IRC logs for #baserock for Tuesday, 2016-04-19

*** gtristan has quit IRC00:18
*** locallycompact has quit IRC00:40
*** gtristan has joined #baserock00:45
*** thrace has quit IRC03:33
*** thrace has joined #baserock03:36
*** persia has quit IRC03:40
*** persia has joined #baserock03:50
*** CTtpollard has joined #baserock07:14
*** bruce_ has joined #baserock07:26
*** rdale has joined #baserock08:07
*** toscalix has joined #baserock08:10
*** anahuelamo has joined #baserock08:19
*** tiagogomes has joined #baserock08:48
*** edcragg has joined #baserock08:53
*** CTtpollard has quit IRC08:54
*** CTtpollard has joined #baserock08:57
*** jonathanmaw has joined #baserock09:01
*** ssam2 has joined #baserock09:03
*** ChanServ sets mode: +v ssam209:03
*** franred has quit IRC09:15
jjardonpaulsherwood: why ?09:24
pedroalvarezjjardon: I guess because it was merged without testing, and it doesn't work?09:26
jjardonpedroalvarez: Does not? Anyway, reverting that will make it not even build09:29
*** franred has joined #baserock09:29
pedroalvarezit even had a -1 from the author of the change..09:30
paulsherwoodjjardon: i'm happy to abandon
*** locallycompact has joined #baserock09:30
paulsherwoodpedroalvarez: we merged the original changhe on the assumption that you were no longer caring about gdp09:31
* paulsherwood has no mechanism to test it, currently09:31
pedroalvarezthen we  have to remove it then09:31
paulsherwoodno, we do not09:32
paulsherwoodwhy is removing it better than leaving it there?09:32
pedroalvarezif we don't care about it anymore, why do we have to maintain it?09:32
paulsherwood*we* do care... you said *you* didn't09:33
*** pedroalvarez has left #baserock09:33
paulsherwoodif i had known you were still interested i'd have waited to let you comment on jjardon's patch09:34
paulsherwoodand i'm trying to establish an alternative way to deal with this (including testing it)09:35
paulsherwoodso we should not remove it, imo09:35
*** edcragg has quit IRC09:40
*** edcragg has joined #baserock09:40
jjardonSo far the subdirectory idea seems the best approach09:47
paulsherwoodwhich subdirectory idea?09:48
jjardonpaulsherwood: yours :)09:48
paulsherwoodack :)09:49
jjardonLast comment in is that rigth?09:51
radiofreedirectories works, as long as it's a wholesale copy09:52
radiofreei don't like the yocto approach of "appending" things09:52
paulsherwoodradiofree: my suggestion was not appending, but only having the delta in (say) the gnome dir etc09:52
radiofreeyou end up having to navigate a maze of recipes trying to figure how wtf is being built09:53
paulsherwoodso core, foundation stay under (say) baserock... and gnome has the gnome strata, while referring to baserock/strata/core and baserock/strata/foundation etc09:53
paulsherwoodi *believe* that ybd would handle this properly already... but haven't tested it for a long time09:54
radiofreeand if gnome needed to change foundation?09:57
radiofreecopy over foundation to gnome?09:57
paulsherwoodand rename it to foundation@gnome09:58
radiofreewell, i doubt it would be gnome changing foundation09:58
* paulsherwood would like to see unique names for definitions09:58
* radiofree has personally never been a fan of foo@bar.morph09:59
radiofreelooks like an e-mail address09:59
paulsherwoodwell, i'm happy to consider another convention for that. but i dislike having multiple foo.morph files09:59
rjekBang paths!09:59
paulsherwoodwhat is a bang path?09:59
rjekWhat we had before @10:00
paulsherwoodoh, ok10:00
rjek-the first !10:00
* radiofree takes what he said about @ back10:00
edcraggis this a new bug in ybd? '6 16-04-19 00:00:57 [/386/386] [stage2-gawk] Try downloading'10:26
edcraggthe counter seems squiffy10:26
paulsherwoodedcragg: sadly, no
paulsherwoodi've been meaning to re-write the counter code10:29
*** richard_maw has left #baserock10:31
*** richard_maw has joined #baserock10:42
*** ChanServ sets mode: +v richard_maw10:42
*** CTtpollard has quit IRC10:44
*** CTtpollard has joined #baserock10:57
*** gtristan has quit IRC11:44
*** ssam2_ has joined #baserock12:36
*** ssam2 has quit IRC12:39
*** bruce_ has quit IRC13:57
paulsherwoodbenbrown_: iiuc you're working on some improvements for trove?14:28
paulsherwoodplease could you email the list describing what you're thinking of? it would be a shame if you did a load of work and folks refused it14:28
*** CTtpollard has quit IRC14:28
*** CTtpollard has joined #baserock14:35
*** bruce_ has joined #baserock14:57
paulsherwoodedcragg: sorry, to clarify, are you saying that your removal of that shuffle fixes the problem, but also slows the builds?15:05
* paulsherwood wonders how it could slow the builds15:05
edcraggpaulsherwood: yes, indeed15:18
SotKits being slowed when instances > 1 right?15:21
edcraggthe same number of instances in both cases, just without randomisation in the installation of dependencies15:25
edcraggi imagine because of increased clashes and timeouts, but it's not clear to me at this stage15:26
edcraggSotK: but it is when instances > 1, simply because there is no randomisation when that is the case15:31
*** fay_ has quit IRC15:31
paulsherwoodedcragg: 'inherent in the definitions, and morph seems to manage15:33
paulsherwoodmeh :)15:33
edcraggmorph has never shown similar problems, like creating non-bootable systems, as far as i know15:34
paulsherwood'ordering is inherent'... is more like 'ordering is accidentally ok', imo15:35
edcraggi do think there is an order in the evaluation of the definitions, normally15:35
SotKin that case, the slowdown is likely because you'll end up with a number of ybd instances all building the same thing the whole time15:35
locallycompactthere is an order by fiat15:35
edcraggby design, busybox is higher up than systemd, so systemd is installed on top15:36
paulsherwoodthanks for investigating, edcragg - i'll think about what to do15:36
benbrown_paulsherwood: I've sent some stuff marked RFC to gerrit already, other things are still being discussed/thought about15:37
paulsherwoodbenbrown_: pls send it to ML... more readers there15:37
edcraggpaulsherwood: np15:39
benbrown_paulsherwood: Very well, I send the patches yesterday fwiw15:40
ssam2_cover letters sent to the mailing list are nice for giving the overall context & intention to a series of patches on gerrit15:41
edcraggSotK: when one process attempts to build something that's already building, it does realise this, and after a timeout parses the definitions again until if finds another random dependency that is still not present in the local cache15:41
franredbenbrown_, do not need to send the patches to the mailing list, just a general description of what you want to achieve with them and the links to the gerrit patches15:42
franredssam2_, snappish :)15:42
edcraggthat's a good idea :)15:42
benbrown_franred: Aye, had no wish to resend the contents of the patches :)15:44
SotKedcragg: does it still do it randomly if you've turned off the randomization?15:49
edcraggSotK: in what i tried, it should still be randomising the order it evaluates the dependencies, but preserving the order that they are installed into the sandbox15:52
paulsherwoodi'll figure it out16:02
*** bashrc has quit IRC16:15
*** bashrc has joined #baserock16:18
SotKedcragg: in that case, I apologise for making assumptions based on my outdated knowledge :)16:19
edcraggSotK: i *think* that's how it works, anyway16:23
paulsherwoodrdale: interesting idea... *can it*?16:25
paulsherwood"Can't this be fixed by writing suitable artifact splitting rules that exclude duplicates?"16:25
*** anahuelamo has quit IRC16:26
rdaleyes, i think so at system assembly time - it depends if it is a problem at build time where artifact splitting doesn't apply16:26
paulsherwoodno, this is only a problem at system assembly time i think16:26
rdalei think it can be fixed with artifact splitting then16:26
* paulsherwood gets excited16:27
*** richard_maw has left #baserock16:27
*** anahuelamo has joined #baserock16:27
edcraggi'm not sure how that would work16:29
edcraggwhat if you want to fall back to something, in the absence of something that would normally be used instead?16:29
*** anahuelamo has quit IRC16:29
paulsherwoodi don't understand your usecase for that, edcragg - most of the situations we have are where busybox is clashing with something else, iiuc16:30
rdalei would split off say 'awk' from the normal busybox-runtime artifact into busybox-awk for instance16:30
edcraggah, ok16:31
*** toscalix has quit IRC16:31
*** fay_ has joined #baserock16:35
edcraggrdale: do you then not install busybox-awk when you have gawk, for example?16:36
*** richard_maw has joined #baserock16:43
*** ChanServ sets mode: +v richard_maw16:43
*** jonathanmaw has quit IRC16:54
*** jonathanmaw has joined #baserock16:56
*** jonathanmaw has quit IRC16:59
*** jonathanmaw has joined #baserock16:59
*** pedroalvarez has joined #baserock17:00
*** ChanServ sets mode: +v pedroalvarez17:00
*** jonathanmaw has quit IRC17:00
*** bashrc has quit IRC17:02
*** cosm has quit IRC17:03
*** ssam2_ has quit IRC17:04
*** locallycompact has quit IRC17:10
*** edcragg has quit IRC17:10
*** cosm has joined #baserock17:16
*** tiagogomes has quit IRC18:03
*** rdale has quit IRC18:07
*** juergbi has quit IRC18:09
*** juergbi has joined #baserock18:12
*** paulwaters_ has joined #baserock19:43
*** paulwaters_ has quit IRC19:45
*** locallycompact has joined #baserock20:36
*** locallycompact has quit IRC22:19
*** persia has quit IRC22:46
*** thrace has quit IRC22:47
*** astrophys has joined #baserock22:51
*** persia has joined #baserock23:04
*** thrace has joined #baserock23:06
*** bruce_ has quit IRC23:13
*** gtristan has joined #baserock23:35
*** edcragg has joined #baserock23:44
*** edcragg has quit IRC23:56

Generated by 2.15.3 by Marius Gedminas - find it at!