[07:00] blr: Evening. [07:09] wgrant: morning! [07:10] blr: I saw that your default clone_from thing landed, testing it now. [07:10] Did you make any progress on the init with alternates? [07:10] wgrant: merged your recent changes into the alternates branch as well now, which is much better - only change in the view is docstrings for repo:repo [07:10] Great. [07:11] the only thing I'm a little uncertain about is that there is now obvious place to cleanup the ephemeral repo - I've added a repo_cleanup() method to both diff functions, but it might be better to have a sweeper do this later [07:11] s/now/no/ [07:12] I think that using a context manager makes sense. [07:13] eg. "with open_repo(blahblah) as repo:" would ensure that it's always cleaned up at the end. [07:45] The autocloning stuff works well, with one unforeseen issue. [07:45] The connectivity check on push uses rev-list --not --all, but --all only looks in the local ref. [07:45] s [07:46] So my first push to ~wgrant/ubuntu/+source/linux took a bit over a minute, as it ended up walking the entire tree when it had no known-good refs. [07:46] It may be worth patching git to let rev-list consider alternate refs as well. [07:47] But it's just a bit slow on first push, not a serious problem. [07:47] Particularly not for sensibly sized repos. [07:50] blr: What d oyou think of the context manager approach to ensure the ephemeral repo is always cleaned up? [07:56] wgrant: yep that sounds sensible, I'll have another pass [07:57] Thanks. [07:57] Let me know how it goes. [10:14] wgrant: open_repo() is a contextmanager now, that should be ready for you [10:15] blr: Let me see. [10:16] lifeless: I think you can do away with repo.ephemeral and cleanup_repo now, since you only call cleanup_repo in the ephemeral case. [10:17] Er [10:17] blr: ^^ [10:17] This ssh latency is really awful for typing. [10:19] wgrant: quite, that's a fairly useless function now. [10:20] ok pushed [10:22] blr: Couple of other trivial comments, then I think it's all good. [10:24] wgrant: hmm actually, in get_commit repo=None is used to prevent us re-initialising repositories in get_commits [10:24] Ah, that's right. [10:24] not certain how to perserve that nicely in the context of the context manager [10:25] blr: Maybe get_commits and get_commit should call a private method which just takes a repo that must already be open. [10:27] that sounds reasonable, I'll sort that out in the morning, getting late [10:27] thanks wgrant [10:27] Night. [10:35] wgrant: https://code.launchpad.net/~cjwatson/launchpad/git-fix-clone-from-owner-default/+merge/257622 if you get a minute, little bit of cleanup [10:37] wgrant: Creating MPs on the webservice now works, though https://code.qastaging.launchpad.net/~cjwatson/germinate/+git/germinate/+ref/exp/+merge/254302 will OOPS in the web UI for now. [11:58] cjwatson: Nice, how's the testing of the remaining KLOC going? [12:11] wgrant: Making progress, just thinking about how to mangle TestRegisterBranchMergeProposalView suitably. [12:11] wgrant: Have you heard anything from IS? [12:13] cjwatson: We're yellow. [12:13] Ah yes, thanks for the score bump. [12:48] wgrant: Is there any equivalent of SQLObjectVocabularyBase for Storm? [12:55] cjwatson: I don't think so. [12:57] I wonder if I should write one. I just noticed that GitRepositoryVocabulary is wrong. [12:58] It would not be a huge task. [16:55] https://code.launchpad.net/~cjwatson/launchpad/storm-vocabulary-base/+merge/257655 https://code.launchpad.net/~cjwatson/launchpad/git-mp-basic-browser/+merge/257674 https://code.launchpad.net/~cjwatson/launchpad/git-mp-register/+merge/257676 are all the MP stuff I have so far, except for my in-progress stuff on preview diffs. [16:56] Should be (barely) enough to go live with. [17:24] cjwatson: did you see recent discussions about git / merge proposal workflows on debian-devel et.al.? [17:25] cjwatson: it would be cool if launchpad would support emailing in git-bundle(1) to create mp just like one can with bzr bundles [17:25] xnox: I noticed the thread, but we have a fair bit still to do before getting to that kind of point [17:26] cjwatson: sure. plus i believe git-bundle has insufficient information as usual in the git world. E.g. bzr bundle encodes public/parent/submit branch locations, and git bundle does not as far as i remember.... [17:27] bzr bundle slso hasn't worked since 2009 [17:28] wgrant: the email interface or the bundles themself?! [17:29] Yes, I wondered why I couldn't find any of this alleged-to-exist code [17:29] i sure used bundles as recent as 2013-2014 e.g. generate it somewhere throw into usb-stick or email and pull from it on the other end.... [17:29] bundles never worked with 2a [17:29] so the LP side was removed [17:34] wgrant: just tested works here.... created --unchanged commit, branched, created --unchanged commit, bzr bundle ../first-branch, then branch first branch again and pulled from the bundle [17:34] got matching revision-ids out of it.... [17:34] * xnox is sad =( [17:43] ah yes, the unchanged test :P [17:43] that crestes little new 2a data [18:15] wgrant: well i did merges and the like... and it didn't fail me. [18:16] althoguht plenty of things are broken with ghost revisions [18:24] hmmm [18:30] wgrant: I can? thanks [18:34] lifeless: 500ms sucks [18:38] aye, where art thou? [18:38] Malta [18:38] nice [18:38] after Austin and London [18:38] I don't miss those trips [18:39] it's certainly a new experimebt in blood caffeine content [20:19] this looks like it might be useful https://github.com/libgit2/pygit2/pull/525 [20:39] blr: I thought about backporting it, but reckoned I could probably just reuse the stuff in LP (actually bzrlib I think) for now and no rush to backport. But I might change my mind if that turns out to be non-trivial :) [20:39] (That is, it looked like it'd be less work to reuse) [20:41] right [20:42] We'll see - I'm going to try to finish the LP side of preview diffs tomorrow [22:51] wgrant: whats em1.rapid.c.c ? [22:51] cjwatson: ^ [22:51] thomi: ^ [22:52] blr: ^ [22:52] any two-letter nicks around ? [22:57] lifeless: replacement for batuan, if you remember that [22:57] cjwatson: no; I've just got hold of the ops VG so we're good [22:57] cjwatson: thanks [22:57] VPN endpoint for misc crap [23:01] lifeless: what was it spamming you with? [23:01] wgrant: git requests. see #openstack-infra. blahdeblah is on it [23:04] hmmm? [23:05] * thomi makes whooshing motion above his head [23:05] hmm [23:05] thomi: outbound NAT gateway for OIL it appears [23:05] not totally implausible [23:06] I know what some of those words mean, so yay! [23:06] for moat of the weird non-IS-managee labs [23:06] wgrant: how's Malta? I'm surprised you're awake [23:06] awake is pushing it [23:07] there's still a load of CI jenkins stuff behind rapid isn't there? [23:07] the CI lab, all the ex-CTD stuff, servenab, hwcert [23:07] contra-terrene devices? [23:07] CTS, that is [23:08] right. not LP anyway :) [23:08] no [23:08] I was pinging clued up people in the most convenient channel :> [23:08] heh [23:08] batuan has ll the stuff you don't want to think about [23:08] lol... and you pinged *me* :D [23:09] thomi: penance for using the wrong channel :). Actually its the CI nature of it that led me to ping you [23:09] ahh [23:09] non-CI doesn't need many many many accesses to the same git repos [23:09] yep [23:09] blissfull ignornace is my motto... [23:09] thomi: and I hear you've done some CI stuff [23:09] anyway, night avain [23:10] wgrant: night, thanks. cjwatson ditto