13:51] <kshishkov> too bad sometimes it's differently obvious to different people
[13:14:02] <superdump> website svn commits only go to -cvslog and not -commits
[13:15:49] <kshishkov> that makes sense too
[13:16:20] <andoma> re the float_to_int16 discussion the other day.. i think the codecs should just output samples in their native format
[13:16:39] <kshishkov> which is?
[13:17:09] <mru> andoma: float output does not preclude scaling it
[13:17:27] <kshishkov> I'd use -1.0..1.0 range
[13:17:43] <andoma> mru: yeah, scaling of float (where it is "for free") can be offered by the codec
[13:18:03] <andoma> s/can/should/
[13:18:20] <kshishkov> my codec can scale by 0 for free!
[13:19:10] <merbzt> my by 1
[13:19:44] <cartman> we get the idea :P
[13:20:14] <kshishkov> and proper one should output numbers in range [2.718281...; 3.141592...]
[14:45:00] <ruggles> andoma: that is the goal, but currently the generic conversion doesn't use the fast functions. once it does the codecs can output what fits best. and even better if we get good planar audio support.
[14:49:23] <BBB> mru wanted planar audio also
[14:49:33] <BBB> I'm all for it, most decoders output planar audio internally anyway
[14:49:42] <mru> yes, planar output is fine
[14:49:55] <mru> something, somewhere still has to interleave it of course
[14:50:08] <mru> but there's no inherent advantage of doing it early
[14:51:56] <BBB> some avaudiofilters (avseq, anyone?) could also work better on planar audio maybe
[14:52:42] <kshishkov> whatever - we'll just need final audioconvert filter
[14:53:37] <mru> filters in general are easier to do planar
[14:54:01] <kshishkov> especially with SIMD
[14:58:01] <cartman> KotH: http://www.kamikazecookery.com/blogs/210
[14:58:56] <kshishkov> cartman: smells like pferdmist
[14:59:16] <cartman> kshishkov: pferwhat?
[14:59:27] <kshishkov> cartman: horse manure
[14:59:40] <saste> astrange: your repo is missing libswscale
[14:59:46] <cartman> kshishkov: ...
[15:00:03] <kshishkov> cartman: that's how they call it in Northern Sauteuerland
[15:00:16] <KotH> cartman: 10 out of 9 people are complete utter idiots... does that mean i have to boycot the world?
[15:00:48] <cartman> KotH: I do so
[15:00:49] <cartman> :D
[15:01:08] * KotH boycots cartman 
[15:01:30] <cartman> :(
[15:02:15] <kshishkov> cartman: number of idiots in Sauteuerland is lower only because it has less people than Turkey
[15:04:28] * cartman hopes to be part of the 1 out of 10
[15:05:55] <spaam> cartman: he said 10/9 ... not 9/10...
[15:06:16] <kshishkov> spaam: he hopes to be that -1 out of 9
[15:06:17] <cartman> spaam: tricky bastard KotH
[15:06:27] <spaam> cartman: i know.
[15:07:05] <spaam> kshishkov: ah :)
[15:54:43] <lu_zero> we do have audio mixing algos in ffmpeg already, isn't it?
[15:54:58] * mru looks around, sees none
[15:55:01] <superdump> i don't think so
[15:55:55] <Tjoppen> nope. not even audioconvert.h is public
[15:56:05] <kshishkov> mixing as in what?
[15:56:17] <spaam> mixtapes
[15:57:14] <av500> lu_zero: out = (L + R) / 2
[15:57:16] <av500> done
[15:57:57] <kshishkov> some decoders have builtin downmixing and upmixing
[15:58:00] <cartman> av500: 2.0
[15:58:02] <cartman> :P
[15:58:12] <av500> kshishkov: yes, dca and ac3
[15:58:19] <av500> and wmapro
[15:58:31] <av500> at least the ones I have :)
[15:58:46] <kshishkov> upmixing 1->2 is also present in some decoders
[16:06:34] <lu_zero> uhmm
[16:06:48] <lu_zero> av500: L + R - L*R you mean =P
[16:22:52] <kshishkov> mru: just in improbable case you've overlooked it - http://www.cc65.org/
[16:28:19] <mmu_man> 6502 \o/
[16:28:25] <mmu_man> You want to port ffmpeg to ORIC ?
[16:29:47] <av500> mmu_man: not all the world knows your french computer
[16:29:58] <av500> only you and ex .yu :)
[16:34:30] <mmu_man> av500: ORIC was british
[16:34:41] <mmu_man> only when the british screwed up was it bought by french :p
[16:35:07] <av500> ah
[16:35:23] <benoit-> mmu_man: isn't that also true for some colonies? :)
[16:35:33] <Tjoppen> ooh, I'll have to look at cc65
[16:36:28] <Tjoppen> assuming it plays well with dasm it should be useful with my atari 2600 experiments
[16:47:44] <mmu_man> Tjoppen: hmm I recall writing a sed script to convert the output of one 6502 disassembler to the one used by cc65
[16:48:07] <mmu_man> I probably put it on miniserve...
[16:48:08] <mmu_man> cf. http://forum.defence-force.org/viewtopic.php?t=635
[16:48:08] <Tjoppen> I'd rather do the opposite
[16:48:55] <mmu_man> http://miniserve.defence-force.org/svn/users/mmu_man/sedoric/ca2xa.sed
[16:49:14] <mmu_man> might still be helpful I hope :)
[16:50:00] <Tjoppen> cool
[16:50:37] <Tjoppen> I wonder if it makes use of the undocumented opcodes. I've found some to be particularly useful, like SAR
[16:51:41] <mmu_man> I used it to disassemble and reassemble SEDORIC automatically
[16:51:50] <mmu_man> the idea was to add Jasmin2 support...
[16:51:53] <mmu_man> never had time
[16:52:06] <mmu_man> cf the makefile http://miniserve.defence-force.org/svn/users/mmu_man/sedoric/
[16:54:02] <lu_zero> firefox has a really strange build system
[16:54:53] <mru> tell us something new
[16:55:13] <av500> at least it has one
[16:56:30] <lu_zero> mru: it seems to be broken right now
[16:56:42] * lu_zero workarounds it...
[16:57:04] <mru> "works around"
[16:59:36] <lu_zero> note taken
[17:02:45] <mru> not "takenoted" :)
[17:03:03] <mmu_man> mans' tree :=D
[17:03:04] <av500> notetaked!
[17:03:22] <mmu_man> or shortened "naked" :)
[17:04:36] <lu_zero> isn't for not acknowledged?
[17:05:18] <av500> NACK'ed
[17:06:03] <lu_zero> jrmuizel: you!
[17:06:08] <mru> I thought it was "nekkid"
[17:06:25] <lu_zero> I'm having some really strange issue building firefox
[17:06:56] <lu_zero> apparently it fails because LIBXUL_LIBS doesn't have -lcairo
[17:07:04] <lu_zero> rings any bell?
[17:07:30] <mru> warning bells, loud ones
[17:09:02] <j-b> s/dont/done
[17:09:19] <mru> yeah, that one didn't quite parse the first time
[17:17:20] <Compn> kshishkov : you seen 'rare exports' yet? horror movie about santa clause
[17:17:28] <Compn> finnish i think
[17:20:11] <peloverde> It looked awesome but I have not seen it yet
[17:26:36] <mru> jannau: patchwork seems to hate mime-encoded subject lines
[17:27:24] <av500> can somebody have a look at that flv index patch?
[17:30:12] <jannau> mru: looks so
[17:30:42] <elenril> jannau: please fix the commit message for the mov patch
[17:30:46] <elenril> it's not a regression
[17:31:44] <elenril> those -- 8< -- lines is some patchwork magic?
[17:31:52] <mru> git magic
[17:32:15] <jannau> elenril: John said it worked before your patch
[17:33:57] <jrmuizel> lu_zero: ?
[17:34:01] <jannau> err, misrembered. will fix the commit message
[17:38:11] <lu_zero> jrmuizel: I'm trying to build firefox from a snapshot a little after beta9 and I got this strange behaviour
[17:38:37] <lu_zero> the workaround I'm using is adding -lcairo to the flags
[17:39:14] <jrmuizel> lu_zero: weird
[17:39:22] <jrmuizel> file a bug?
[17:39:37] <lu_zero> jrmuizel: I wanted to confirm it first
[17:39:42] <jrmuizel> sure
[17:39:50] * jrmuizel heads to lunch
[17:41:38] * lu_zero is thinking about heading to bed...
[17:47:57] <mru> BBB: what are the constraints on the 'h' argument to vp8 epel MC?
[18:07:50] <TheFluff> does av_realloc() guarantee aligned memory?
[18:08:07] <mru> what are you doing?
[18:08:39] <TheFluff> using it in an application that links against ffmpeg
[18:08:39] <mru> but the answer is no
[18:08:50] <TheFluff> ok
[18:09:22] <TheFluff> I'm reading frames from a file to a buffer, enlarging it when needed
[18:09:32] <TheFluff> and then feeding that buffer to lavc
[18:09:44] <TheFluff> so I guess I shouldn't do that and use av_malloc instead
[18:09:50] <mru> why don't you allocate enough to begin with?
[18:10:00] <mru> realloc is slow
[18:10:01] <TheFluff> because I don't know the maximum frame size
[18:10:26] <TheFluff> I guess I could scan the file first but that's probably even slower
[18:10:46] <mru> start with a sensible guess, then every time you find a larger frame, allocate a bigger buffer and keep that size
[18:10:59] <mru> you might add an exponential factor too
[18:11:09] <TheFluff> that's what I'm doing, except I start at 0 instead of a sensible guess
[18:11:17] <mru> if you do that, you shouldn't need to enlarge it more than a few times
[18:11:40] <TheFluff> lazy coding uses realloc the first time too so I don't have to handle the unallocated case in a special way
[18:11:48] <TheFluff> but I guess I should Do It Right
[18:11:56] * mru bets TheFluff doesn't handle realloc failure properly
[18:12:24] <TheFluff> I do, I raise an error saying "out of memory"
[18:12:27] <TheFluff> ;V
[18:12:38] <TheFluff> (hasn't been triggered yet to my knowledge)
[18:12:41] <mru> and leak the old buffer
[18:12:50] <TheFluff> no, and abort
[18:14:53] <mmu_man> re
[18:16:51] <ruggles> TheFluff: regarding starting at 0, many movies start with black & silence which could have small frame sizes for VBR, so it could still be way too low even after the first realloc.
[18:17:36] <TheFluff> yes, the current implementation is highly inefficient
[18:17:46] <TheFluff> I'm very aware of this
[18:18:26] <av500> what is the issue to alloc 256k or so for the frame?
[18:19:25] <TheFluff> none, I just inherited lazy coding that does it in the laziest way possible, which is if (framesize > bufsize) enlarge_buffer(framesize + some constant) where enlarge_buffer calls realloc
[18:19:53] <TheFluff> init the bufsize to 0 and you don't have to do anything special at all
[18:20:00] <TheFluff> no special cases etc
[18:21:06] <av500> + constant -> *= 2
[18:21:49] <TheFluff> yeah but that still leaves me with unaligned memory so I really should take the time and write the five lines of code required to do it right
[18:34:42] <Compn> peloverde : i just finished watching it. very strange movie.
[18:44:40] <BBB> mru: h is either 4, 8 or 16, and it is the same as the width, or half or double of that (i.e. for width=8, you can have 8x4, 8x8 or 8x16)
[18:45:01] <BBB> mru: also, if width == 16, then x and y (last two arugments) are always even
[18:45:11] <BBB> mru: so you don't need to implement fourtap for width == 16
[18:48:21] <mru> and for width 4, h >= 4?
[18:50:36] <BBB> h is either 4 or 8
[18:50:42] <BBB> for width == 16, h is 8 or 16
[18:50:50] <BBB> for width == 8, h is 4, 8 or 16
[19:33:47] <j-b> I am so sad I must be missing all the fun
[19:35:26] <Jumpyshoes> fun?
[19:35:54] <mru> Jumpyshoes: sarcasm detector broken?
[19:36:15] <j-b> Jumpyshoes: sorry, missing the <irony> tags
[19:37:08] <Jumpyshoes> lol, i'm bad at sarcasm
[19:37:21] <mru> learn, quickly
[19:38:32] <j-b> Jumpyshoes: it will be a useful tool for your career, srsly
[19:38:54] <j-b> Jumpyshoes: just try to avoid the being-too-cynical-to-appreciate-life trap
[19:39:47] <Jumpyshoes> i should learn
[19:44:30] <thresh> it's not a trap really
[19:44:37] <thresh> but the only possible way of survivng
[19:46:17] <j-b> no, life is Awesome!
[19:46:45] <j-b> you need to be a bit cynical and use irony&sarcasm, but you need to be positive about life
[19:49:20] * thresh puts off his Marvin hat
[19:55:21] <j-b> thresh: I must re-read those books
[20:00:49] <_av500_> j-b: but in the original russian!
[20:01:09] <thresh> j-b: I like the movie too
[20:01:41] <j-b> av500: ? ?? ?????? ??-??????
[20:04:39] <thresh> yeah that's what I say in the mornings
[20:05:52] <thresh> when i'm like that: http://www.youtube.com/watch?v=umxQfF4dobw
[20:23:16] <mru> http://blog.regehr.org/archives/364
[20:31:32] <j-b> rofl
[20:31:45] <j-b> yet another angry mail from top HP executive to me
[20:32:02] <j-b> for "VLC is destroying speakers of our laptops"
[20:32:26] <_av500_> lol
[20:32:33] * elenril always knew vlc is evil
[20:32:54] <roxfan> wow, how do you do that?
[20:33:19] <j-b> roxfan: well, the UI allows you to go above 100% of the audio input
[20:33:32] <roxfan> nice
[20:33:33] <j-b> roxfan: so the sound will saturate (obviously)
[20:34:01] <j-b> roxfan: but, saturating sound is still outputted through Win32 drivers to their hardware
[20:34:21] <j-b> I am trying to tell them that we aren't modifying the output
[20:34:40] <mru> does it cause physical damage?
[20:34:47] <j-b> destroys the speaker
[20:34:50] <j-b> according to them
[20:34:51] <mru> if so, the laptops are rather poorly designed
[20:34:59] <j-b> my opinion too
[20:35:04] <j-b> we are using Win32 APIs
[20:35:08] <j-b> not hooking the drivers
[20:35:13] <mru> laptop speakers are like 0.5W
[20:35:46] <mru> it's not like driving a few kW too much from a power amp into a too small speaker
[20:35:48] <j-b> so, I've received the 20th mail on the thread, and they are getting angry
[20:35:59] <j-b> and angrier
[20:36:16] <iive> tell them it is driver problem.
[20:36:22] <thresh> lol
[20:36:34] <j-b> I mean, it is like if I was listening to some German Metal group
[20:36:37] <thresh> j-b: what do they want from you?
[20:36:39] <j-b> the sound saturared
[20:36:44] <j-b> thresh: fix our code
[20:37:00] <j-b> so they "can actively promote VLC"
[20:37:04] <j-b> I lol'd on that one
[20:37:23] <j-b> iive: well, to me too, it seems obvious
[20:37:26] <elenril> haha
[20:37:38] <elenril> wait until they threate to actively promote mplayer instead
[20:37:41] <thresh> j-b: well tell them you did fix the code
[20:37:47] <_av500_> j-b: make em pay
[20:37:48] <thresh> and watch them burn another laptops
[20:37:50] <thresh> :)
[20:38:01] <wbs> j-b: ask what they'll do if you don't fix it, add a "please don't use VLC on this computer, it might break" sticker on all computers? :-)
[20:38:09] <j-b> elenril: well, VLC, mplayer, this is the same
[20:38:21] <wbs> that'd sell well
[20:38:30] <j-b> wbs: lol
[20:38:41] <j-b> but, I am a bit afraid that I am missing something
[20:39:35] <j-b> but what?
[20:40:15] <ruggles> this is float output?
[20:41:02] <j-b> f32l or s16l
[20:41:28] <ruggles> how can s16l go over 100%?
[20:41:41] <mru> how can anyone be stupid enough to build a system capable of self-destructing?
[20:41:41] <_av500_> s17l
[20:41:49] <roxfan> 2.0 = 200% ?
[20:41:52] <mru> just put a goddamn power limiter in the output stage
[20:41:52] <j-b> ruggles: well, it doesn't, obviously
[20:42:16] <j-b> ruggles: can it?
[20:42:28] <mru> of course not
[20:42:33] <mru> s16 uses the full range
[20:43:31] <j-b> so, s16l is out of problem I guess
[20:43:34] <j-b> so is spdig
[20:43:44] <j-b> *spdif*
[20:44:02] <mru> I doubt the hw accepts anything other than s16
[20:44:23] <j-b> mru: so, am I crazy?
[20:44:29] <_av500_> no
[20:44:44] <_av500_> their amp stage is crap
[20:44:55] <_av500_> they must limit the sound
[20:44:55] <j-b> seeing the code, we try fl32 before s16l
[20:45:02] <mru> j-b: do you know which part is breaking?
[20:45:07] <_av500_> even we do that
[20:45:09] <mru> output stage or speakers?
[20:45:15] <_av500_> speakers
[20:45:18] <_av500_> i guess
[20:45:21] <j-b> mru: they claim "speakers destroyed"
[20:45:33] <mru> coming from an executive that could mean anything
[20:45:42] <_av500_> no bandpass on the speaker stage
[20:45:46] <mru> it's his way of saying "no sound" in a more techy-sounding way
[20:45:49] <_av500_> even we do that
[20:46:04] <j-b> "Open Source Program Office",  (ITO CT &
[20:46:49] <j-b> "Open Source Program Office",  "ITO CT & CA and Open Source Lead", "Open Source and Linux Technology Architect", "central CTI"
[20:46:58] <j-b> I mean, those guys are not kidding :)
[20:47:27] <j-b> serious business
[20:48:14] <_av500_> j-b: always ask for a sample product
[20:48:16] <j-b> and of course, I know next to nothing in audio
[20:48:26] <j-b> av500: free laptop!
[20:48:32] <_av500_> yup
[20:48:44] <j-b> I would prefer a new server
[20:48:49] <j-b> so I can host a ffmpeg-fork
[20:48:54] <j-b> </sarcasm>
[20:49:03] <mru> j-b: ask him to test vlc on a server
[20:49:05] <mru> hope it breaks
[20:49:17] <j-b> well, www.v.o is a HP server
[20:52:21] <j-b> still, to me the fault is theirs
[20:54:22] <{V}> why is an executive even contacting you about it. Sounds like something one of their more technically inclined people should address (and probably not by contacting you)
[20:57:32] <ruggles> which systems don't HAVE_6REGS?
[20:58:10] <mru> x86_32 with pic and -fno-omit-frame-pointer or something
[20:59:49] <ruggles> ok. vector_fmul_window sse & 3dn2 require 6regs, but with memcpy+5regs it's still faster than C on my system.
[21:07:21] <lu_zero> nice =)
[21:08:03] <mru> how can you replace a mul with memcpy?
[21:20:01] <Compn> j-b : ohhh, wheres the thread about bad laptop speakers? :D
[21:20:05] <Compn> or is it private mail hehe
[21:20:07] <kierank> Yuvi: do bugs in qt aac encoder get fixed if we report then
[21:20:08] <kierank> them*
[21:20:28] <BBB> Yuvi is online?
[21:20:29] <BBB> ooo
[21:20:31] <BBB> hi yuvi
[21:21:07] <ruggles> mru: just memcpy src0 to dst then load/save to dst
[21:21:22] <mru> how can that be faster?
[21:21:31] <ruggles> faster than the C version
[21:22:08] <ruggles> sse2 6regs: 6671 - sse2 5regs: 8761 - C: 12721
[21:22:21] <mru> ah, now you're making sense
[21:22:32] <mru> but who cares about not having 6 regs free?
[21:22:43] <mru> besides, it can just be written in yasm instead
[21:22:45] <mru> problem solved
[21:23:40] <ruggles> true
[21:23:41] <BBB> x86-32 always has 6 regs
[21:23:42] <j-b> Compn: private mail
[21:23:48] <BBB> just put it under #ifdef HAVE_7_REGS or so
[21:23:53] <BBB> we don't care about broken systems
[21:24:32] <ruggles> it's currently under HAVE_6REGS
[21:24:56] <ruggles> and falls back to C otherwise
[21:25:13] <ruggles> i was just wondering if it's worth it to optimize the 5 register case
[21:25:22] <mru> it's not
[21:25:39] <mru> that only happens in some bizarre combination of bad choices
[21:25:58] <ruggles> good enough for me
[21:30:24] <j-b> Compn: i can fwd you the mails
[21:30:35] * BBB agrees with mru
[21:32:22] <Compn> j-b : sure, patriotact at gmail.com
[21:32:39] <j-b> what?
[21:32:44] <Compn> my email
[21:34:03] <Compn> to forward the mails to...
[21:34:40] <spaam> eh what? dont you have a rr.com mail?
[21:35:11] <kierank> ah yes compn's epic email
[21:35:19] <Compn> i do, mostly used for the mailing list
[21:35:20] <lu_zero> which one?
[21:35:34] <Compn> but i use the gmail when i post it in public because it has better spam protection
[21:35:48] <Compn> kierank : glad someone likes it :)
[21:36:02] <spaam> Compn: since when is ffmpeg-devel not public?
[21:36:06] <Compn> its a constant reminder that at any time, my government can read what books i take out of the library
[21:36:26] <Compn> spaam : well that and i signed up for the list before i got gmail
[21:36:59] <spaam> ok :)
[21:37:36] <Compn> projects at mplayerhq.hu forwards to my gmail too
[21:38:19] <Compn> i was getting ffmpeg project emails about every month until you guys hid the projects page on the about navigation
[21:39:03] <Compn> er nm, i cant remember if that mail was ever up there
[21:39:13] <spaam> :)
[21:49:50] <BBB> wbs: ping
[21:49:58] <wbs> BBB: pong
[21:50:01] <BBB> \o/
[21:50:06] <BBB> wbs: do you feel like cleaning up https://github.com/rbultje/ffmpeg/commit/4900f0ec82720229c7293c242114555d3843dccc ?
[21:50:35] <BBB> there's quite some work left to do, it's a asf rtp payload parser when streamed by realmedia servers (I know, don't ask)
[21:50:50] <BBB> it crashes on exit, there's a few hacks that may be preventable, etc.
[21:51:03] <BBB> sample stream is provided in the commit message
[21:51:47] <wbs> BBB: sure, I can give it a look in a while
[21:52:20] <BBB> awesome
[21:52:33] <BBB> I'm gonna clean up some other patches to push into that tree
[21:59:40] <mru> ruggles: do you want a login on my ppc?
[22:00:16] <ruggles> sure
[22:00:30] <mru> send an ssh pubkey
[22:00:38] <mru> and pick a username
[22:02:53] <ruggles> sent
[22:04:10] <mru> ruggles: ssh saracen.mansr.com should work
[22:04:20] <mru> usual dev tools are available
[22:04:26] <mru> samples in /misc/samples/
[22:04:35] <mru> mphq and some other random stuff
[22:04:57] <mru> machine is dual g4 800-ish MHz
[22:05:03] <ruggles> awesome. thanks.
[22:07:03] <ruggles> how do i send my patch?
[22:07:18] <mru> send it where?
[22:07:36] <ruggles> to the ppc. so i can test it.
[22:07:38] <mru> you have git, push and pull to your heart's content
[22:13:05] <ruggles> i can setup a git clone on the ppc, but how do i make it accept push from local?
[22:13:16] <mru> from local?
[22:14:02] <mru> on your pc, do "git push saracen.mansr.com:path/to/clone master"
[22:14:05] <mru> or something like that
[22:14:28] <mru> wait, it'll bitch about pushing into a checked-out branch
[22:14:43] <lu_zero> BBB: http://ffmpeg.pastebin.com/SRRxSQSx
[22:14:48] <lu_zero> I hate it
[22:14:53] <lu_zero> got better ideas?
[22:15:54] <wbs> lu_zero: uh, can't we allocate it once we know rt->nb_rtsp_streams at startup?
[22:16:02] <lu_zero> wbs: even more ugly
[22:16:31] <BBB> exactly
[22:16:38] <BBB> it's ok to do that, but not in fetch_packet
[22:16:43] <BBB> malloc during init is fine
[22:17:00] * j0sh would like a way to add/remove streams at runtime
[22:17:34] <ruggles> mru: what if i just init an empty repo on the ppc, then push?
[22:17:44] * elenril would like a pony
[22:17:51] <mru> ruggles: it'll still be checked out
[22:18:03] <mru> you can push to a temp branch and merge that into master
[22:18:32] <ruggles> ok. i'll give it a shot.
[22:28:41] <Yuvi> kierank: well, unreported bugs probably won't get fixed
[22:28:41] <Yuvi> beyond that, "probably"
[22:28:41] <Yuvi> BBB: hey
[22:28:52] <ruggles> mru: worked fine. i checked out a temp branch then pushed to master.
[22:29:04] <kierank> Yuvi: yes we reported it
[22:29:05] <kierank> thanks
[22:29:21] <mru> Yuvi: you mean they'll be snuck into a later update under "improve bulgarian translation"
[22:29:44] <Yuvi> mru: usually it's "bug fixes" as the entire changelog ;)
[22:30:07] <mru> bios updates are really bad like that
[22:30:19] <mru> the official release notes always seem totally innocent
[22:30:28] <mru> yet they mysteriously fix things
[22:30:33] <bcoudurier> merbanan, are you around ?
[22:30:39] <merbanan> yes
[22:32:06] <lu_zero> j0sh: that isn't a problem where you add the change logic you can add the av_realloc ^^;
[22:33:25] <lu_zero> http://ffmpeg.pastebin.com/tdvNNdPE <- better?
[22:36:12] <wbs> lu_zero: yes, looks better to me. won't you need to allocate it somewhere else for the rtsp demuxer, though? that only allocates it for the sdp demuxer
[22:36:59] <lu_zero> wbs: rtsp->sdp->rtp
[22:37:40] <wbs> lu_zero: did you test it yet? ;P sdp_read_header isn't called anywhere in the rtsp demuxer
[22:37:49] <lu_zero> uh?
[22:37:51] <lu_zero> meh
[22:39:51] <lu_zero> ff_sdp_parse fooled me =P
[22:41:22] <lu_zero> ff_rtsp_setup_input_streams ...
[22:41:46] <lu_zero> remind me why we have it in another file while it is called in the other one...
[22:42:36] <wbs> since it's only used in the rtsp demuxer (while it's called by common rtsp code in rtsp.c), it was moved there as part of trying to split it up a bit more in october
[22:43:44] <lu_zero> right
[23:08:04] <mru> BBB: so which vp8 epel variants are not used?
[23:08:34] <mru> I don't see width4 h4 or v4 in my profiles
[23:11:24] <Yuvi> those are for if either the x or y component has no subpel at all iirc
[23:11:44] <Yuvi> but the other is epel
[23:12:53] <mru> the function pointer tables have entries for h6, v6, h4, v4, h6v6, h6v4, h4v6, h4v4 for widths 4, 8, and 16
[23:13:07] <mru> but some of them don't appear to be used (much)
[23:13:47] <mru> BBB said width 16 4-tap isn't used
[23:13:51] <Yuvi> yeah
[23:13:55] <Yuvi> 4 tap is chroma-only
[23:14:28] <Yuvi> and I'd imagine most chroma mvs have subpel for both components
[23:14:45] <Yuvi> or none at all
[23:15:11] <mru> what are the mixed 4/6-tap ones for?
[23:15:43] <Yuvi> qpel position for (x,y), epel for the other
[23:15:50] <Yuvi> x or y I mean
[23:30:56] <jannau> is the git-svn ffmpeg repo still somewhere?
[23:31:39] <mru> it's still there
[23:31:44] <mru> but out of public view
[23:32:18] <j-b> hello jannau
[23:37:41] <Dark_Shikari> oh awesome, more ffflames
[23:39:54] <jannau> hi j-b, I still haven't tested the crystalhd decoder in vlc
[23:40:52] <j-b> still, point 8) is quite right
[23:41:30] <Sean_McG> I hear those are awesome
[23:42:15] <j-b> crystald?
[23:42:31] <Sean_McG> aye
[23:42:37] <jannau> that sounds like a wild exaggeration. I would say they work to some extent
[23:43:14] <Sean_McG> well as in they let a dinky Atom play a 1080p stream
[23:43:32] <j-b> Sean_McG: they are not too bad
[23:43:34] <jannau> and the driver has much room for improvement
[23:43:41] <j-b> Sean_McG: and the API is broken, but not too much
[23:43:49] <j-b> but the drivers is not up to standards yet
[23:46:52] <jannau> mru: and the non-public way to access it?
[23:47:07] <mru> same git url as before
[23:47:16] <mru> why?
[23:49:18] <jannau> indeed, I assumed ffmpeg and ffmpeg.git are the same repos
[23:49:30] <mru> sorry for the confusion
[23:49:33] <lu_zero> got fooled as well ^^
[23:49:43] <Sean_McG> ditto
[23:49:54] <mru> we'll make them the same eventually
[23:50:15] <mru> but I didn't want to break things for anyone overnight
[23:50:36] <Sean_McG> I thought that's what mainline development was all about?
[23:50:40] * Sean_McG ducks and runs!
[23:54:22] <Sean_McG> anyone know where I should ask about metadata tagging for a show like oldskool Doctor Who, where you had separate "stories" within a season?
[23:55:43] <mru> ask elenril, he's our master of all things meta
[23:56:33] <mru> not too familiar with old dr who, are those a plot line spanning several episodes?
[23:56:47] <mru> but several such multi-episode stories per season
[23:56:50] <Sean_McG> aye
[23:57:27] <mru> I guess I'd just tag them "title: part X" or something
[23:57:41] <mru> of if sw can handle it, part as a separate tag
[23:58:04] <Sean_McG> I suppose I could do that... I tried with episode ID but iTunes still doesn't sort them properly