[21:41] morning folks === wallyworld_ is now known as wallyworld [22:45] hi waigani [22:45] wallyworld: morning [22:45] hey [22:45] thumper: hello [22:45] wallyworld: https://bugs.launchpad.net/juju-core/+bug/1360286 [22:45] Bug #1360286: ppc64el and i386 unittests cannot find tools

[22:45] wallyworld: my guess is axw's bootstrap refactoring [22:45] wallyworld: do you think we should revert or fix? [22:46] wallyworld: blocking CI [22:46] could be, i'll look if my network connection satys up long enough :-( [22:46] let me look first [22:46] thumper: checking the user exists I can only see an unexported func state.checkUserExists [22:47] waigani: what is this for? [22:47] thumper: AddEnvironmentUser [22:47] thumper: checking if user exists before adding it [22:47] waigani: ok, there are two main possibilities [22:48] waigani: 1) check at the apiserver level using state.User(name) to load the user [22:48] 2) look inside the actual state method to add an environment user [22:48] I'd probably prefer 2. [22:48] thumper: okay, also how do we tell if they are local? [22:49] waigani: the tag Provider method is "local" [22:49] thumper: thanks [22:54] thumper: the failing test can be removed since the logic has changed - i'll do a branch with it skipped to unblock ci and confirm with andrew and then remove it [22:55] wallyworld: cool [23:15] wallyworld: http://juju-ci.vapour.ws:8080/job/github-merge-juju/buildTimeTrend [23:15] wallyworld: FYI looks like Build #374 (21-Aug-2014 07:52:39) axw environ-bootstrap-upload-tools is the one that started big time changes [23:22] thumper: our build times have also been going up because tests are really unreliable and we now fail back to -p 2 most times [23:22] ah [23:23] * thumper suggests that we only check once [23:23] fail faster if we are going to fail [23:27] trouble is the 2nd run normally works [23:27] but is much slower than with full parallelism [23:28] we just need to fix the freaking tests [23:36] wallyworld: agreed [23:37] the replicaset test seems to be overly fragile [23:37] and long [23:37] can we not force that out of our normal test run into something more suitable? [23:37] and write a robust test? [23:37] you would think so [23:38] also, jujud tests are horrible [23:38] and there's a failing uniter test