**** BEGIN LOGGING AT Wed Jul 07 02:59:56 2010 Jul 07 03:43:13 wow Jul 07 03:43:44 so how well does it run on an n900? Jul 07 07:22:21 Is there a easy way to get NXClient on Meego for netbooks? Jul 07 08:04:52 MAKANKOSAPPO! Jul 07 08:04:55 Há! Jul 07 08:15:32 morning all Jul 07 08:16:11 morning slaine Jul 07 08:19:21 hi guys Jul 07 08:19:41 I am torn between using webkit or the mx-toolkit for this project http://whyareyoureadingthisurl.wordpress.com/2010/06/14/introducing-the-wasiliana-mail-client/ Jul 07 08:19:54 could someone give me the advantages of using mx-toolkit over webkit? Jul 07 08:20:03 (and HTML/CSS interface) Jul 07 10:52:06 ugh why is it i always fall over with the damned browser package? with moblin i couldnt get things to work using firefox, now with meego i cant get things to work with chromium :( Jul 07 10:52:53 FunkyPenguin: oh, but you have a change here to get things not working with fennec as well Jul 07 10:53:28 Aard, im not even looking at handheld yet! Jul 07 10:53:52 i need to get netbook out the door first, then i'll look at handheld (if i get the hardware) Jul 07 10:54:28 netbook works fine for basic testing, if you don't Jul 07 10:54:53 * FunkyPenguin has 1 week to get things working and built Jul 07 12:05:33 <_krypton> FYI : Video of the MeeGo running on a Lenovo Pad at http://www.youtube.com/watch?v=6muBs8LUlYU Jul 07 12:06:10 this is a pretty old version of Meego , so just dont mind the flakiness Jul 07 12:40:58 " Jul 07 12:41:00 Also as many have stated, Moorestown is not really ready yet for Mobile space consumption. The next die-shrink will be the first real Intel foray into the mobile handset space." Jul 07 12:41:10 -vasra2 Jul 07 12:46:26 7" Cortex-A8 tablet (android) $200-250 http://shanzai.com/index.php/bandit-gadgets/tablets/1360-first-look-a81-sub-usd-200-mountable-tablet-sports-removable-battery-android-21-cortex-8-processor-and-features-galore Jul 07 12:59:10 re: moorestown, probably the most true thing said about the entire platform as of yet. Jul 07 13:31:04 woohoo, we're no.1 in ireland Jul 07 13:31:06 http://www.irishtimes.com/business/tech50/ Jul 07 13:31:36 congrats! Jul 07 13:31:53 Thanks Stskeeps Jul 07 13:31:59 load of BS though ;) Jul 07 13:32:24 hehe Jul 07 13:32:29 slaine: Congrats .... calls for a beer ;) Jul 07 13:32:46 I think so Jul 07 13:32:55 We're due to have a company bbq soon actually Jul 07 13:40:07 Hey folks. Does anyone know how to install rdesktop on meego? Jul 07 13:40:44 My boss loves meego, we just need to see if we can install rdesktop somehow, and if we can, we're going to put it on all of our check out netbooks. Jul 07 13:41:18 should be fairly easily to port Jul 07 13:41:22 if not a rpm exists already Jul 07 14:10:11 hi Jul 07 14:10:21 is MeeGo netbook based on a distro? Jul 07 14:10:55 meego core ;) Jul 07 14:14:01 i know the question may be stupid :) i cant test it atm and i would like to know if i can work on it (develop) Jul 07 14:14:31 "work on it" ~ code my projects Jul 07 14:14:41 there's a netbook sdk Jul 07 14:15:44 wasnt talking about developing meego, but developing on it Jul 07 14:15:48 yeah, i know Jul 07 14:15:50 * josemoreira feels like a total newbie Jul 07 14:17:33 http://wiki.meego.com/Getting_started_with_the_MeeGo_SDK_for_Linux Jul 07 14:17:46 well im gonna steal someones pen and try it Jul 07 15:17:57 when are official meego meetings? Jul 07 15:19:47 #meego-meeting Jul 07 15:28:16 FACEPALM Jul 07 15:34:58 rektide: there's a schedule on the wiki Jul 07 15:40:07 and here's a link to the meeting schedule: http://wiki.meego.com/MeeGo-Meeting_IRC_Schedule Jul 07 15:46:13 DawnFoster: would you like me to transfer ownership of #meego-dev over to you? Jul 07 15:46:41 b-man: probably a good idea, thanks! Jul 07 15:46:55 b-man: can you also add stskeeps? Jul 07 15:47:03 sure :) Jul 07 15:47:23 1 sec, gotta re-login ;P Jul 07 15:47:27 thanks, again. Things are so spread out right now, I'm not sure who owns what :) Jul 07 15:50:08 hmm, giveownership is not recognized :S Jul 07 15:50:47 /chanserv set #meego-dev founder dawnfoster Jul 07 15:50:56 ah Jul 07 15:50:58 ty Jul 07 15:52:11 b-man17: this is why I always give stskeeps access, too - he seems to know all of the IRC commands :) Jul 07 15:52:22 i call it brain damage Jul 07 15:52:22 :P Jul 07 15:52:27 lol Jul 07 15:52:36 :-))) Jul 07 15:52:37 can i still be a mod? :) Jul 07 15:52:38 ircmaster Jul 07 15:52:39 :P Jul 07 15:52:43 god, this connman bug is annoying.. Jul 07 15:52:51 b-man17: sure Jul 07 15:52:59 ty :) Jul 07 15:53:38 * b-man17 gives now Stskeeps ownership Jul 07 15:53:46 *now gives Jul 07 15:54:08 if you already switched to dawn, think it's too late :) dawn should /chanserv access #meego-dev add stskeeps +F Jul 07 15:54:19 oh Jul 07 15:54:33 woops Jul 07 15:54:35 once the switchover is done Jul 07 15:54:49 DawnFoster: did you get ownership? Jul 07 15:55:08 still noted as you as founder, what did chanserv respond to you? Jul 07 15:55:35 no, I don't have access (I just tried to add stskeeps) and got (notice) You are not authorized to execute this command. Jul 07 15:55:51 ok, i'll rty again Jul 07 15:55:55 *try Jul 07 15:56:01 DawnFoster: execute the following: /msg ChanServ SET #meego-dev FOUNDER DawnFoster Jul 07 15:56:04 that should complete the step Jul 07 15:56:39 (a good way to avoid getting blamed for suddenly being owner of #goatse, they verify you actually want it..) Jul 07 15:56:54 lol Jul 07 15:56:55 im so glad magical command line incantations still exist Jul 07 15:57:00 stskeeps: that worked :) Jul 07 15:57:10 there we go, yes Jul 07 15:57:21 :) Jul 07 15:57:52 stskeeps: I just gave you +F access Jul 07 15:57:53 is everyone set? xD Jul 07 15:58:00 DawnFoster: k Jul 07 15:58:21 b-man17: thanks, I think we got it all straightened out Jul 07 15:58:36 awesome :) Jul 07 15:59:08 b-man17: and thanks again for offering to transfer it! Jul 07 15:59:44 no problem :) Jul 07 16:00:03 i think it's better off with someone official anyhow xD Jul 07 16:10:55 thiago_home: thank you for hitting the nail on the head with a cruise missile :) Jul 07 16:12:57 * thiago_home wonders what nail he hit Jul 07 16:13:24 your latest reply to the openness thread Jul 07 16:17:54 I've just sent the email Jul 07 16:17:59 * thiago_home wrote it on the plane Jul 07 16:23:50 arjan: probably a stupid question, but how would you enable system-wide coredumps in meego core? Jul 07 16:24:06 am debugging a connmand issue (crash on startup) Jul 07 16:24:27 and yes, i tried ulimit -c unlimited before running it Jul 07 16:30:24 Stskeeps: check /proc/sys/kernel/core_pattern Jul 07 16:30:39 on the N900 by default it has a path that doesn't exist, so the core dumps go nowhere Jul 07 16:30:46 :nod: Jul 07 16:30:55 also note you need write perms on the current directory Jul 07 16:31:22 sounds obvious, but the first time I tried core-dumping on the N900, it wasn't creating anything (over NFS-mount) Jul 07 16:31:54 arjan: are you around? Jul 07 16:33:38 im booting meego on an asus-pe1050 and all i get is a black screen Jul 07 16:33:54 this is after boot menu Jul 07 16:34:08 running from a pen Jul 07 16:34:25 (maybe that _is_ meego?) :) Jul 07 16:35:04 :) Jul 07 16:35:26 +1 for power consumption Jul 07 16:36:44 removed 'quiet' from the boot entry, kernel boots and then Jul 07 16:36:48 BSOD again Jul 07 16:36:55 lol Jul 07 16:37:03 or, BSOS Jul 07 16:37:11 bsos? xD Jul 07 16:37:13 Black Screen Of something Jul 07 16:37:23 dunno, cant see Jul 07 16:37:25 :p Jul 07 16:37:42 maemo can have the wsod xD Jul 07 16:37:44 what graphics chip does that hardware have ? Jul 07 16:37:59 ehum good question Jul 07 16:38:43 its a asus 1005 pe Jul 07 16:38:46 let me see Jul 07 16:39:15 josemoreira: you trying to run meego? Jul 07 16:39:22 yes Jul 07 16:39:54 does your processor have SSSE3 support? Jul 07 16:40:20 dunno its an intela tom Jul 07 16:40:24 :D Jul 07 16:40:42 EeePC 1005PE: http://meego.com/devices/netbook/supported-hardware-platforms Jul 07 16:41:09 josemoreira: can you ssh into it? Jul 07 16:41:40 have no idea Jul 07 16:41:53 dont know the ip Jul 07 16:42:36 http://en.wikipedia.org/wiki/Intel_Atom says it has ss3, i guess Jul 07 16:42:42 I don't think ssh is enabled by default Jul 07 16:42:53 The blank screen is likely a graphics driver issue Jul 07 16:42:56 http://en.wikipedia.org/wiki/Intel_Atom#N4xx_Series Jul 07 16:43:05 slaine, i believe so also Jul 07 16:43:06 Is this from the live key or after the install Jul 07 16:43:14 live usb key Jul 07 16:43:30 can you switch to a VT ? Jul 07 16:43:35 it goes blank when trying to display the gfx bootloader i guess Jul 07 16:43:51 ahhhhhhhhh Jul 07 16:43:56 it was the brightness :P Jul 07 16:44:01 :) Jul 07 16:44:08 forgot about that Jul 07 16:44:10 that's a first Jul 07 16:44:12 even on linux i have probs Jul 07 16:44:43 ususally i need to add acpi=linux_vendor or somehting to the kernel Jul 07 16:44:54 ps.: whats the password? :) Jul 07 16:45:27 rofl Jul 07 16:45:39 josemoreira: username is "root" and password is "meego" Jul 07 16:46:18 makes sense :) Jul 07 16:46:30 woo cooal Jul 07 16:47:18 CosmoHill: and network login is enabled by default? :) Jul 07 16:54:27 you mean ssh? Jul 07 16:55:49 yep Jul 07 16:59:18 i really wonder why noone just makes a netbook .ks, ask to put NVIDIA drivers in same directory, .. bam, a nvidia netbook image.. Jul 07 16:59:53 heh Jul 07 17:00:06 I'll suggest that to NVidia tomorrow Jul 07 17:00:10 (some generalisation and abstraction needed) Jul 07 17:00:20 it is all libGL afterall Jul 07 17:00:22 rather, Friday Jul 07 17:00:26 I have a meeting with them Jul 07 17:00:34 of course, I won't be able to tell you anything about their reaction Jul 07 17:00:57 Stskeeps, at the very least enable the nouveau driver in the kernel Jul 07 17:01:01 :S Jul 07 17:02:16 send a patch? :P Jul 07 17:02:32 moo all Jul 07 17:02:57 any meego kernel guys in here? Jul 07 17:03:05 thiago_home: have a meeting with someone who doesn't play poker Jul 07 17:05:36 trip0: http://meego.gitorious.org/meego-os-base/kernel-source , patch away Jul 07 17:06:26 Stskeeps, i've already got a clone for that Jul 07 17:06:35 i need to figure out the "right way (TM)" to do it though Jul 07 17:06:41 cuz i dont' think i did it right Jul 07 17:32:46 nouveau is actually blacklisted too trip0 Jul 07 17:32:55 just so you're aware Jul 07 17:33:07 slaine: ?? Jul 07 17:33:12 why? Jul 07 17:34:07 arjan: the module blacklist as nouveau in there Jul 07 17:35:33 I black listed nouveau when I installed nvidia on redhat Jul 07 17:35:42 but why would it be black listed on meego? Jul 07 17:36:18 that'd be a bug Jul 07 17:42:48 arjan: did I tell you I implemented the second-based timer for Qt? Jul 07 17:44:35 how do i logout? :) Jul 07 17:45:09 josemoreira: you close the lid or press the power button to turn off Jul 07 17:45:50 it's not a multi-login system Jul 07 17:46:05 silly unix users Jul 07 17:46:06 slaine, ok, it was to change language settings (running from pen drive) Jul 07 17:46:08 :P Jul 07 17:47:12 thiago_home, second-based? Jul 07 17:47:20 as opposed to millisecond-based Jul 07 17:47:29 theres me thinking it was girlfriend code Jul 07 17:47:51 you mean for operating tick tock clocks etc? Jul 07 17:48:14 what does it gain practically that the ms based one didnt cope with before? Jul 07 17:50:00 thiago_home: nice Jul 07 17:50:32 system wide synchronized? Jul 07 17:50:38 speaking of qt timers Jul 07 17:50:47 does qt know if the application has focus Jul 07 17:50:50 or is on the screen at all ? Jul 07 17:50:51 synchronised to the monotonic clock's full-second (0 milliseconds) Jul 07 17:50:57 (and does it get notification if that changes?) Jul 07 17:51:14 not in the basic event loop, but the X11 dispatcher can set that info Jul 07 17:51:22 basically Jul 07 17:51:22 you can get focus events for windows Jul 07 17:51:33 we can do timer rounding in the kernel Jul 07 17:51:43 to 10 ms? (MONOTONIC_COARSE) Jul 07 17:51:45 and can do it precise if visible, waaay roudned up if not visible Jul 07 17:51:48 yeah Jul 07 17:51:52 APIs don't cahnge Jul 07 17:52:00 how do I tell the kernel? Jul 07 17:52:02 the rounding happens kernel side, and will coincide with other events Jul 07 17:52:05 or should I change the timer in Qt? Jul 07 17:52:08 via prctl() Jul 07 17:52:20 lcuk: system wide seconds based timers would mean that apps with timers wake up on whole seconds only, instead of one app waking up at 1.25, other at 1.35, etc. When everything wakes up simultaneously, there's a bigger period of cpu idle between timer events, which saves much power Jul 07 17:52:33 (but I'm just guessing) Jul 07 17:52:49 prctl() with PR_SET_TIMERSLACK to be honest Jul 07 17:53:12 lcuk: many apps have "every 10 seconds more or less" kind of timers Jul 07 17:53:23 lcuk: there's huge value in aligning them all up to the same point Jul 07 17:53:25 sure Jul 07 17:53:27 10000 Jul 07 17:53:29 (power value) Jul 07 17:53:37 well, I implemented three timer modes in Qt Jul 07 17:53:43 by making the API in seconds, it means you give explicit permission to round Jul 07 17:53:49 the HighPrecision one is the current model, which is exact to the millisecond Jul 07 17:53:55 that's fine Jul 07 17:53:57 the VeryCoarse one is the second-based one Jul 07 17:54:05 it always rounds the timer to a second and wakes up in a full second Jul 07 17:54:07 so this isnt useful for clocks Jul 07 17:54:18 lcuk: no but for many other backgroudn things it is Jul 07 17:54:27 the third one is an adaptative mode called simply Coarse Jul 07 17:54:46 thiago_home: the prctl() one is nice in that it impacts the thread as a whole, for any timers it does Jul 07 17:54:49 it will use a 5% error to make the timer wake up at defined intervals Jul 07 17:54:58 so youc an turn it coarse/precise based on global events, from a local decision point Jul 07 17:55:33 so if you're at 40412.123 and ask for a 400 ms timer, it will try to use a multiple-of-200 ms boundary Jul 07 17:55:56 ah nice Jul 07 17:56:04 slowly self aligning Jul 07 17:56:34 the nice thing about the prctl is that it gives timers a range Jul 07 17:56:43 but if some event happens inside that range, the timer will happen right then Jul 07 17:56:44 so it will wake up at 40412.333 (210 ms later), 40412.543, 40412.753, 40412.963, 40413.173, 40413.383, 40413.593 Jul 07 17:56:54 so if you set a 20 msec range, you'll effectively align on vblank events etc Jul 07 17:56:56 then it wakes up at 40413.800 and from then on precisely 200 ms Jul 07 17:57:08 How new a distro do I need to find a prctl man page? :) It sounds interesting Jul 07 17:57:22 wouldnt people wanting vblank somehow tie into vblank event etc? Jul 07 17:57:25 * thiago_home doesn't have PR_SET_TIMERSLACK in his prctl(2) manpage Jul 07 17:57:34 arjan: anyway, what I want is to set the timerslack per select/poll Jul 07 17:57:38 since theres no way of knowing which system you are running on or even if you have a vblank Jul 07 17:57:44 meego should 'remember' the last top icon people clicked, after returning from an application Jul 07 17:57:45 i.e., the app is going to sleep, it knows the timers pending and how slack they are Jul 07 17:57:55 otherwise, I need two syscalls per sleep Jul 07 17:59:17 thiago_home: agreed... but Ulrich said nobody would use it, and he'd not support it in glibc until there was a user Jul 07 17:59:25 if Qt will use it then we can go to uli together and plead the case ;) Jul 07 17:59:35 * thiago_home raises his hand Jul 07 17:59:35 if Qt has a meaningful way of doing this... that'd be good Jul 07 17:59:54 my current code only handles the first (soonest) timer Jul 07 18:00:08 what happens if the soonest timer has a more lax constraint than the next one? Jul 07 18:00:23 what you do is sort the timers on "start + slack" not on "start" Jul 07 18:00:36 basically the end of the contraint time is the hard deadline and sort on that Jul 07 18:00:44 hmm Jul 07 18:00:48 of course, I could determine a max slack (20 ms?) and look into the timers there Jul 07 18:00:50 that does not work if userspace does it as well Jul 07 18:01:06 although it'll be correct if you sort by end of range Jul 07 18:01:14 and then pick the timer that pops up first and just use that Jul 07 18:01:19 its not optimal Jul 07 18:01:22 * arjan needs to ponder more Jul 07 18:01:29 (for kernel we have more liberty) Jul 07 18:01:32 what exactly are you trying to do? Jul 07 18:01:41 * CosmoHill hands arjan some coffee to help the pondering Jul 07 18:01:42 I spent several hours considering the best algorithm for Qt Jul 07 18:02:26 the current algo is just for synchronising wakeups Jul 07 18:02:51 since the objective is that all apps synchronise to the same point in time, it shouldn't be a waste Jul 07 18:03:52 for example, plasma-desktop has tens of 2-second timers. Without my patch, it was causing 3.3 wakups/sec. Jul 07 18:04:10 and now? Jul 07 18:04:15 with the patch, my laptop produced 0.9 (I don't know where powertop got that number) Jul 07 18:04:40 debugging info revealed it slept for 1.5 sec, woke up, slept 500 ms, woke up, restarted the cycle Jul 07 18:04:42 thiago: you do know that is is 0.9 wakeups pers second more than needed? =) Jul 07 18:05:24 on the laptop, it took more than 500 ms to initialise all timers, which meant that some timers got aligned to a different 500-ms boundary Jul 07 18:05:35 on my workstation in the office, all 62 2-second timers were sync'ed Jul 07 18:05:51 RST38h: of course. I can't help if you try to shoot yourself in the foot. :-) Jul 07 18:06:34 thiago: seriously, is it the mobile version or the netbook? Jul 07 18:06:41 plasma-desktop Jul 07 18:06:46 not plasma-mobile or plasma-netbook Jul 07 18:06:57 ah, got it Jul 07 18:07:18 Ok, let us see... Jul 07 18:07:30 You have got N processes, each having a set of timers Jul 07 18:07:46 Timers inside a process are supposed to be synced by phase Jul 07 18:08:05 But between sets, you do not care Jul 07 18:08:20 between apps you still care Jul 07 18:08:24 hence roudning to a sytem timer Jul 07 18:08:34 I am not sure that I care Jul 07 18:08:41 my battery cares Jul 07 18:08:43 If an app wants to tie itself to the system timer, it can Jul 07 18:08:44 and thus I care Jul 07 18:08:53 Wait, let us forget the battery for now Jul 07 18:08:59 no, let's not forget the battery Jul 07 18:09:05 lets not forget the battery Jul 07 18:09:06 the battery is the entire objective of this exercise Jul 07 18:09:18 Battery has nothing to do with what I am currently trying to figure out Jul 07 18:09:30 if it weren't for the battery, we'd try to get apps to wake up at different times, so as to not compete for CPU time Jul 07 18:09:39 Consider a thread creating several timers. That thread probably cares about the phase between its timers Jul 07 18:09:52 what do you mean by phase here? Jul 07 18:09:54 sometimes Jul 07 18:09:54 BUT, if we look at two threads or two tasks or two apps, they do not care Jul 07 18:10:06 most timers in practice are very very coarse and more for background maintenance tasks Jul 07 18:10:07 thiago: period between timer0.start tim and tier1.start time Jul 07 18:10:22 and the app really does not care how/when it runs Jul 07 18:10:30 if the thread really cares about that, you can use the HighPrecision timer Jul 07 18:10:36 The way I understand it, you want to merge as many individual timers together as you can Jul 07 18:10:39 most apps that I've seen don't Jul 07 18:10:46 so that one wakeup could trigger multiple timers in many apps Jul 07 18:10:49 right? Jul 07 18:10:51 right Jul 07 18:10:52 arjan: is there an OBS setup guide for meego for companies wanting to tailor meego for their hardware ? Jul 07 18:10:57 RST38h: that's the inverse proposition of what we try to do here Jul 07 18:10:59 abby working on the gibbs case of when he shot the drug cartel guy Jul 07 18:11:11 RST38h: we try to redcue the number of times we wake up to service timers Jul 07 18:11:12 arjan: you want to have as many wakeups as possible? Jul 07 18:11:27 RST38h: grouping timers together is a means to get there Jul 07 18:11:32 RST38h: what you're saying is that if I start a 100 ms timer now, wait 23 ms, then start another 100 ms timer, the thread wants the second timer to always wake up 23 ms after the first one? Jul 07 18:11:44 but there really is no difference if you wake up 2 times per second, if you have 10 - 10 or 18 - 2 Jul 07 18:11:48 those distributions are equivalent Jul 07 18:11:59 ok Jul 07 18:12:12 thiago: correct Jul 07 18:12:15 while under a grouping objective, the 18-2 would be favoured Jul 07 18:12:25 RST38h: well, my patch would undo that Jul 07 18:12:41 RST38h: the way I see it, if you want to keep the phase between timers, you need to use some construct to do that Jul 07 18:12:51 thiago: I suspect you want to keep that, at least in one thread Jul 07 18:13:00 arjan: if you're interested, this is the current core algo: http://pastebin.ca/1896153 Jul 07 18:13:02 RST38h: only if you explicitly ask for it Jul 07 18:13:13 RST38h: also if you don't keep synchronizing, you will not keep that 23 msec Jul 07 18:13:17 ok, so all timers are independent Jul 07 18:13:23 exit latencies and other system tasks will skew your timer Jul 07 18:13:24 yes, timers are independent Jul 07 18:13:34 even simpler Jul 07 18:13:38 there's no mechanism to make them lock to each other right now Jul 07 18:13:52 and you have not got alarmd yet? Jul 07 18:13:55 thiago_home: it seems Intel IT has blocked pastebin.ca Jul 07 18:14:02 damnit Jul 07 18:14:16 cliffhanger ending of the season of ncis Jul 07 18:14:23 arjan: proxy-us lets it through Jul 07 18:14:25 bastards Jul 07 18:14:30 (or did, a few hours ago) Jul 07 18:15:35 arjan: http://qt.pastebin.com/dDuTmB4g Jul 07 18:16:01 this time with syntax highlighting Jul 07 18:16:11 the pastebin plasmoid doesn't set the highlighting to C++ Jul 07 18:16:52 looks reasonable Jul 07 18:17:09 hmm... lines 33 and 40 are wrong Jul 07 18:17:22 they should really be += Jul 07 18:19:16 arjan: I'd love to discuss with you how to pass the current coarseness to the kernel Jul 07 18:19:51 and I'm not afraid of making syscalls myself :-) Jul 07 18:20:03 just another select() system call with an extra argument Jul 07 18:20:18 it's not liek we don't already have 4 variants of poll/select anyway Jul 07 18:20:27 would epoll be more useful? Jul 07 18:20:28 well poll not select, lets pretend select doesn't exist for new syscalls Jul 07 18:20:30 nah Jul 07 18:20:37 poll is already ms-based Jul 07 18:20:40 epoll solves the "webserver with thousands of connections" Jul 07 18:20:48 depends which poll you use Jul 07 18:20:58 I haven't debugged why, but with poll(2), plasma-desktop wakes up less often than with select(2) Jul 07 18:21:22 thiago_home: that's odd Jul 07 18:21:22 ah, ppoll has timespec Jul 07 18:21:40 ppoll() yeah Jul 07 18:22:05 one extra detail: I don't control the actual sleep when the Glib mainloop is enabled Jul 07 18:22:14 different animal Jul 07 18:22:23 is that the common model ? Jul 07 18:22:38 yes, if Qt was compiled with Glib support Jul 07 18:22:39 like required for dbus or something Jul 07 18:22:45 no, it's not required for D-Bus Jul 07 18:22:57 it's required by the Gtk+ style and the GStreamer backends to Phonon and QtMultimedia Jul 07 18:23:24 the former I care less about Jul 07 18:23:32 but gstreamer for qtmultimedia we require Jul 07 18:23:36 yeah Jul 07 18:23:49 it can be disabled (i.e., use the Qt loop) by use of an environment variable Jul 07 18:23:53 the netbook theme should really be native not some gtk+ thing Jul 07 18:23:58 but gstreamer is essential Jul 07 18:24:44 the event loop selection is per-thread. I don't know if that helps the gstreamer case. Jul 07 18:25:41 on the subject: should we use eventfd instead of pipes for "hey, thread, wake up" ? Jul 07 18:27:28 is there a meeting tonight? Jul 07 18:28:02 good point, I meant to ask that too :-) Jul 07 18:31:10 thiago_home: doesn't look like there is one Jul 07 18:31:17 I'd imagine that too Jul 07 18:31:32 thiago_home: does eventfd have tangible gains? Jul 07 18:32:34 "The kernel overhead of an eventfd file descriptor is much lower than that of a pipe, and only one file descriptor is required (versus the two required for a pipe)." Jul 07 18:32:37 from the manpage Jul 07 18:33:19 but, of course, that creates the problem I described to you once, w00t_ Jul 07 18:33:37 * w00t_ thinks back Jul 07 18:33:42 it means we use different codepaths on Linux, thereby testing less the others Jul 07 18:33:43 signalfd would be good Jul 07 18:33:46 right Jul 07 18:33:49 but glib and qt don't have any signal API Jul 07 18:33:52 aah Jul 07 18:33:55 Robot101: signalfd is to get a signal firing Jul 07 18:33:58 my screen went weird Jul 07 18:34:06 Robot101: there's an API for that for the QWS version of Qt Jul 07 18:34:06 yeah, but nobody actually solves it atm, which is crap Jul 07 18:34:25 ooh Jul 07 18:34:29 would be good to move to core Jul 07 18:34:46 it's just a (Qt) signal emitted from QCoreApplication when the (Unix) signal is caught Jul 07 18:35:08 do you need to get a signal? Jul 07 18:35:10 any tips for a tripe boot install with Meego? i have win7 and Mint atm Jul 07 18:35:16 triple Jul 07 18:35:40