IRC logs for #baserock for Saturday, 2015-11-21

*** edcragg has quit IRC00:09
*** locallycompact has joined #baserock01:08
*** locallycompact has quit IRC01:36
*** locallycompact has joined #baserock01:37
*** locallycompact has quit IRC02:04
*** locallycompact has joined #baserock02:05
*** locallycompact has quit IRC02:10
*** tlsa_ has joined #baserock02:18
*** SotK_ has joined #baserock02:18
*** tlsa has quit IRC02:19
*** SotK has quit IRC02:19
*** bwh has quit IRC02:19
*** bwh has joined #baserock02:19
*** gtristan has quit IRC02:32
*** locallycompact has joined #baserock07:44
*** locallycompact has quit IRC07:52
straycatcan we set up mason to build the devel system? if there's a release then folks should just be able to pull down all the artifacts for a system09:41
straycatpedroalvarez, around?10:41
pedroalvarezKind of10:44
pedroalvarez(phone)10:44
pedroalvarezJust switched them back to build ci.morph10:44
straycatokay well i think we need https://gerrit.baserock.org/1480 to fix mason10:45
pedroalvarezI tried to upgrade morph on them before reading your request, and everything fails...10:45
pedroalvarezRight, that fixes the error10:47
pedroalvarezThat will break mason with current version of morph10:47
straycatokay, i'd assumed mason had upgraded itself at this point10:48
pedroalvarezAnd that was true, but I will wait until it builds at least the devel system10:49
straycatokay that sounds good10:49
* richard_maw has been fiddling with build sandboxing again10:50
richard_mawI've got safe, rapid git cloning10:51
straycatcool :)10:52
richard_mawby cloning with --shared, and then bind-mounting the objects tree into the chroot and making a new objects/info/alternates file to use the bind-mounted tree10:53
richard_mawnext up is looking at using sandboxlib, then artifact caches10:54
straycati didn't know about --shared, that's neat10:54
richard_mawyeah, if you can trust it, then it's great10:55
richard_mawthe reason why I spent so long on gitcache, was that I needed to make --shared safe10:56
richard_mawso there's file locking to prevent you breaking the repository if you use the api10:58
straycatsounds good :)10:59
richard_mawthe git worktree stuff has a more clever implementation for the subset of the problem that we share11:01
richard_mawbut it's not exposed at that level11:02
richard_mawand requires a very new version of git11:03
richard_mawbut I had to patch linux-user-chroot to support bind-mounting files from the host in11:04
straycatmultiple worktrees share the same set of objects though? so you can't break the repo by removing an object that's still referenced by the other repo?11:04
straycat(so i don't see what the subset is)11:18
richard_mawwe both have the problem of needing to know what is in use from other repositories11:28
* pedroalvarez fears tools that depend on forks of other tools 11:28
richard_mawyeah, I'm going to submit my luc patch when I get around to it11:29
straycatrichard_maw, oh okay, i thought worktrees were just copies11:30
richard_mawworktrees are clever, assigning a different index, head and checkout to the same refs namespace, I need independent ref namespaces, and each to act as completely separate repositories, rather than a shared view of the same repository11:33
straycatokay, so in the case of worktrees it's what's in use from other worktrees, rather than other repos11:39
richard_mawyes, there's a lot of common behaviour, but unfortunately there is a key difference that makes sharing code difficult11:53
*** tlsa_ is now known as tlsa12:04
*** tlsa has quit IRC12:09
*** wdutch has quit IRC17:44
*** wdutch has joined #baserock17:45
*** jjardon has quit IRC17:45
*** jjardon has joined #baserock17:49

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!