IRC logs for #baserock for Friday, 2015-09-11

*** Walkerdine has joined #baserock00:23
Walkerdineannnnnd my build failed00:23
*** zoli___ has joined #baserock00:29
*** zoli__ has quit IRC00:29
WalkerdineGot the error "g++: internal compiler error: Killed (program cc1plus)" So I guess the jetson ran out of ram00:30
*** zoli__ has joined #baserock00:46
*** zoli___ has quit IRC00:46
WalkerdineBuilding again to see if building the artifacts had anything to do with it00:54
*** zoli__ has quit IRC01:19
WalkerdineHmm it still failed. I keep running out of RAM01:57
*** zoli__ has joined #baserock02:19
*** zoli__ has quit IRC02:23
*** zoli__ has joined #baserock02:24
*** zoli__ has quit IRC02:27
*** zoli___ has joined #baserock02:27
WalkerdineI added swap space on an SD card I'll see how that works out02:53
*** zoli___ has quit IRC03:03
*** zoli__ has joined #baserock03:03
*** zoli__ has quit IRC03:50
*** zoli__ has joined #baserock03:50
*** zoli__ has joined #baserock04:51
WalkerdineSwap space is working out so far04:52
*** zoli__ has quit IRC04:57
*** zoli__ has joined #baserock05:52
*** zoli__ has quit IRC05:57
*** zoli__ has joined #baserock06:31
*** paulw has joined #baserock06:38
*** zoli__ has quit IRC06:45
*** zoli__ has joined #baserock06:46
*** zoli__ has quit IRC06:48
WalkerdineHey it worked! Finally!06:55
radiofreeWalkerdine: which component was running out of memory?07:16
WalkerdineOkay now that I deployed my upgrade how do I get back to the baseline? I thought I'd be able to select which one it booted into07:24
radiofreesystem-version-manager list07:27
radiofreeand probably you want to do `system-version-manager set-default factory`07:27
WalkerdineHow do I do that if I'm in the Genivi demo platform?07:27
WalkerdineI was probably supposed to do that first wasn't I07:28
WalkerdineWait jk07:34
WalkerdineUnless I send the command over ssh or something07:34
WalkerdineYep sorry still getting the hang of this07:37
paulsherwoodkeep going :)07:37
WalkerdineI'm finally up and running! Now I can start to actually do something07:39
WalkerdineThanks for all the help07:40
paulsherwoodyou've been very patient :)07:40
WalkerdineIs there a way to do this using a VM that emulates an ARMv7? Or is the jetson really the only option07:43
*** tiagogomes_ has joined #baserock07:49
WalkerdineWelp anyway its time for bed. Goodnight everyone07:50
*** rdale has quit IRC07:50
WalkerdineOr should I say goodmorning07:50
pedroalvarezbaserock never sleeps :)_08:03
*** bashrc has joined #baserock08:04
*** toscalix_ is now known as toscalix08:05
*** mariaderidder has joined #baserock08:13
*** paulw has quit IRC08:13
*** zoli__ has joined #baserock08:19
*** paulw has joined #baserock08:22
*** zoli__ has quit IRC08:24
*** zoli__ has joined #baserock08:27
*** zoli___ has joined #baserock08:36
*** zoli__ has quit IRC08:37
*** zoli___ has quit IRC08:37
*** zoli__ has joined #baserock08:42
richard_mawpaulsherwood: hmm, I can't leave comments in-line on the web UI. I'll send you an e-mail.08:42
*** rdale has joined #baserock08:48
paulsherwoodWalkerdine: we'll have news on that soon, i hope09:04
paulsherwoodrichard_maw: i could send the patch to the ml?09:06
richard_mawI've already generated the patch and am mid-response.09:06
richard_mawIf you want the review to be public I can send it to the list.09:06
paulsherwoodok by me09:07
richard_mawpaulsherwood: If you're sure it's ok. It can serve as a useful example.10:04
*** zoli___ has joined #baserock10:06
*** zoli____ has joined #baserock10:08
*** zoli___ has quit IRC10:08
paulsherwoodrichard_maw: are you going to trash my code? :-)10:08
paulsherwoodplease go ahead10:08
*** zoli__ has quit IRC10:08
KinnisonIt's a way for everyone to learn10:09
richard_mawpaulsherwood: you've made some of the most useful mistakes for explaining some of the intricacies of locking10:09
* richard_maw thinks the biggest wart on the flock API is that you can't pass a timeout10:10
Kinnisonthreads and alarums10:11
* Kinnison hides10:11
rjekalarm() and a signal handler10:11
richard_mawKinnison, rjek: Yes, I'm aware of how you have to get around it. The result is that potentially the least ugly way of doing a blocking lock with a timeout is to shell out to flock10:12
richard_mawpaulsherwood: review sent10:20
*** zoli____ has quit IRC10:24
*** zoli__ has joined #baserock10:24
*** paulw has quit IRC10:32
*** paulw has joined #baserock10:33
*** zoli__ has quit IRC10:56
*** zoli__ has joined #baserock10:56
*** zoli__ has quit IRC10:56
paulsherwoodrichard_maw: rebuttal attempted :)11:21
*** paulw has quit IRC11:29
*** zoli__ has joined #baserock11:45
WalkerdineIs the functionality of working on a virtual arm just not there right now for baserock or there just isnt a well documented way of doing it?13:29
KinnisonARM emulation is expensive13:30
radiofreewell you can always run it in qemu-arm but it'll be very slow13:30
Kinnisonthe best emulators seem to achieve a hundred MHz per GHz of x86 or thereabouts13:30
paulsherwoodKinnison: on that basis my scaleways results are not so bad :)13:32
Kinnisonpaulsherwood: indeed13:32
WalkerdineSo would VMware be my best option then?13:36
Kinnisonvmware for ARM?13:37
WalkerdineOr do they not have support for that13:37
rjekI don't know if qemu-kvm does ARM systems yet13:37
rjekBut VMware do not have an ARM product13:38
paulsherwoodWalkerdine: i've been building arm baserock systems on machines recently, it's quite fun, but still quite slow13:40
rjekpaulsherwood: How many have you tried using at once?  How does it compare to a Moonshot m700 cartridge?13:41
* rjek keeps meaning to play with Scaleway13:41
paulsherwoodrjek: one machine doing the ci:
rjek20 hours?13:42
paulsherwood5 in a herd got it down to 11:10:0013:43
*** zoli__ has quit IRC13:43
rjekIf only they'd slung an mSATA card under each CPU :)13:43
* paulsherwood can't find any moonshot logs13:44
*** zoli__ has joined #baserock13:44
*** Walkerdine has quit IRC13:47
*** Walkerdine has joined #baserock13:55
*** zoli___ has joined #baserock14:06
*** zoli__ has quit IRC14:08
*** rdale has quit IRC14:12
paulsherwoodrichard_maw: can we talk briefly here about this locking thing?14:16
paulsherwoodmy point is i only need to know the set of files at the point i take the exclusive lock. once i know that set, i can take as long as i like trying to delete them, without a lock14:17
paulsherwoodand if other ybds do the same, it doesn't matter so long as they do so in the try block14:18
richard_maw1. converting the exclusive lock into a shared lock is sufficient to prevent you entering the situation where you have two YBDs attempting to remove files at the same time14:20
richard_maw2. just throwing them into a try-except is not sufficient14:20
richard_mawlet me finish14:20
richard_mawbecause a) when it throws an exception because the file it wants to remove no longer exists, it doesn't then resume removing the rest14:20
richard_mawso when they do stomp on each other, the most you can hope for is that only one of the removes stops working14:21
richard_mawwhen if they both try to remove the same set of files in different orders, they will both stop trying to remove things14:22
richard_mawand b) the try-except is in the wrong place to prevent that14:22
richard_mawsince if you get an exception, it will jump into that handler14:22
richard_mawand then proceed to start another build14:22
paulsherwoodas that all?14:22
paulsherwood1. i want an exclusive lock only to be sure no other instances are building at this point. if i have a lock, i can be 100% that the to_delete files can go.14:24
paulsherwooddo you agree with the above?14:24
paulsherwooda shared lock does not give me that confidence14:26
paulsherwoodany other ybd hitting the lock skips cleanup14:27
richard_mawyes, I *get* that point, it's that you then release the lock, so there's a period that isn't locked, rather than atomically converting it into a shared lock, so no *other* YBD can decide it knows everything and can start deleting files14:27
paulsherwoodthe shared lock has a different purpose. it says i'm using the tmpdir.14:29
paulsherwoodfor the occasional situations where more than one ybd believe they know what to delete, they may trip over each other14:29
paulsherwoodbut that's still ok, for me - partial cleanup is better than none. and i think that case is rare14:29
paulsherwood(and they don't crash)14:31
KinnisonI think you're wilfully ignoring a valid issue just because you think it'll be rare14:32
KinnisonWhich is your perogative14:32
Kinnisonbut you're also missing the point that there is a period where you have no lock14:32
paulsherwoodi'm not missing the point.14:33
KinnisonIn traditional locking strategies, you don't acquire more locks after releasing any14:33
Kinnisonhence the atomic conversion recommendation from richard14:33
paulsherwoodKinnison: to be clear, the impact of this issue is that it doesn't clean optimally once in a while14:35
paulsherwoodnot that anything will crash or builds will fail spuriously14:35
paulsherwood(except ENOSPACE)14:36
KinnisonIf you state that "When ybd runs, it will make an attempt to clean up temporary directories, which may or may not succeed to a greater or lesser extent"14:36
Kinnisonthen it meets your implementation14:36
* Kinnison would maintain that this is a worthless assertion14:37
rjekThat's horrible14:37
Kinnisonbut that's your perogative14:37
Kinnison"when you run this, it might maybe do something useful, or not, YMMV"14:37
Kinnisonnot my favourite tool definition14:37
* rjek wonders what the problem with making it never fail is.14:37
richard_mawin the very least *please* wrap your subprocess with its own try-except, so that it doesn't attempt to start another build when two builds attempt to clean up the same thing14:38
paulsherwoodrjek: so far, the problem with that is i'm getting loads of information about locking as if i don't know what i'm talking about, rather than a few lines to improve the implmenetation14:38
richard_mawI promised review, not patches.14:38
richard_mawand I've posted examples14:38
paulsherwoodthank you14:39
paulsherwoodi'm promising code that works effectively, not 100% efficiently14:39
KinnisonI'd argue that "incorrectly" or "intermittently" doesn't equal "effectively" but you're welcome to disagree14:47
* Kinnison does, however, not care since we'll just clean up /src/tmp *before* invoking ybd14:47
*** mariaderidder has quit IRC15:00
paulsherwoodin other news, i got so involved in this i've missed my flight :/15:01
*** mariaderidder has joined #baserock15:12
*** zoli___ has quit IRC15:15
paulsherwoodmaybe it serves me right :)15:44
*** mariaderidder has quit IRC16:47
*** bashrc has quit IRC16:49
*** tiagogomes_ has quit IRC18:23
*** toscalix has quit IRC18:26
*** zoli__ has joined #baserock18:34
*** zoli___ has joined #baserock19:12
*** zoli__ has quit IRC19:14
*** zoli__ has joined #baserock19:27
*** zoli___ has quit IRC19:27
*** zoli__ has quit IRC19:38
*** zoli__ has joined #baserock19:38
*** zoli__ has quit IRC19:56
*** zoli__ has joined #baserock19:56
*** zoli___ has joined #baserock20:33
*** zoli__ has quit IRC20:33
*** Walkerdine has quit IRC20:35
*** jjardon has quit IRC23:03
*** jjardon has joined #baserock23:09

Generated by 2.15.3 by Marius Gedminas - find it at!