IRC log of #maemo-ssu for Tuesday, 2012-05-08

*** smoku has joined #maemo-ssu00:04
*** _rd has quit IRC01:10
*** LaoLang_cool has joined #maemo-ssu02:33
*** smoku has quit IRC02:47
*** smoku has joined #maemo-ssu02:47
*** gri has quit IRC03:19
*** gri has joined #maemo-ssu03:25
*** trbs has quit IRC03:26
*** LaoLang_cool has quit IRC03:30
*** smoku has left #maemo-ssu03:57
*** arcean has quit IRC04:15
*** amiconn has quit IRC05:54
*** amiconn_ has joined #maemo-ssu05:54
*** amiconn_ is now known as amiconn05:54
*** nox- has quit IRC06:47
*** Sc0rpius has quit IRC07:31
*** Sc0rpius has joined #maemo-ssu08:12
*** gri_ has joined #maemo-ssu08:15
*** lartza__ has joined #maemo-ssu08:16
*** lartza_ has quit IRC08:21
*** infobot has quit IRC08:21
*** wmarone_ has quit IRC08:21
*** andre__ has quit IRC08:21
*** ZogG has quit IRC08:21
*** IronLegend has quit IRC08:21
*** gri has quit IRC08:21
*** Raimu-X has quit IRC08:21
*** guly has quit IRC08:21
*** StyXman has quit IRC08:21
*** Raimu has quit IRC08:21
*** RST38h has quit IRC08:21
*** AndrewX192 has quit IRC08:21
*** claw has quit IRC08:21
*** ChanServ has quit IRC08:21
*** wmarone_ has joined #maemo-ssu08:22
*** andre__ has joined #maemo-ssu08:22
*** ZogG has joined #maemo-ssu08:22
*** Raimu has joined #maemo-ssu08:22
*** RST38h has joined #maemo-ssu08:22
*** IronLegend has joined #maemo-ssu08:22
*** AndrewX192 has joined #maemo-ssu08:22
*** claw has joined #maemo-ssu08:22
*** Raimu-X has joined #maemo-ssu08:22
*** guly has joined #maemo-ssu08:22
*** StyXman has joined #maemo-ssu08:22
*** ChanServ has joined #maemo-ssu08:22
*** barjavel.freenode.net sets mode: +o ChanServ08:22
*** lartza__ has quit IRC08:23
*** lartza_ has joined #maemo-ssu08:23
*** pabs3 has joined #maemo-ssu10:13
*** pabs3 has left #maemo-ssu10:13
*** dafox has joined #maemo-ssu10:16
*** Pali has joined #maemo-ssu10:38
*** BCMM has joined #maemo-ssu11:25
*** arcean has joined #maemo-ssu12:14
*** BCMM has quit IRC12:41
*** RST38h_ has joined #maemo-ssu12:42
*** wmarone_ has quit IRC13:06
*** RST38h_ has quit IRC13:17
*** RST38h_ has joined #maemo-ssu13:23
*** RST38h_ has quit IRC13:31
*** wmarone_ has joined #maemo-ssu13:47
*** psycho_oreos has joined #maemo-ssu13:57
*** RST38h_ has joined #maemo-ssu15:09
merlin1991X-Fade: ping15:23
DocScrutinizermerlin1991: seems he's listening on #harmattan15:27
DocScrutinizeror at least did, 20 minutes ago15:28
*** DocScrutinizer has quit IRC15:56
*** DocScrutinizer has joined #maemo-ssu15:56
*** RST38h_ has quit IRC16:03
merlin1991probably went for lunchbreak16:09
merlin1991or other stuff :D16:09
*** DocScrutinizer has quit IRC16:22
*** DocScrutinizer has joined #maemo-ssu16:22
*** DocScrutinizer has quit IRC16:30
*** DocScrutinizer has joined #maemo-ssu16:30
*** DocScrutinizer has quit IRC16:33
*** DocScrutinizer has joined #maemo-ssu16:33
merlin1991DocScrutinizer, MohammadAG name for the rewrites component?16:34
DocScrutinizermerlin1991: name for repo?16:40
merlin1991nope component inside cssu repo16:40
DocScrutinizercssu-optional16:40
merlin1991it went down as "replace"16:40
DocScrutinizerwell, more specific16:41
merlin1991since we have free and non-free16:41
DocScrutinizerwhere would things like orientation-lock reside?16:42
merlin1991free as an opt-in16:42
merlin1991free is for new and already oss packages16:42
merlin1991non-free is for everything nokia is kind enough to hand over (certman applet)16:42
merlin1991and replace is for everything that replaces former non-free stuff16:43
merlin1991and we'll introduce a new metapackage, yay :/16:43
*** kent_autistic has joined #maemo-ssu16:43
DocScrutinizeraaah16:44
DocScrutinizer:-)16:44
merlin1991I went for the new component way since with an all new repo we'd have a hard time ensuing cssu is installed aswell (since we do depend on cssu for that optional stuff)16:45
merlin1991s/ensuing/ensuring/16:46
DocScrutinizer:nod:16:46
DocScrutinizerI think we should think a few days about it, discuss it16:46
merlin1991X-Fade: is putting the server side into place as we speak, how we end up using it we can still discuss :)16:47
DocScrutinizerthe purpose however should be clear: have a "monolitic" CSSU install that only comes with mandatory stuff, and have a way to allow user to select optional packages in a finegrained though convenient way16:47
merlin1991yep16:48
DocScrutinizerfor optional packages we need to distinguish a few cases: stuff that replaces maemo-mp components, stuff that depends on cssu, ... <other>?16:49
merlin1991yeah, I suggest we do a proper meeting next week?16:49
DocScrutinizerfor each case we have to consider dependencies carefully16:50
DocScrutinizerI'm "free" next week, so why not16:50
merlin1991"free" ? :D16:50
DocScrutinizerholiday16:50
merlin1991ah :)16:50
DocScrutinizeranother awesome 1.5 weeks16:51
merlin1991Pali, arcean, MohammadAG: got time next week monday @16 UTC (18 CET) ?16:51
DocScrutinizermerlin1991: I'd like to invite jaffa and a few others to discuss the implications with 'us'16:54
merlin1991sure16:54
DocScrutinizerX-Fade as well, of course16:54
DocScrutinizerthere might be pitfalls at least me doesn't realize16:55
DocScrutinizerthere's nothing as annoying as a pitfall you fall in because you haven't anticipated it16:56
X-FadeWell, things might break. But it is hard to predict :)16:56
DocScrutinizerhi X-Fade :-D16:57
*** psycho_oreos has quit IRC16:57
merlin1991things can't break even more than the ke-recv disaster :D16:57
DocScrutinizerwhat's you system-architect take on the topic?16:57
DocScrutinizermerlin1991: well, we might see urge to do a full rollback if anything in our design doesn't hold16:58
MohammadAGmerlin1991, yes16:58
X-FadeWell, that would involve me deleting some files from the tree and re-indexing the repo.16:59
X-FadeNot that big of a deal.16:59
DocScrutinizernot exactly a big deal, as long as this happens for T17:00
DocScrutinizerfor S however it was a mayor PITA, regarding user experience17:00
X-FadeYou could push a different meta, just for testing this.17:00
DocScrutinizersimply means early pitfalls aren't as deep and hurting as late ones17:01
merlin1991no worries I'd leave quite some testing period before this comes anywhere near stable17:01
DocScrutinizermerlin1991: sure thing :-)17:01
X-FadeHehe, and that too.17:02
DocScrutinizerI never really wrapped my head around that components:free/non-free thingie17:02
merlin1991it's mainly a policy thing17:03
merlin1991you could slap everything into 1 component and still run the same project17:03
DocScrutinizerbut how does it impact the way e.g. HAM works?17:03
merlin1991apt doesn't care since we are still in the same repo, only ham might get ideas which is why we have to livetest it :)17:04
*** RST38h_ has joined #maemo-ssu17:04
merlin1991as far as I see it we have to decide on a policy how we handle replacements and opt-ins from now on and then just try it with the next update17:04
DocScrutinizermy vision is along a new "category" besides "all" and "games" and "$foo", in HAM17:04
DocScrutinizerprobably this idea won't fly though17:05
merlin1991would require as slight update to ham17:05
merlin1991nothing major though17:05
merlin1991btw if you go down that path I'd suggest 2 categories for ham17:05
merlin1991one for "cssu opt ins" and one for "foss replacements"17:06
DocScrutinizersince the new stuff only applies on CSSU systems and those systems are supposed to have patched HAM anyway, it might work17:06
merlin1991sure patched ham would be part of the core mp17:06
DocScrutinizersure, 1 new cat, 2, whatever17:06
kent_autisticsince it already involves HAM, i hope something is being done about the slowness of HAM.17:07
kent_autisticand forget FAM17:07
DocScrutinizerwell, fapman et all is indeed a problem when opting for this path17:08
merlin1991kent_autistic: Pali did stuff to ham, though last time I tested his patches it didn't work properly, he made some fixes in the meantime so I'll reevaluate his code gain with the coming cssu udpate17:08
merlin1991DocScrutinizer: I don't see why17:08
kent_autisticnice to know thanks!17:08
DocScrutinizerNFC about how fapman works, but we need to at least consider *if* it would17:08
merlin1991apt would still work and fapman calls apt17:09
merlin1991so it is going to still work17:09
DocScrutinizerhmm, ok17:09
DocScrutinizerif fapman doesn't care about categories at all, then fine17:09
DocScrutinizer"fine"17:09
DocScrutinizer:-P17:09
merlin1991well fapman might not have the category visible in the ui17:10
merlin1991but why would we care?17:10
DocScrutinizer:shrug:17:10
* DocScrutinizer for sure won't17:10
merlin1991since we "know" that merely using fapman is bound to fubar your system if you got cssu17:10
DocScrutinizeras long as we are clear about it17:10
DocScrutinizerseems I'm once more guilty to have started something17:11
DocScrutinizerand then leaving for RL ;-D17:11
* DocScrutinizer waves17:12
merlin1991:D17:12
merlin1991hf17:12
DocScrutinizerthanks17:12
Raimu:)17:12
DocScrutinizermerlin1991: please don't forget to properly announce meeting17:12
merlin1991currently trying to find out who's got time17:13
merlin1991so far only you and mag replied :)17:13
DocScrutinizerand X-Fade17:13
DocScrutinizerI'd also appreciate javispedro joining17:14
merlin1991phew I remembered my pin17:15
X-FadeOk, I think I touched every script for it to work with -testing.17:15
DocScrutinizerbest practice to announce such stuff by introducing the plan in large picture and asking participation/comments of whom-it-may-concern17:15
merlin1991I'll post to maemo-devs17:16
DocScrutinizergreat17:16
DocScrutinizero/17:16
merlin1991but for now I'll write some emails to find out if people even have time :D17:17
X-FadeSo ping me when you pushed something and things break :)17:17
merlin1991X-Fade: currently we target to have a meeting about this first next monday so first thing pushed probably happening at some point next week :)17:17
X-Fademerlin1991: no problems, we'll see what happens.17:18
*** kent_autistic has quit IRC17:24
merlin1991X-Fade: still here?17:26
X-Fademerlin1991: Sure17:27
merlin1991http://maemo.org/packages/view/libxau6/ is still a problem in the repo17:27
*** freemangordon has joined #maemo-ssu17:27
merlin1991(extras-devel)17:28
X-Fademerlin1991: Yeah, I guess a big cleanup is in order. It is just a massive pain..17:28
MohammadAGit's more than that, it reboot loops systems17:28
freemangordonmerlin1991, MohammadAG, I think we finally have ROCK STABLE thumb2 on n900, what do you think about introducing that in CSSU?17:30
merlin1991oh hey freemangordon :D17:30
freemangordon:D17:30
merlin1991read your mail ;)17:31
MohammadAGdepends, rocks shatter with rain :P17:31
freemangordonmerlin1991, yeah :D17:31
merlin1991perfect so I only need a yes from arcean and Pali and we're good to go :)17:32
freemangordonwell, my question re thumb was "in general", i.e. should i put more time in that or is it a no go17:32
freemangordonjust an initial statement :)17:33
freemangordonPali, BTW how is kubuntu (besides being slow as hell)17:33
MohammadAGif it doesn't break ABI, sure17:33
freemangordonbreak ABI? I am using it on my primary device17:34
merlin1991about thumb2, whilst I see the advantages of it I'm not so sure if cssu is really the right target for it, for me it feels more like a thing for seperate repo17:34
freemangordonthe only phone I have17:34
MohammadAGfreemangordon, kubuntu's good17:34
MohammadAGbut needs good resources17:34
freemangordonMohammadAG, 12.04 is hardfp thumb compiled ;)17:34
merlin1991freemangordon: starting with the fact that you need a different gcc inside scratchbox17:34
freemangordonand before latest thumb2 fixes in uboot was SIGILLing every now and then17:35
freemangordonmerlin1991, that is why CSSU is the right place AIUI17:35
merlin1991freemangordon: does it work only by flashing a different kernel?17:35
merlin1991s/only by/by only/17:36
freemangordonyep, we need thumb errata workaround17:36
merlin1991meaning without uboot?17:36
freemangordonyep17:36
freemangordonthere is still no such kernel, nut it is easy to be done.17:36
freemangordon*but17:36
freemangordonit was easyer for me to patch u-boot17:37
merlin1991if we really go through with that we'd need to provide scratchbox tool packages aswell17:37
freemangordonbut the same could be implemented in kernel, it is just a matter of SMC call17:37
merlin1991since anybody should be able to grap cssu sources compile them and run17:37
freemangordonmerlin1991, gcc4.6.2 ready for SB is on your server :P17:37
merlin1991which isn't the fact anymore if you can't do it with the default maemo scratchbox17:37
freemangordon^^^17:38
freemangordonBTW it *might* be possible that we don't need 4.6.217:38
DocScrutinizergeneral procedure to "upstream" kernel patches: spot a problem, understand the problem, REPRODUCE the problem, find a fix, PROVE fix actually id effective on previously demonstrated testscase that reproduces the problem. Mail to LKML with patches, get REVIEW, wait for upstream to include it to master branch, *then* include it to local branch17:38
freemangordonDocScrutinizer, I have only one life to live17:39
DocScrutinizerthumb is not only a kernel patch, it's a system wide issue17:39
freemangordonand the issue has beer reported and fixed back in 200917:39
freemangordon*been17:39
freemangordonwe are not discovering the hot water here17:39
freemangordonthe fix is not mine, but from TI17:40
DocScrutinizerso where's the problem to simply pull the 2009 patches into our local branch?17:40
merlin1991freemangordon: I do appreciate the effort you put into this, but I'm inclined to think that it is somewhat out of the scope of cssu17:40
freemangordonmerlin1991, ok17:40
DocScrutinizerI agree with merlin1991 here17:41
merlin1991btw MohammadAG what's your take on the whole issue?17:42
freemangordonmerlin1991, again, take your time with MohammadAG to discuss it, if you need some info just ask me.17:42
merlin1991freemangordon: is there anything in maemo thumb2 atm?17:43
freemangordonyep17:43
merlin1991besides the stuff on your device ;)17:43
freemangordonx-loader and nolo17:43
merlin1991but nothing in userspace or other parts?17:44
freemangordonno way17:44
freemangordonyou remember modest, ain't?17:44
freemangordon:D17:44
merlin1991hehe17:44
merlin1991basically my reasoning against this for cssu is, that it's too big of an unkonw / too big of a change to the system to be considered a simple "update"17:45
freemangordonmerlin1991, it does not work like that17:45
freemangordonimplementing the patch just enables thumb2 compiled binaries to work without crashes17:46
freemangordonit does not turn automagically all binaries to thumb-compiled :D17:47
merlin1991I know17:47
freemangordonso it could be made st6ep by step17:47
merlin1991but if you ship the patch I'm sure you want to ship loads of thumb compiled binaries aswell :D17:47
freemangordonno17:47
freemangordonslowly, one after another, first being Qt, as no critical part of the system depends on it17:48
DocScrutinizerso what's the benefit then?17:48
freemangordonand it is HUGE17:48
freemangordon20-30% less RAM usage for Qt apps17:48
freemangordonand Qt could be moved to NAND(NOR)17:48
merlin1991freemangordon: the other problem is, that we'd essentially break everything for anybody using a custom kernel17:49
freemangordongiving additional boost to load times17:49
freemangordonmerlin1991, what kernel do you think of?17:49
DocScrutinizerlook, we need to take responsibility for each single binary we ship. So if you ship a new thumb-enabled Qt, you basically find yourself as co-maintainer of every single Qt-based app on the system, doing triaging on bugtracker, trying to tule out it's your thumb thing that caused that problem17:49
merlin1991for example I ran power46 for the injection drivers quite long17:50
freemangordonDocScrutinizer, and what is different with the situation as it is now?17:50
merlin1991there is the btrfs kernel17:50
merlin1991and other17:50
merlin1991+üs17:50
freemangordonmerlin1991, the only kernel that is still maintained is KP17:50
merlin1991the problem I see with it, that we can't ship it as a single update17:50
merlin1991freemangordon: does not mean it is the only one used17:50
freemangordonmerlin1991, 20 lines patch could be easyly backported to any custom kernel we might have17:51
DocScrutinizerwon't happen17:52
merlin1991the point is cssu should work for *everyone*, I'm sure we could get another repo in place as "cssu-thumb2" and I'd have no problem adding it there17:52
DocScrutinizerfull ack17:52
freemangordonthat is a fork17:52
merlin1991ye17:53
DocScrutinizerand maybe even eventually port that 20liner patch to anything cssu-core, IF we ever would dare to link CSSU to a kernel version17:53
merlin1991normally you fork stuff when you can't achieve what you want within the bounds of the original project ;)17:53
MohammadAGquestion17:53
merlin1991...17:53
MohammadAGif we simply add the patch to all kernels17:53
MohammadAGand include omap1 patched kernel in CSSU17:54
MohammadAGwouldn't that cover all cases?17:54
freemangordonyep17:54
DocScrutinizerno17:54
freemangordonyes17:54
merlin1991MohammadAG: anybody with his own kernel flashed is still fubar17:54
DocScrutinizeryep17:54
merlin1991that's why I'd say have a fork of cssu that is thumb217:55
DocScrutinizerit's honestly WAAY beyond the scope of CSSU anyway17:55
MohammadAGsigh17:55
freemangordonguys, all kernels we have are omap1 based, and if you are good enough to build and flash your own kernel, should be able to backport that patch17:55
freemangordonand who we are talking about 5-10 geeks?17:55
DocScrutinizerthat's getting quantitative argumentation now17:56
DocScrutinizerCSSU is supposed to be compatible17:56
freemangordonDocScrutinizer, was CSSU compatible when Qt 4.7.4 was brought in?17:57
DocScrutinizerI dunno, but if it wasn't it was a major mistake anyway17:57
freemangordonor it broke a couple of applications and custom setups?17:57
MohammadAGit's like telling Nokia not to have a newer kernel in an update cause custom ones will still be old17:57
DocScrutinizernot the first mistake in CSSU17:57
freemangordonMohammadAG, agree17:57
merlin1991freemangordon: Qt 4.7.4 broke stuff because if was borken itself17:58
merlin1991when we fixed that (optimisation flag) everything was compatible17:58
MohammadAGbasically, if omap1 is shipped, it'll overwrite any custom kernel in place17:58
freemangordonmerlin1991, after that17:58
DocScrutinizerMohammadAG: nope, as Nokia wouldn't come up with an incompatible kernel17:58
MohammadAGwe can ship both omap1, and kp can include the patch upstream17:58
MohammadAGDocScrutinizer, this doesn't break anything the way I understand it17:58
DocScrutinizerit does17:59
DocScrutinizerfor all those that have any custom built kernel17:59
MohammadAGshipping only the kernel doesn't17:59
merlin1991btw what's the big deal with doing a fork for thumb2?17:59
DocScrutinizerand honestly why would we want to link CSSU to a new kernel?17:59
MohammadAGtoo many forks17:59
MohammadAGcause this patch goes to waste without it17:59
DocScrutinizernonsense18:00
freemangordonyeah, I don't want to be the only developer and maintainer of such a fork. Not that I like the idea of forking18:00
MohammadAGif kp gets the patch, the binaries won't be shipped in extras18:00
DocScrutinizeryou can port this patch to PK any time you want18:00
MohammadAGand you can port this patch to your custom kernel too18:00
DocScrutinizerbut WHY would we link it to CSSU?18:00
merlin1991I mean hey, I'd probably swap over to the thumb2 fork and propose to handle the stable branch aswell, but I just see it conflicting with the mission statement of the original cssu18:01
DocScrutinizeryes18:01
freemangordonmerlin1991, what is the idea all of us to fork to a new CSSU-thumb leaving non-CSSU thumb to stagnate?18:01
merlin1991freemangordon: why would non-thumb stagnate18:02
freemangordonwhat will change by changing the name?18:02
merlin1991it would be like debian x86 and x6418:02
merlin1991essentially just differntly compiled18:02
merlin1991only that apt does not differ between "armel thumb" "armel non thumb"18:02
DocScrutinizerfreemangordon: what's the alternative?18:02
merlin1991so we have to apply that on a different level18:02
merlin1991freemangordon: I'd just keep cssu non-thumb in sync with everything not thumb related18:03
freemangordonBTW soon or later we need a kernel in CSSU if we want a bugfixes on kernel level18:03
DocScrutinizersure18:03
freemangordonand how is that different the?18:04
freemangordon*then18:04
DocScrutinizerand we had to consider *carefully* if we really want to do that18:04
DocScrutinizerfor above elaborated reasoning18:04
freemangordonwell, there are bug reports against omap1 kernel, should we mark them as WONFIX because someone could have flashed his own kernel?18:05
freemangordonsigh18:05
merlin1991freemangordon: those are 2 disticnt discussions :D18:05
DocScrutinizerkernel bugs get fixed in kernel. AIUI PK is all about that18:05
freemangordonKP is not omap1 bugfixing cousing, if you don't beleive me, ask Pali18:06
merlin1991freemangordon: I don't see why you're so against branching off a cssu with thumb that keeps in sync for main cssu and vice versa18:06
freemangordonmerlin1991, I don't want forks, that is why. Otherwise I am ok, if you become a maintainer18:07
DocScrutinizerouch18:07
freemangordonbtw apt-get problem could be solved by using epoch?18:07
merlin1991freemangordon: I meant that with apt a lil different18:07
merlin1991for me using thumb or not is kinda like running x86 or x6418:08
merlin1991it's something the user decides18:08
freemangordonyet another reason - I am just not good enough with debian packaging to become a maintainer of such a fork18:08
DocScrutinizerthen you evidently don't have the manpower to pull such project18:09
merlin1991freemangordon: I'd happly do that for you18:09
freemangordonnever pretended I have, that is why I am here18:09
DocScrutinizeras I explained above, you'll need to maintain means triage any bug against your thumb version18:09
freemangordonDocScrutinizer, I think I am good enough on bug-fixing and problem discovering, ain't?18:10
DocScrutinizerfor $random-app of extras?18:10
freemangordonsure18:11
merlin1991DocScrutinizer: your argument is moot anyway18:11
merlin1991anyhting that breaks due to thumb is surely inside freemangordons scope of interest18:11
merlin1991and everything else is inside the scope of "main" cssu18:11
DocScrutinizeryeah, but you need to triage18:11
merlin1991since I guess freemangordon does not plan to move anyhting outside main cssu into his thumb project18:12
freemangordongood gues18:12
DocScrutinizerQt?18:12
freemangordon*guess18:12
merlin1991DocScrutinizer: again, if it's qts fault then --> main cssu, if it's thumbs fault --> freemangordon18:12
freemangordonQt (as everything else) will be just thum-compiled18:12
DocScrutinizerfunction calls?18:12
freemangordoninterworking?18:12
DocScrutinizer[2012-05-08 16:49:56] <DocScrutinizer> look, we need to take responsibility for each single binary we ship. So if you ship a new thumb-enabled Qt, you basically find yourself as co-maintainer of every single Qt-based app on the system, doing triaging on bugtracker, trying to rule out it's your thumb thing that caused that problem18:13
freemangordon[17:48] <freemangordon> DocScrutinizer, and what is different with the situation as it is now?18:14
freemangordonthat is not productive18:14
merlin1991I think we can stop here and say that freemangordon is more than willing to triage anything related to thumb18:14
freemangordonas is with anything non-related18:15
merlin1991and that we also have a strong concensus against thumb into "main" cssu18:15
DocScrutinizerthe triaging is way more complex, compared to a proper bugfix to Qt that maybe ideally even gone upstream and reviewed18:15
merlin1991complexity is not an argument against it though18:15
DocScrutinizernot an argument against own fork, sure18:16
freemangordonmerlin1991, if you and MohammadAG agree on "forking" then someone should do testing-thumb and testing-stable branches on gitorious18:16
freemangordons/testing-stable/stable-thumg/18:17
DocScrutinizer\o/ we once more found a common sense18:17
DocScrutinizer:-D18:17
merlin1991freemangordon: even you could do that18:17
merlin1991since branches are only realted to being able to push ;)18:18
freemangordonmerlin1991, ok, will try :)18:18
* DocScrutinizer tries to finally leave for RL now18:18
merlin1991freemangordon: as a starter I could set you up with another repo on my server18:19
freemangordonyeah. I know you like that :p18:19
merlin1991lol18:19
merlin1991to release this to the general public we'd need that "do I want to flash the kernel" thingy18:20
merlin1991and a patched omap118:20
merlin1991though ofc the do I really want to flash package in your case only checks against compatible kernels and flashes otherwise18:21
freemangordonmerlin1991, I don't think so, after all you should manually add the repo, so you will have that info before installation18:21
merlin1991freemangordon:  you manually add cssu aswell and we still argue about how to ship kernels :D18:21
freemangordonand as the kenel there will be in fact KP there is no point to ask such questions18:22
DocScrutinizeryup18:22
DocScrutinizerjust create a dependency in your repo core apps to KP18:22
freemangordonno, kernel-feature-thumb-errata ;)18:23
DocScrutinizerKP on isntallation is asking this chitchat anyway18:23
freemangordonyes18:23
DocScrutinizerfreemangordon: ack18:23
merlin1991basically any thumb app has to have that depends18:24
DocScrutinizerbasically you'd need that dependeny in all your thumb-compiled pkgs18:24
freemangordonmerlin1991, agree18:24
DocScrutinizerhehe, merlin1991 been faster than me18:25
freemangordonbut of course, he is younger :P18:25
DocScrutinizer*sigh*18:25
DocScrutinizerindeed18:25
* merlin1991 sometimes wonders how old doc really is18:25
freemangordon40?18:25
merlin1991anyway, kernel-feature-thumb-errata sounds good18:26
freemangordonyeah. Now I just have to convince Pali to fork his own kernel :D18:26
DocScrutinizererr, is that really needed?18:27
merlin1991freemangordon: why?18:27
merlin1991doesn't the normal system still work with the errata enabled?18:27
DocScrutinizerAIUI the patch in kernel is compatible?18:27
freemangordonno, I was joking18:27
merlin1991fu :D18:27
freemangordonyeah, normal system could be a little bit slower, but otherwise it is compatible18:28
freemangordonafter all the patch clears BTB on every context switch18:28
freemangordonthumb does not come without cons18:29
DocScrutinizereeew18:29
DocScrutinizerare there numbers on overhead in uSeconds for one context switch, and an idea how many context switxhes 'usually' happen during normal usage?18:31
freemangordonAFAIK CONFIG_HZ is 12818:31
freemangordonso that is the absolute maximum18:31
DocScrutinizersure?18:31
DocScrutinizeraren't here possible contect switches on calling kernel functions as well?18:32
freemangordonbut I am not aware of any hard numbers. But the rumors stskeeps spreads that the system becomes very very slow are not true18:32
* DocScrutinizer scratches head18:33
freemangordonDocScrutinizer, I think that is what scheduler uses. Could be wring of course18:33
freemangordon*wrong18:33
DocScrutinizerthat's the maximum stay-in-focus time for a particular context, until scheduler will preempt18:34
DocScrutinizerthere *could* be way more context switches / second though, if task itself is causing the context switch, buzzword function-call18:35
freemangordonDocScrutinizer, I am not knowledgeable enough, sorry, could only share my end-user experience. and it is that multitasking is better, applications start faster, and the system slowdown is not irritating.18:36
DocScrutinizerfair enough18:36
freemangordonBTW my swap right now (after more than 24hours of usage) with 4 im account online and BT handsfree is 106 MB18:37
freemangordonwould you share yours?18:38
DocScrutinizerthere however could be particular usecases where this actually hurts a lot - e.g. media playback using the DSP. AIUI each call to DSP functions will cause a context switch. On video playback this might show, on 'normal' system usage you'll never encounter it. Just as an example of a possible scenario18:38
freemangordonDocScrutinizer, no way18:39
freemangordontrust me on that, I spend good enough time playing with DSP18:39
DocScrutinizerit was a (somewhat) made up example18:40
freemangordonI don't argue there is no slowdown, but (in theory) it should be overcompensted by better code density18:41
freemangordonand less memory usage18:41
DocScrutinizerjust explaining why it's not exactly easy to estimate the impact18:41
freemangordonagree18:41
freemangordonanyway, the major slowdown factor now is swapping, agree?18:41
DocScrutinizeryes18:41
DocScrutinizerusually18:41
freemangordonso, you can imagine the impact of having 30-40 MB more free RAM18:42
freemangordonand even if you have to discard the executable page, by the time you will need it again, you will have to read 30-40 % less data from the disk18:43
freemangordon(nand, flash, whatever)18:43
freemangordonsounds good (at least in theory) :)18:44
Palihi18:45
freemangordonhi18:45
Palifirst, merlin1991 monday I can be online after 18:00 UTC18:46
merlin1991damn18:47
freemangordonmerlin1991, BTW isn't 16 UTC == 19 CET?18:47
freemangordonsummer time?18:48
freemangordonnah, scratch that18:48
freemangordonit is 19 here18:48
freemangordon:)18:48
merlin1991yep :D18:48
Palifreemangordon, kubuntu seems be stable, no craches but slow and some problems with touch input (but I did not used special installer, only prepaired image)18:48
freemangordonPali, it is not slow, it crawls :D. I installed Ubuntu instead of it, it is much faster. No crashes here too.18:49
merlin1991freemangordon: dunno kubuntu is fast for me18:49
freemangordonif you wish, I can give you xorg.conf, which fixes touch screen ad the keyboard maping18:50
Paliso this is problem with KDE, akonadi, nepomuk, etc kde ...18:50
freemangordonmerlin1991, which version?18:50
merlin1991latest18:50
freemangordon12.04 on n900?18:50
merlin1991though I have disabled akonadi18:50
Palimaybe I do not have 3d drivers18:50
PaliI used prepaired image (maybe for beagleboard)18:51
freemangordonmerlin1991, we are talking about k/ubuntu 12.04 on n90018:51
merlin1991oops :$18:51
Palione question, will be obs also for personal repos? or only for central extras?18:51
freemangordonPali, yes, it is for BB, but runs on n900. Though extremely slow. Ubuntu 12.04 is much much better, though I hate unity desktop18:52
freemangordonand the fact it seems there is no lxde :(18:52
merlin1991Pali: as far as I understood it the maemo obs should have personal repos aswell18:52
Paliok18:53
freemangordonPali, did you read through backscroll?18:53
Paliyes, but there are a lot of lines18:53
Palibut, stable thumb in cssu seems good18:53
freemangordonwell, it should be enough if you got the general idea18:54
freemangordonand it is that merlin1991 and DocScrutinizer don't want thumb in CSSU :D18:54
freemangordoninstead we will do internal "fork"18:54
freemangordona repo wich will mirror CSSU state with thumb-compilde binaries18:55
freemangordonand will have KP (under a different name) too18:55
merlin1991freemangordon: why do you want to put kp in there actually?18:56
merlin1991why not simply have kp with the errata enabled in extras?18:56
* DocScrutinizer hands merlin1991 a cookie18:56
freemangordonbecause it does not make sense CSSU (even thumb-forked one) to depend on application in extras. not to mention the problems we have with repos on maemo.org18:57
MohammadAGcross repo depends...18:58
freemangordoni.e. right now Pali cannot promote KP in extras, because DB is crewed as usual18:58
freemangordon*screwed18:58
DocScrutinizerI thought the dependency was kernel-feature-thumb-errata, and no matter where from to pull the pkg providing this property19:00
freemangordonexactly19:00
freemangordonbut we should provide the "default" kernel19:00
merlin1991then I'd put patched omap into the "cssu" repo and have a patched kp in extras19:01
DocScrutinizerI'm all happy with 'default' KP not including this patch, as you yourself elaborated it can slow down 'normal' systems19:01
freemangordonbecause otherwise we force users to enable -testing or -devel repos enable. merlin1991, why we need omap1 kernel?19:01
merlin1991because we also like those users who still use omap1 and don't want kp, I dunno19:03
DocScrutinizerfreemangordon: I know it's difficult to *set* this "flush cache please" bit, but could it get *cleared*?19:04
freemangordonkp is better, it has bugfixes against CSSU, stuff like that19:04
freemangordonDocScrutinizer, it is equally hard19:04
DocScrutinizer:-/19:04
freemangordoni.e. you should issue a call to Secure Monitor with the new contents of Aux Control reg. But if I understand your question, it is NP to be doone through sysfs entry19:05
DocScrutinizercould we switch the bit during uptime, at all? I.E. could KP learn to enable/disable thumb-support (and thus remove negative impact of thumb patch for 'normal' systems)?19:06
DocScrutinizeryeah19:06
DocScrutinizer:-D19:06
freemangordonDid I get it correctly, you want that bit enabled by default, and userspace tool to clear it19:07
freemangordonyeah :D19:07
*** dafox has quit IRC19:07
freemangordonthat could be done19:07
DocScrutinizergreat19:07
DocScrutinizerI'd say this should go into 'standard'19:07
DocScrutinizerKP19:07
freemangordonyou'd better tell that to Pali :D19:08
DocScrutinizerI'm just stating my notion, not telling anybody19:08
DocScrutinizerfreemangordon: btw this would be a great way to actually do the proof-of-effectiveness of that patch19:09
freemangordonDocScrutinizer, the best way is k/ubuntu, it is thumb compiled all the way19:10
freemangordonand it crashes every now and then without that fix19:10
freemangordonSIGILL every 20-30 seconds19:11
DocScrutinizeryou should be able to stop modest from segfaulting (in thumb version), and also see some slight (or huge) slowness when running a benchmark on a ARM binary19:11
freemangordonDocScrutinizer, the whole OS is a better testthan just modest19:12
DocScrutinizermaybe, but I prefer clear small well defined testbeds19:12
DocScrutinizer:-)19:12
*** RST38h_ has quit IRC19:13
DocScrutinizerand after all it's not hard to install one thumb binary (modest or whatever) on top of KP and then run that test for an hour with thumb-patch enabled, then disable the patch and watch it segfault19:14
DocScrutinizersimilarly to check impact on performance for normal apps19:15
freemangordonDocScrutinizer, it is no different, the bug is triggered by chance, AIUI only TI could actually do a real test case by playing with only God knows which debug registers in OMAP. But by having all user binaries thumb-compiled the chance is bigger. I told you, ubuntu 12.04 SIGILLs every 2030 seconds without the patch19:15
*** NIN101 has joined #maemo-ssu19:15
freemangordon20-3019:16
freemangordonAnd that could be confirmed by anyone who've tried it on n90019:16
DocScrutinizerwell, I'm not going to install a complete new system, to verify the patch. Too much overhead19:17
freemangordonI did it :)19:17
freemangordonSo did Pali19:17
DocScrutinizeryou should try to get the patch upstream, and for that you'll need a better (smaller more convenient to use) testbed to allow upstream maintainer to do that test him/herself19:18
freemangordonDocScrutinizer, no way this to go upstream AIUI. It relies in SM call, which is specific for every device AFAIK. Think of it as calling an undocummented BIOS function.19:20
DocScrutinizeranyway, the performance test is quite straightforward without any need for a single thumb binary19:20
freemangordonAnd TBH I won't go through all that fuss just to see my name in kernel.org contributors list(or whatever it will appear) :P19:21
DocScrutinizerso such sysfs (or proc?) node is for sure more useful than e.g. smartreflex19:21
freemangordonyeah19:21
freemangordonwhy is that? ever tried SR in KP50?19:22
DocScrutinizerno, just used a well known example19:22
freemangordonwell, try it, you'll be surprised. Seriously19:23
DocScrutinizerI'd even dare to suggest the thumb patch should be off by default in KP, and you include a line to system-init to enable it, on cssu-thumb19:23
freemangordonDocScrutinizer, sounds sane19:24
DocScrutinizeryour thumb pkgs could even depend on that mini pkg that inserts that line enabling thumb patch19:24
DocScrutinizerwhich in turn would depend on kernel-feature-thumb-errata19:25
freemangordonyeah19:25
freemangordonok, time to say bb19:26
DocScrutinizero/19:26
* freemangordon is afk19:26
*** Pali has quit IRC19:43
*** Pali has joined #maemo-ssu19:47
*** trbs has joined #maemo-ssu20:52
*** NIN101 has quit IRC20:56
*** NIN101 has joined #maemo-ssu21:50
*** _rd has joined #maemo-ssu21:56
*** _rd has quit IRC22:28
*** Pali has quit IRC22:40
*** Pali has joined #maemo-ssu22:41
*** Pali has quit IRC22:48
*** arcean has quit IRC23:10
*** arcean has joined #maemo-ssu23:14
*** _rd has joined #maemo-ssu23:32
*** ekze-nyan has joined #maemo-ssu23:54
*** ekze has quit IRC23:54
*** nox- has joined #maemo-ssu23:58

Generated by irclog2html.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!