/srv/irclogs.ubuntu.com/2012/05/14/#ubuntustudio-devel.txt

len-dtUsing polkit is almost too easy...00:21
len-dtWith regard to changing runlevel as a user. I was able to to set up telinit to run from user space... so I can use runlevels to control what performance level I am running at.00:24
len-dtThis is probably not a good idea as telinit has more function than just changing runlevel.00:24
len-dtHowever, it would be easy to write an application who's only job is to take user input and allow only three or four actions... changing runlevels to one of rl2-rl5.00:27
len-dtAnd throw away anything else... even a bash script could do that.00:28
ailolen-dt: I read your post on the mail list before, but haven't had the time to devote to US today. Did you read about Ulatencyd on another post?00:33
len-dtI have. It is hard to know what is useful for audio. What would be really helpful would be test cases.00:34
len-dtBoth test cases that have lots of stress and those that are more "realistic".00:34
len-dtThat would help us to find out is any of these things are really helping or needed.00:35
scott-workgood morning everyone12:25
len-dtscott-work, been a while since I've seen that name here ;-)13:17
knomeyeah, but he's not really on :<13:23
len-dtknome, he probably has lots of catch up to do.13:35
scott-worksorry, i'm back again14:00
scott-worki really would like to talk about blueprints this week and get them resolved and to kate (release manager) by the end of the week14:01
scott-worklen-dt: would you like to head up an "OTB improvements" blueprint?14:01
scott-worki shoudl be posting to the mailnig list with a list of things i think we need to get done along with some others might might consider14:01
scott-workof course, i want people's suggestions and input as well14:01
ailoscott-work: Me and Len have been putting some work on system tuning. Mostly len, so far. And we've set up pages on the wiki for documenting both what we have now, what is possible, and lastly, what we would like to have14:43
ailoI think we should also go more deeply with package selection. Making more detailed analysis of work flows14:43
ailoProbably not a lot missing, but it would be good to document that throuroughly as well14:44
ailoscott-work: The wiki pages are still mostly empty. The three last items on this page https://wiki.ubuntu.com/UbuntuStudio/TeamResources14:45
scott-workailo: that's good to hear and i agree about package selection as well14:46
scott-workailo: shnatsel had some interesting ideas as well14:47
ailoscott-work: I think len-dt and me will go ahead and explore those things, and hopefully that will result in some good analysis of what might or might not be useful for tuning the system for better performance14:51
scott-workthat would be out-f'ing-standing :)14:51
ailoI think we'll probably need a -controls application in the end. Would you agree len-dt?14:51
scott-workailo: i was going to suggest that as well14:51
scott-worki've not been a fan of the -controls idea for a while, but it has mainly because no one seemed to be able to clearly define its purpose anymore14:52
scott-workbut it sounds like we are clearly defining some purposes now14:53
ailoAs len-dt suggested in hit post, I think, many things can not be done in userland. You set it, and reboot, or at least restart your session. Also, you'll want to be able to toggle things separately, since some settings only apply to some machines. A -controls application could also inform the user how each setting can help them. 14:53
ailoWell, time to do some laps in the park :P14:58
ailolen-dt: I think the first thing I want to do is to finish working on kernel related docs and set up kernel builds for testing on my PPA. Get back to you when I have something up for testin/reading16:12
scott-workit looks like the kernel time will most likely assume responsiblity of the -lowlatency kernel if they can collapse some of the other kernels they are supported16:13
scott-workwell, some are going to be collapsed (i.e. combined) and others are going to be dropped16:13
ailoIt's probably not going to make much of an impact for anyones performance, I don't think, tweaking a few kernel options, but who knows. And I just want to learn all about it to be sure16:13
ailoscott-work: I think that's great. I do hope we can have some serious input into what configs the kernel should have though16:14
scott-workailo: i would hope so...and i think we have time for that as well still16:14
scott-workthey need to get he blueprints done this week as well, then go through the work of minimizing their supported kernels before they address ours16:14
ailoscott-work: Let's have a meeting before then?16:15
scott-workabsolutely!16:15
scott-worki was hoping this weekend actually so we can finalize the blueprints and some other ideas/tasks16:16
* micahg waves to scott-work16:16
scott-workailo: len-dt:  i don't know if you know about my blog, it has several noted items from UDS including things related to studio16:16
ailoscott-work: Haven't read it for a while. What was the link again? ;P16:16
scott-workmicahg: hi!  i reversed my position, after we work through bluepritns i want to help with syncing and FTBFS as soon as possible to get the exposure16:17
* scott-work is sorry for being wishy-washy and this subject micahg, but has his mind made up very firmly now16:17
scott-workailo: http://dullass.blogspot.com/16:17
scott-works/and this subject/on this subject16:21
scott-workheh, micahg left already :/16:23
knomehah:)16:23
knomeyou want to fail to build from source?16:23
scott-workknome: i want to learn16:23
scott-work;)16:23
knome;)16:23
scott-workailo and len-dt :  feel free to leave comments in the blog, don't be shy :P16:24
knomescott-work, so, where's the blog articles for days 3-5 ?16:24
scott-worki'm still working on them :P  i'm ADHD and sometimes have trouble keeping my attention on certain things16:25
scott-worki have notes from all of it and i will be finishing them up in the next two days16:25
ailoscott-work: You only have one letter combination? I have all of them16:26
knomescott-work, diagnosed ADHD, or non-literallt ADHD?16:26
knome*non-literally16:27
scott-workknome: literally, i have been diagnosed and i am taking adderall16:34
scott-worki still get a little flaky by getting distracted and hyper at times (i'm kinda feeling that now)16:35
knomescott-work, okay, good to know :)16:35
scott-workbut i can sustain much better concentration in the short term though16:35
scott-workbefore i HAD to also follow all the conversatiosn that others are having around me as well as my own conversation16:35
knomewell i do that too...16:36
scott-workmicahg: i see you are back in channel18:53
micahgwell, about to go off for a few hours :)18:53
astraljavascott-work: You? Hyper?!19:35
scott-workhehe, sometimes very badly19:35
scott-workastraljava: sometimes is because very, very obvious because i can talk really, really fast at times and it might seem that i don't breath between paragraph-like sentences19:36
astraljavaI will have to travel to Texas to witness that. :D19:37
scott-workwe'd be glad to have you stay at the house if you did19:40
astraljavaWe have a few employees in the States, so I may take you up on that offer. :D19:42
scott-workthat would be super cool :)19:44
astraljavaIndeed it would. I haven't been further into your continent than... Atlanta. :)19:44
astraljavaOne of them lives and works in Colorado. That's at least where I'd wanna go. It's not TOO far to hop over to your place from there. :)19:45
astraljavaFortunately our company is very pro-travelling.19:46
len-dtailo, controls or setup or config or whatever. Ya. any performance mode thing needs a config app so that the user can make it work for their particular use. Things like wireless or not. Mysql or not... are things atht come to mind.20:18
astraljavaNot even just wireless, but networking overall.20:21
len-dtscott-work, ailo while a lot of things are out of userland, polkit is easy to use to get around that. The easiest and most secure method would be to use runlevels. That way up to 4 different setups can be defined. I have been able to make strong changes without either booting or relogging in.20:22
len-dtastraljava, exactly. 20:26
astraljavaI was reminded again at work today how vulnerable even modern unices are to networking problems.20:27
astraljavaUI freezes totally, even on supposedly well-working multitasking OS.20:28
astraljava...on a dual-core machine.20:28
knome..with memory built from stardust?20:28
len-dtastraljava, the thing with networking is things like qmidinet and qmidicontrol (I think I got it right) and for that matter lyricue (which also uses mysql :-P20:28
astraljavaknome: Have you been inhaling something you shouldn't, again?20:29
astraljavalen-dt: gah!20:29
astraljavaBut okay, I get the point.20:30
knomenot admitting anything.20:30
len-dtastraljava, breathing you mean like city air?20:30
astraljavaWell, he lives considerably closer to downtown than I do.20:30
knomethough you have that dust coming in all the time20:31
knomei live considerable closer to a (small) forest too20:31
knomes/e/y20:31
len-dtWhere we used to live downtown had the cleaner air and the wind used to blow the crap up the valley to the farmland.20:31
astraljavaYeah I guess that's true. There are many more buses going just underneath my balcony than anywhere near his apartment.20:34
knomeso it's probably just too clean air i'm breathing... or the methane :|20:34
astraljavaRight, that's it.20:34
astraljavaNo wonder you married your wife, it's because of her profession. :D20:35
knome;]20:35
len-dtastraljava, it is unlikey that someone would use lyricue along with audio applications, but quite possible that a musician would have lyricue and audio apps on the same machine. lyricue being a video app that displays video and lyrics.20:35
knomeyeah, good to have backup if i faint of my own smell20:35
astraljavalen-dt: Right, I can understand that.20:36
astraljavaBut then there are all those other runlevels.20:36
astraljavaLike you said already.20:36
len-dtYup, mysql seems to be runlevel friendly, NetMan is not. But it does handle being shutoff ok.20:37
len-dtAnd NM could be shut off (and turned on) by a runlevel friendly upstart script.20:38
len-dtastraljava, I would like to do as much as I can that is upstart friendly (or driven) as that seems to be the future of things. Any of the upstart *.conf can have a *.override file that takes it's place without deleting the original... very handy. Also the override file does nothing if there is no conf file to match... also handy as it means if we (or the user) tries to shut off something not installed there is no problem.20:44
astraljavalen-dt: Sounds good indeed.20:49
scott-worki wonder how regular ubuntu devs would view us considering runlevels21:00
len-dtscott-work, should be ok. I think RL2 should be the same as vanilla ubuntu and there should always be a path back to that.21:01
scott-worklen-dt: that's good to know. i now very little about runleves so i worry about such things21:02
scott-worklen-dt: do you think you would be up to leading a blueprint for OTB performance improvements?21:02
len-dtscott-work, right now rl 3,4 and 5 are the same as rl2. That is not used.21:03
len-dtscott-work, OTB?21:03
scott-workout of the box21:03
scott-workor Out of The Box :P21:03
len-dtscott-work, I can try. You mean how the setup is at install or live DVD?21:04
scott-workwell, kinda21:04
scott-workthere isn't too much pressure to it, to be honest21:04
scott-workdo you know much about blueprints?21:05
scott-worklen-dt: ^^^21:05
len-dtscott-work, I get the idea of what they are for... but not much on the formal language of them.21:05
scott-workformal language is minor aspect, don't worry about that21:06
scott-workso, to show our intended work for the cycle we make blueprints and the work items within the blueprints21:06
scott-workthis one might have only a few generalized work items like "explore possible ways to improve OOTB experience" and "implement what we can"21:06
len-dtSo I can come up with a doc, which you will vet and post?21:06
scott-workor might be more details about which options we might try and such21:07
scott-workyeah, i would then vet it then it needs to go up the chain to the release manager who would approve it and make it part of our formal "ubuntu studio for q cycle blueprint", which acts like a header blueprint21:07
scott-workyou don't get in trouble if not everything is completed or if things don't work out and there is nothing we can do to improve the OOTB experience21:08
scott-workbut i can't keep up with every blueprint and make sure progress is going on, especially when i'm not very familiar with some of the things you are talking about21:08
len-dtscott-work, ok. There are two streams of thought with this, one that with a modern system there is performance to spare anyway and the other which is we can do much better with tweaks.21:09
scott-worklen-dt: well, if we can reasonably and safely make tweaks and yield appreciable improvement, i say we do it21:09
scott-workor at least consider it21:09
len-dtscott-work, I know I can make magrinal machines better (like my netbook for example), but they are less noticable on a good machine.21:10
len-dts/magrinal/marginal/21:10
scott-worklen-dt: that might be a good work item - to verify which improvements yield good results on which machines21:10
scott-workbtw, https://help.launchpad.net/Blueprint for those who want to learn more about blueprints21:10
len-dtscott-work, thanks.21:11
scott-workso, we could say one work item is to brainstorm which tweaks might work21:11
scott-workanother to test them21:11
scott-workanother to make a decision which to explore for reals21:11
scott-workanother to implement one (or several)21:11
scott-workanotehr to validate the improvements21:11
scott-work(oh, i suppose one might be to make a baseline reading earlier)21:12
len-dtYes, scott-work I have already found a tweak on one of the linux audio/music pages that does not work as stated in ubuntu.21:12
scott-worki think generally each work item is suppose to be a day or three of worth of work21:12
ailoTo get to the bottow with all of the system configs, we need much more time than that.21:13
len-dtscott-work, right now I am trying to gather info on what tweaks others have tried and to see if they work for me.21:14
scott-workailo:  more time than...?21:14
ailoFor each work item21:14
scott-workoh, right...i understand :)21:14
len-dtIt looks like I will need to at the very least get more memory just for testing.21:14
scott-workthe documentation will be the same ailo 21:14
scott-workuser documentation, developer documentation, testing docuementation, all that stuff21:15
ailolen-dt: Remember that I have 5 machines or more, so we could just as well do a team effort on the testing. 21:15
scott-worklen-dt: we set the goals/work items for the blueprint, if you feel that we will only get through testing and making a suggestion for next cycle then that is okay21:15
scott-worklen-dt: we should just set a reasonable goal for Quantal (including some buffer probably) and try to get that done21:16
len-dtscott-work, there are some things that have already been proven to work for some people, I would like to include those. I will be using some of them in my machine anyway.21:16
astraljavaI get your point, BUT IT'S THE RELEASE AFTER LTS!! :D21:16
scott-workastraljava: LOL, yeah i know, many people want to go crazy!  but i don't want to scare len away :P21:17
scott-worklen-dt: that's fine, and keep in mind that i'll help you with developing the work items if you want21:18
len-dtastraljava, while this may have been LTS, I think we are really part way into the vision of where we want to go at this time.21:18
astraljavascott-work: You'll scare him away by putting on restraints. :D21:18
len-dtastraljava, scott-work :-)21:18
astraljavalen-dt: No, you misunderstood me. I fully support you! :)21:18
astraljavaLet's rock! :D21:18
len-dtastraljava, just laughing along with you21:19
astraljavaGood. :)21:19
scott-workhehe, good point astraljava  :)21:20
scott-worki took it the other way :P21:20
scott-worklen-dt: but seriously, the work items are really another minor issue, but it would be very helpful to have you (who understands these items more than i) driving this blueprint21:21
scott-workor ailo could do it, or you both could21:21
scott-worki just wanted someone else but me (who doesn't have time to learn all the stuff you already are working on) to drive the blueprint21:21
scott-worki can still make the blueprint and with your help fill int he work items21:22
len-dtscott-work, astraljava I am thinking that even though my workflow app is not very polished, I will use it to set up rl based performance modes. The gui is throw away (or not) but the mode switching is pretty basic. The config for it can be more of the same or use the US controls framework if someone feels more comfortable working with that.21:22
scott-workoh, one reason i haven't been on much lately is not only because i was at UDS but also because my old laptop died21:23
len-dtI will try to come up with an overall document in the next few days.21:23
scott-worki am unsure what i am going to set up at the house21:23
scott-worklen-dt: don't stress about getting final form or even doing a blueprint at this point, we have until early next week to get it done21:23
len-dtscott-work, netbooks are cheap... and do surprisingly well.21:23
scott-workwe can brainstorm about what things we want to explore and then start parsing into steps and then get that into the blueprint21:24
scott-workthere will also be other blueprints, like some of the documentation and live-dvd fixes/improvements21:24
len-dtanyway, I am off for a bit... it's about 2:30 time to go get kids.21:24
scott-worki am leaving for home as well21:24
ailoI'm fine with len-dt doing the blueprints. I'm still on documenting kernel configs, and will probably continue with that and the wiki for the rest of the week. I will however think about what we could bring up at the next meeting, when we discuss the next cycle21:25
ailoIf we are to make good use of each others machines for testing, I believe documentation will be essential, so that would be the best place to start21:27
astraljavaI was just recently shown a link to some test automation, but haven't had a chance to check it out yet.21:32
ailoastraljava: Standard type of testing? Give us a link if you find it21:33
ailoI mean standard, like what you would do for regular Ubuntu, not multimedia21:33
astraljavaailo: I'll look into it and see how it could be utilized.21:35
astraljavaBut not now, I have to sleep.21:35
ailosweat sheep21:36
ailosweet*21:36
astraljavaNo I don't do sheep, sweaty or not. That's jussi's territory.21:37
ailoHave a good restful night of sleep then21:39
knomenighty astraljava 21:44
len-dtWho is good to talk to about security matters? I want to use pkexec to run telinit. Should I do so directly? As far as I know telinit can tell upstart to reread it's configuration (telinit q) or restart (telinit u) Which while problematical (upstart doesn't handle this well) doesn't seem to be a security problem21:59
len-dttelinit 0 or 6 the user already can do (reboot or halt). that leaves telinit s or S which goes to single user mode.22:01
len-dthmm, single user as root I think, sounds bad, but in ubuntu anyone can get there with a reboot to recovery mode.22:03
len-dtoutside of ubuntu, if the user needs to login even in recovery mode, then this will be true in RL1 as well.22:07
len-dtAny holes in my thinking?22:07
ailosingle user as root sounds bad to me22:30
len-dtI think it is safer than running telinit from a sh script which filters the command line variables. I am not sure there are not combinations that might tell the script to do something nasty.22:31
len-dtailo, Ya, it sounds bad. Ubuntu's fault that, most distros require a password to get into a console in runlevel 1.22:32
len-dtThose that use sudo with no root password allow user login others require the root password.22:33
ailoBut even so, you don't log in as root for user tasks. Only to change the system or doing some monitoring you are unable to do as user22:36
ailoBut I guess you are just exhausting all options22:37
ailoI don't know much about run levels and init scripts, so all I can offer there is to help doing some testing later on, if it becomes a valid option to try22:38
len-dtWith ubuntu, if you have physical access... you are root!22:40
len-dtAll you have to do is reboot to rescue mode.22:40
len-dtailo, this is one thing I always thought was odd about ubuntu.22:41
len-dtI could never have done that with Slackware.22:41
ailoYou mean it's not very secure in that sense?22:41
ailoIf you have physical access, unless disks are encrypted, you can always put the disks into another machine22:42
len-dtAny system that someone has physical access to can be entered anyway. Boot from your own CD.22:42
len-dtYa, encryption would be the only way around that.22:43
ailoI wonder if encryption works if you try booting into rescue mode, not as the user? I think it does22:43
len-dtRL 1 is pretty secure even having root access. The network is shut down so no external shells are available. I think it shuts down serial access too.22:44
ailoWell, don't really know. Only that I've had problems with some software trying to use encrypted drives belonging to my own user22:44
ailoPretty radical thing to do, in order to get slightly better performance22:45
len-dtencrypted partitions still work because the partitions are only accessible with the key... root or not.22:46
len-dtHowever your point is valid. It is best to wrote a binary program that only takes input to change through RL 2 to 5.22:47
len-dtailo, I was just feeling lazy ;-)22:47
len-dtailo, it has been a long time since I last coded c ... I will probably have to look up each thing I do :P22:48
ailoWhat bugs me is that on probably most machines, you might only need one or two configs diffing from the -generic kernel in order to get full blown low latency performance (also having PAM privileges set up), but then you have this and that hardware configuration which does not work properly. 22:49
ailoIt's gotta be the fault of the kernel in the end22:50
ailoAnd all it contains22:50
ailoAs we talked about, drivers and what else22:50
len-dtailo, ya my netbook is a perfect case... wireless has to be removed and then... solid as a rock.22:51
len-dtailo, well till memory runs out.22:51
ailoThis is why I feel that the standard setup is the best default setup, where there may be a few configuration options that could never hurt and thus could also be added by default, but leaving the rest for trouble-shooting more or less22:53
len-dtailo, I would in some ways like to double my memory and kill swap. shnatsel says that is a bad idea. I guess swap is used for hibernation22:53
ailoAnd for those options that should be tweakable, because they either do not do anything useful for most people, or may even do the opposite, an application would be best22:54
ailohibernation.. now often does that work? I've had a problem with that on a lot of machines22:54
ailoBut should probably not be disablel :P22:55
shnatsellen-dt: http://rudd-o.com/linux-and-free-software/tales-from-responsivenessland-why-linux-feels-slow-and-how-to-fix-that22:55
ailodisabled*22:55
len-dtailo, RL2 should always be left default anyway. shnatsel Ya I read it.22:56
shnatsellen-dt: I think you need some filesystem caches anyway22:57
len-dtshnatsel, there seems (even in the comments for that article) to be some disagreement on his aproach.22:57
len-dtcaches in swap?22:57
shnatsellen-dt: yeah I agree it needs further research22:57
shnatsellen-dt: there are no caches in swap22:57
shnatsellen-dt: linux is not stupid22:58
len-dtI thought not.22:58
shnatsellen-dt: so either use zcache or swap to zram, and the latter is preferable because it's used only when you're really running out of ram if you've decreased swappiness22:58
len-dtshnatsel, with the use of zram, which I assume looks like swap to the machine, what happens then to hibernate?22:59
ailolen-dt: About those tweakable options. Since they probably don't do a lot to most people, "performance mode" does not seem appropriate to me. More like "troubleshooting mode". I wouldn't use that term, but just add the options to a audio-settings application, which might as well be a completely rewritten ubuntustudio-controls, and have that application easily accesible to the user23:00
shnatsellen-dt: I haven't tested that myself, but it's stored in RAM and it can't be swapped into itself, so the worst case is as much data as fits into zramswap being compressed and then swapped on the disk23:00
len-dtailo,  the name doesn't matter to me.23:00
shnatsellen-dt: which is in fact faster than usual hibernation because it basically means lzop-compressing data before writing it to disk23:01
shnatsellen-dt: I'm negotiating replacing lzop with lz4 btw23:01
shnatsellz4 is exactly the same except it uncompresses much faster23:02
len-dtshnatsel, I'm more worried about some of the data being mistakenly stored in zram instead of swap and there for vanishing when power goes.23:02
shnatsellen-dt: I believe those parts of memory are never swapped into itself23:03
shnatsellen-dt: otherwise it'd make no sense23:04
len-dtailo, quite definitely I (personally) want to be able to switch because I want to be able to use wireless when not doing audio too.23:04
len-dtshnatsel, ok, personally I never use hibernation anyways, I always shutdown.23:05
shnatsellen-dt: I've popularized swapping to zram with http://www.webupd8.org/2011/10/increased-performance-in-linux-with.html which triggered adding my scripts to repos of several distros. I've received several bug reports, but nobody complained about hibernation.23:05
shnatsellen-dt: neither do I23:05
len-dtSo my questions are more for if I am helping someone out.23:06
len-dtI do want to try zram though. shnatsel and I think I will (for personal interest more than anything) try running swap-less but adding more memory. It seems to me it is not too hard turning swap on and off... well it used to be easy ;-)23:08
shnatsellen-dt: sudo swapoff -a23:08
shnatsellen-dt: sudo swapoff23:08
shnatselone of them23:08
len-dtshnatsel, Ya, that command is still there. 23:09
len-dtshnatsel, would swapoff kill zram as well?23:11
shnatsellen-dt: yes23:11
shnatsellen-dt: and it's not easy to bring it back :P23:12
len-dtshnatsel, so I would want to do swapoff _special_23:12
len-dtThat is specify the partition.23:12
shnatsellen-dt: prediction: you'll get very annoyed by swapless setup because of out-of-memory kills with apps losing data.23:12
shnatsellen-dt: swapoff --help or man swapoff23:13
len-dtshnatsel, thats what I'm looking at.23:13
shnatsellen-dt: to get a taste of out-of-memory kills, press Ctrl+Alt+PrintScreen+F23:14
shnatsellen-dt: it triggers one :)23:14
len-dtshnatsel, I agree it would be bad for normal use, but If I don't have enough memory for my audio stuff to all fit in memory in the first place I think I would rather know that before I hit the record button.23:14
len-dtIn other words, swap off would be a specific use only 23:15
shnatsellen-dt: write a wrapper script checking the amout of memory available on app startup. I've got one for ensuring all my memory modules are in place and connected.23:15
shnatselit runs on boot though23:15
ailolen-dt: Yes, but the switch is only helpful to a few people, so it won't be interesting to most people to use. Anyway, implementation is step 223:18
ailoEnd of this day for me too. I'll be around tomorrow again23:19
len-dtailo, GN23:19
len-dtshnatsel, I'm going to play with zram... and maybe zcache.23:20
shnatsellen-dt: zcache is not easy to set up23:21
shnatsellen-dt: to get the code I mean23:22
len-dtOk, I'm looking through your blog just now... I'll get to that.23:22
len-dtHard it may be, but something of use to me I think.23:23
len-dtshnatsel, The first thing for me though, is 1Gig of ram is really too small. Doing a witch hunt to see what modules/daemons/apps I can kill to get more ram space is probably counter productive. I think US wants to be an "up to date" OS. To do that I think I need to update my machines. I need at least 2G ram, but will try to end up with 4G on my desktop which is the main audio machine anyway.23:36
len-dtI think zcache is a good idea as it reserves space for cache.23:37
shnatsellen-dt: I've been hitting swapping deaths with 2Gb, but with zramswap all is fine now23:37
shnatsellen-dt: and Firefox doesn't eat up gigabytes of memory anymore23:37
shnatselin fact it uses only 100mb for gmail while chrome uses 200mb23:38
len-dtHow well does disk read/write work without cache space?23:38
shnatseland I'm comfortable with 1Gb on my laptop too23:38
shnatsellen-dt: no idea23:38
len-dtshnatsel, so you would recommend setting up zram first.23:39
len-dtAnd let cache take care of itself.23:40
shnatsellen-dt: yes23:40
shnatsellen-dt: and I'm not very fond of zcache personally, though I haven't really looked into it23:40
len-dtshnatsel, is it worth using zram with the LZ1 alog. on an atom based machine?23:42
shnatsellen-dt: lzo, yes, I think so23:42
shnatsellen-dt: try and see23:42
len-dtOk, what split would I use for 1G23:42
shnatsellen-dt: just install my zramswap-enabler23:43
shnatsellen-dt: and see if it suits you23:43
len-dtshnatsel, is there anything there not compatible with 12.04?23:44
shnatsellen-dt: already resolved :)23:44
len-dtHmm, it depends on the generic kernel.23:48
shnatseloh, I haven't added that kernel flavour to dependencies, sorry23:49
shnatselwhat's the package name?23:49
len-dtzramswap-enabler23:49
shnatselno, for that kernel of yours23:49
len-dtOr do you mean owr kernel.23:49
len-dtI can't look just now... I'm DL the generic kernel...23:50
len-dtOur package is linux-lowlatency-pae23:55
len-dtI don't know what the amd64 package is though.23:55

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!