[09:35] * cjwatson starts trying to saw up his Git MP work from the plane into manageable pieces [09:35] blr: You probably didn't see, but I got MPs working on Saturday evening \o/ [09:35] (minimally, anyway) [09:36] wgrant,blr: Are we holding the team meeting at the calendared time tonight? [09:39] That time's a bit awkward for me this week but I guess we might be able to get away without it. [09:43] It's exactly as awkward for me as it normally is :-) [09:44] Heh, true. [09:45] How goes the release sprint? [09:46] The usual amusing crop of installer bugs. [09:46] Except all with added systemd fun this time. [09:47] eg. oem-config breaking because of a shadow bug from two years ago because systemd. [09:47] Only two last-minute sabdfl changes so far. [09:50] cjwatson: Are you comfortable enough with your MP changes to make schema review sensible at this stage? [09:53] wgrant: Yeah, my threshold was "works locally", which it does. [09:53] Great, will look soon [09:54] Do you think I need to ask thumper etc. if they're OK with removing BMQs? [09:54] Or just assume that if they haven't finished it in the last 5+ years they're unlikely to now? [09:54] Meh. [09:54] They're unused, and have been disabled for like five years. [09:54] If someone wants to revive them, they can feel free. [09:54] * cjwatson nods [09:55] VCSes are cool like that. [09:55] :-) [09:55] Vanishing. [10:08] cjwatson: https://code.launchpad.net/~wgrant/launchpad/changelog-formatting/+merge/256898 [10:08] Quality tests. [10:11] wgrant: Nice, r=me. [10:12] Thanks, bugged me for five years or so... [10:13] [10:14] cjwatson: Not oversized? I'm disappointed. [10:14] wgrant: Hm? [10:14] Your next git branch is just under 800 lines! [10:15] Ah :-) [10:15] I have another smallish one, but then need to think a bit ... [10:15] The whole diff's 4kish, right? [10:16] 4060 right now, though I'll need to write some more tests. [10:20] cjwatson: boom, btw [10:22] Grr. [10:24] So much for having carefully run all the code tests and thought that was probably enough ... [10:29] Heh [10:29] Ah, maybe I didn't apply the changed security.cfg first. [10:30] I'll just revert security.cfg, not worth the hassle. [10:31] Yep [10:40] Testfix landing. [11:22] wgrant: Do you know why the existing Branch:+subscribe etc. permissions are the way they are? I was copying from those. [11:23] cjwatson: Probably from before private branches. [11:24] The lack of a ViewAndAnyPerson-ish thing has always been annoying. [11:25] What would the difference be between that and just View here, given that an anonymous user can't subscribe to things? [11:25] Oh, but View would *allow* an anonymous user to subscribe. [11:26] Exactly. [11:26] I guess it's sufficient to just rely on you not being able to get at the Branch/GitRepository to subscribe to it in the first place if you don't have View. [11:27] The portlets could be launchpad.View at least, though. [11:30] Yeah, that's the way it works now, but it's a bit unfortunate. [11:30] Particularly given that +subscribe is the access control view. [11:30] I'll put something in Asana for it. [11:37] cjwatson: Can I upgrade DF? [11:37] Sure. [11:42] cjwatson: Have you timed the new check constraints? [11:43] wgrant: I don't have the scrollback, but that was what I was doing during the wrap-up when I accidentally failed to do it in a transaction :-) [11:43] wgrant: IIRC they were something like 0.2 seconds each [11:43] Oh,heh. [11:44] Right. [11:44] It's not a big table. [11:45] Yeah, on DF I get 153.458 ms, 95.952 ms, 94.037 ms, 69.866 ms, 125.432 ms [11:46] So the whole patch might end up being as much as a second or two but shouldn't be much worse. [11:47] cjwatson: Does different_git_refs really handle a null dependent_git_repository proprely? [11:49] wgrant: Should do. ROW (NULL, NULL) != ROW (1, 1), e.g., and I tested without a prerequisite locally. [11:50] cjwatson: And one_vcs doesn't check that the VCS is consistent. [11:50] Just that each of source, target and dependent only has ne. [11:50] source_git_repository | source_git_path | target_git_repository | target_git_path | dependent_git_repository | dependent_git_path [11:50] -----------------------+----------------------+-----------------------+-------------------+--------------------------+-------------------- [11:50] 1 | refs/heads/ibm-nvram | 1 | refs/heads/master | | [11:52] wgrant: Hm, yes, that's true. [11:52] Should be (source_branch IS NULL OR target_branch IS NULL OR dependent_branch IS NULL) != (source_git_repository IS NULL OR target_git_repository IS NULL OR dependent_git_repository IS NULL), I think. [11:53] Er, s/IS NULL/IS NOT NULL/g [11:56] Fixed. [12:02] cjwatson: I think that new one_vcs allows me to completely omit target or source, as long as one of the others is set. [12:02] But apart from that it looks reasonable now. [12:03] Oh, that's true. [12:07] Will fix after lunch. [13:17] wgrant: How does that look to you now? [13:17] ALTER TABLE BranchMergeProposal [13:17] ADD CONSTRAINT one_vcs CHECK (((source_branch IS NOT NULL AND target_branch IS NOT NULL) != (source_git_repository IS NOT NULL AND target_git_repository IS NOT NULL)) AND (dependent_branch IS NULL OR source_branch IS NOT NULL) AND (dependent_git_repository IS NULL OR source_git_repository IS NOT NULL)); [13:19] cjwatson: That works. [13:19] brushing off my boolean logic [13:19] I want IMPLIES [13:19] A truth table would be more compact... [13:20] Yeah [13:24] p → q ≡ ¬p ∨ q, but. [15:31] wgrant: Did you intend to review db-git-mp in the MP itself? [15:38] cjwatson: Er, yeah, if you've pushed that. [15:38] Sorry, distracted by OpenStack. [15:39] I pushed that change, yes. [15:39] How goes the restacking? [15:39] Finally got the x86 cloud back up with trusty+kilo, doing builds, next step is to see if nova-compute works on cameron. [15:40] And then weneed to hack up the imemaker bits so we can hulksmash them onto cameron for ppc64el only. [15:43] works on cameron> Hopefully you won't have done the kilo upgrade for nothing ... [15:46] Huh, I wonder why I made GitRepository.name readonly=True? [15:47] cjwatson: It'll at least be easier to fix on kilo! [15:47] cjwatson: name needs a setter, at least. [15:48] Branch.name doesn't have one. I assume you just get a constraint violation if you try to make it equal to an existing one. [15:49] Ugh. [15:49] And trying to set GitRepository.owner crashes, bah. [15:51] I'll be able to QA subscriptions eventually ... [20:29] cjwatson: excellent, great news :) [20:31] should have the cross-repo diff soon, yesterday was a bit of a write-off. [20:52] I can imagine [20:52] Not been hugely useful today myself ... [22:03] wgrant,blr: Are we meeting?