[06:16] === Top 10 Time Out Counts by Page ID === [06:16] Hard / Soft Page ID [06:16] 131 / 5921 Archive:+index [06:16] 64 / 255 BugTask:+index [06:16] 21 / 263 Distribution:+bugs [06:17] 9 / 115 ProjectGroupSet:CollectionResource:#project_groups [06:17] 8 / 15 NullBugTask:+index [06:17] 6 / 37 Milestone:+index [06:17] 5 / 202 POFile:+translate [06:17] 5 / 87 Archive:+packages [06:17] 5 / 4 ProjectGroup:+milestones [06:17] 5 / 0 BinaryPackageBuild:+retry [06:18] wgrant: still interested in DistroSeries:+queue [06:19] /srv/launchpad.net/production/launchpad-rev-12004/lib/lp/soyuz/browser/../templates/distroseries-queue.pt [06:19] - Line 281, Column 11 [06:19] - Expression: [06:19] ... [06:19] LocationError: (None, 'filesize') [06:19] That's not meant to happen. [06:20] 0 LocationError: (None, 'filesize') [06:20] bah [06:20] 10 LocationError: (None, 'filesize') [06:20] I think Translations may be expiry overzealously. [06:20] s/expiry/expiring/ [06:20] well [06:20] the page should handle gc'd content anyhow [06:20] No. [06:20] no? [06:21] It's not allowed to be GC'd... [06:21] obviously nothing is enforcing that [06:21] I'd accept not /meant/ to be GC'd [06:21] There's nothing we can sanely do if it is GC'd. [06:21] It is forbidden. [06:21] But someone broke it. [06:21] but permitted [06:22] By the DB? Sure. [06:22] It is not possible to sanely apply that constraint in the DB. [06:22] I can think of at least one way. [06:23] Mortifyingly expensive triggers? [06:23] separate tables [06:23] so the rules in each table are succinct [06:23] is bug 262767 related? [06:23] <_mup_> Bug #262767: Expire translations-tarball file contents once they get successfuly processed [06:25] https://bugs.launchpad.net/soyuz/+bug/592417 [06:25] <_mup_> Bug #592417: LocationError in DistributionSourcePackageRelease pages

[06:26] The idea in 262767 is what I was going to blame. But it seems to not be implemented anywhere yet. [06:27] 592417 is about real things. Not custom uploads. [06:27] So it's not really related. [06:27] bug 685401 [06:27] <_mup_> Bug #685401: LocationError in DistroSeries:+queue [06:27] lifeless: So, it could just not show them. But the point of the page is to allow packages to be accepted. [06:27] Can't accept something that's not there. [06:28] somethings thats gc'd isn't there, right ? [06:29] Right. [06:29] so theres no need to accept it [06:29] add a garbo job/fix the expiry code to /really/ delete it. [06:29] and we're done, no? [06:29] Hmmm? Expiring an unaccepted queue item doesn't make sense. [06:29] It is illegal. [06:29] It *should* crash. [06:30] I don't see why,. [06:30] you may have some axiom you need to share. [06:30] Hmm, actually, do you have the URL for that OOPS? [06:30] yes [06:30] in da bug [06:31] No query string? [06:33] I think it must have been the Done queue. [06:33] The query string will tell us. [06:34] Maybe Translations did actually run the query that they suggested. [06:34] queue_state=3&queue_text=&start=126240 [06:34] Aha, it is indeed Done. [06:34] So we can probably reasonable hide expired files in those cases. [06:34] But Translations is still doing something that is technically illegal. [06:35] And will break once we unfuck custom upload handling. [06:35] whats the illegality [06:37] lifeless: Before acceptance: you're deleting part of an upload that hasn't yet been processed. [06:37] After acceptance: you're deleting part of an upload that will be used during a copy once we remove 2004-era insanity. [06:39] Perhaps for concerningly more broad values of 'we' than would be traditional. [06:40] wgrant: if you don't want the upload processed, why is deleting it a problem? [06:40] wgrant: what copy? [06:42] lifeless: If you don't want the upload processed, you should reject it. [06:42] lifeless: Any copy. [06:42] Translations has removed all old translations tarballs. [06:42] wgrant: so, will you expand the bug appropriately? And suggest the right way forward to achieve their goals [06:42] ? [06:42] lifeless: I will. [06:43] keeping old data because it was used once, is nuts. [06:43] To an extent. [06:45] We need to actually develop a policy. [06:45] And make the code respect it. [06:45] Not develop over 5 years. [06:46] we have a policy - don't delete anything ever. Except where we need to (which is broadly everywhere), then we can, if we need to. [06:47] Well. [06:47] The Soyuz policy until 18 months ago was "never delete anything". [06:47] Then it was changed. [06:47] Without actually making the code work like that. [06:47] Or thinking much about the consequences. [06:48] waheee [06:48] The fact that the hashes are stored in the deleted LFC is a significant problem :( [06:48] don't worry, md5 has been broken right? [06:49] I really don't like deleting data until we're really, really sure. [06:50] We have lost far too much already. [06:51] -> nz [06:51] See you. [06:51] Damn. [17:59] off topic - but anybody here happen to be a god with vim modelines in source files? === pjdc_ is now known as pjdc [18:46] mtaylor: yo [18:47] lifeless: so, I have vim set up in my .vimrc to use spaces and to to do 2-space tabs (since that's what we do in drizzle) ... but I'm editing this file that uses tabs at the moment... and I was hoping there was an easy way to put a modeline into the file so that vim would behave properly [18:49] noexandtab [18:49] blah [18:49] noexpandtab [18:49] sw=8 [18:49] that sort of thing [18:49] lifeless: you make me happy [18:49] its a gift [18:49] probably sts=8 too [19:00] lifeless: works perfectly. thanks [19:00] de nada [19:48] morning [20:32] it's going to be a quiet day [20:33] how can you tell? [20:42] I'm on leave [20:47] and wally is on leave [20:47] and mwhudson is more quiet than normal since the linaro move [20:47] so I'm mostly left to myself... [20:47] * thumper goes to make a coffee [23:17] lifeless: is the launchpad copy of testtools up to date? [23:17] lifeless: it seems to be lacking some matchers [23:19] thumper: EndsWith ? [23:19] jelmer: yeah [23:20] also, I'd find IsNot(None) easier to read than Not(Is(None)) [23:20] thumper: That was added fairly recently, and IIRC the revision after the revision that lp uses. [23:21] thumper: I was planning to update lptools to a newer revision anyway, jml added some nice code to help with debugging of "lost" deferreds. Perhaps I should just do that.. [23:22] jelmer: sounds good to me [23:27] tip is broken. [23:28] jelmer: ^ [23:28] jelmer: if you wanted to review my fix for the critical bug, that would be good. [23:28] lifeless: Sure, I can do that. [23:30] lifeless: Ah, that one! I should've remembered. [23:33] gym time