[01:44] Hmmm. [01:44] I think someone needs to make dpkg -b stab people unless I_SWEAR_THAT_I_AM_DEBUILD=1 is set. [01:53] hmm [01:53] need to make incomplete-* actual statuses. [02:04] :( 105 timeouts. [02:04] So close. [02:04] 36 / 18 Distribution:+bugtarget-portlet-tags-content [02:04] :/ [02:04] thats pretty good [02:04] Hmmmm [02:04] 20 of the timeouts are from edge. [02:05] yup [02:05] gotta get that redirect in [02:06] aaaa, it executes the query twice [02:06] *That's* not meant to happen. [02:06] Oh god. [02:07] It's not checking tags_cloud_data's emptiness in TAL, is it... [02:07] Yes, yes it is. [02:07] * wgrant blargs. [02:09] wgrant: double query? I did mention that :> [02:19] lifeless: So, grant an exception to Distribution:EntryResource:searchTasks and drop to 9s on Monday? :D [02:19] Hmm. [02:19] Not all the soft timeouts are listed. [02:20] There's 65 that aren't on the list... I guess it doesn't show if there aren't any hard timeouts :( [02:25] lifeless: https://code.launchpad.net/~wgrant/launchpad/bug-736002-seriously-this-time/+merge/58016 [02:26] wgrant: are you sure it works? [02:26] Yes. [02:26] why does it work? [02:27] Because it's a dict, not a resultset. So the define evaluates it, and now we have a non-lazy object. [02:27] And I have a postgres log. [02:27] Thanks. [02:27] ah, it wasn't a cached property? [02:28] No. Could do that, I guess, but it's so easy to avoid.. [02:29] possibly useful belt and bracer [02:29] given that there is no test [02:29] It's hard to give a useful test, since it's dominated by variable numbers of other queries :/ [02:29] \i/ [02:36] :( [02:36] chaenomeles still has no 'logs' rsync module :( [02:37] tell losa on monday [02:38] Oh hm. [02:38] These OOPS counts are wrong :/ [02:38] oops-tools doesn't know about the new wampee prefixes, nor chaenomeles at all. [02:38] heh [02:39] bug xyz about that. [02:39] That explains why they're mostly edge. [02:40] lifeless: I should file a bug on oops-tools about that? [02:41] there are bugs for auto adjusting already [02:42] there is a wiki page explaining how to update things laready [02:42] uhm [02:42] if its hard to find or whatever then yes, a bug. [02:44] I gather it has to be done through the web UI now. [02:44] By an unknown set of privileged people. [02:46] Ah, there we go. [02:46] New prefixes added. [02:46] You can import them from lp-p-c without auth. [02:46] nicely done [02:46] if only it was in cron [02:51] ..... [02:52] IE10 will not run on Vista. [02:52] Grar. [02:57] lifeless: Do you know how to rescan the last couple of days' OOPSes? [02:58] AIUI each scan scans all oops prefixes known [02:58] runs every 7 minutes [02:58] so may have a bit of a backlog, but should catch up pretty soon [02:58] Ah, true. [03:43] As expected p-s-c has done nothing over the weekend. [03:44] StevenK: on any of the 4 services? [03:44] lifeless: Er, it runs on loganberry only? [03:44] + staging [03:44] + qastaging [03:44] + dogfood [03:45] dogfood doesn't run it, and we need to re-import the database onto mawson in 1.5 weeks anyway [03:45] :) [03:45] staging won't need to after this weekends re-import [03:46] And qastaging, "meh" ? [03:46] lifeless: And also, it doesn't hurt anything if it stops running early -- it isn't like your index change, where stuff completly breaks if it's NULL [03:47] you sure about that ? [03:48] The only thing that uses SPR.changelog is DSD base version calculation -- if it can't find the ancestry, it will return early, not OOPS. [03:57] lifeless: He's right. [03:58] It's not like BugMessage:+index. [03:58] Er. [03:58] BugMessage.index [03:58] * lifeless shrugs [03:58] lifeless: I can see I've not convinced you, so meh [03:59] I would be most comfortable with a situation where we don't have partial migrations done [04:00] but its your call; you are doing the work. [04:01] As you would have seen on the list, I'm somewhat around this week, so I'm just going to not care until Friday [04:02] lifeless: I also seriously question having to perform this (very expensive) migration on four seperate instances [04:21] wow [04:21] my net is so farked [08:13] Project windmill build #184: STILL FAILING in 1 hr 0 min: https://lpci.wedontsleep.org/job/windmill/184/ [08:29] Project devel build #645: FAILURE in 5 hr 33 min: https://lpci.wedontsleep.org/job/devel/645/ [08:34] Sigh. That test needs a rewrite to not be racy. [20:15] Yippie, build fixed! [20:15] Project devel build #646: FIXED in 5 hr 34 min: https://lpci.wedontsleep.org/job/devel/646/ [20:28] Project windmill build #185: STILL FAILING in 1 hr 0 min: https://lpci.wedontsleep.org/job/windmill/185/ [20:50] morning [21:49] Has something changed in the ~vcs-imports owner recently? [21:49] I now get a rejection email from launchpad-bugs@ every time I change a vcs import [21:50] I think we'll nuke that team (-bugs) [23:14] hi all [23:56] thumper: hey [23:56] lifeless: hi [23:56] thumper: how does enum sorting work with queries [23:56] what do you mean? [23:57] thumper: I'm looking at making INCOMPLETE_WITH_RESPONSE and INCOMPLETE_WITHOUT_RESPONSE real status values in the db [23:57] bug 759467 [23:57] <_mup_> Bug #759467: incomplete-with-response searches require complex searches < https://launchpad.net/bugs/759467 > [23:57] see lib/lp/bugs/interfaces/bugtask.py and BugTaskStatus/BugTaskStatusSearch [23:57] ok... [23:58] so from my brief checks we just issue order by status [23:58] where status is the enum column [23:58] an int in the schema [23:58] and we have no special handling [23:58] *but* [23:58] BugTaskStatusSearch [23:58] has a sort_order attribute [23:58] and I'm lke 'wtf'