[02:10] <StevenK> wgrant: https://code.launchpad.net/~stevenk/launchpad/distroseries-spec-preload/+merge/180046 is updated, and it failed ec2 last night, but the only failures were related to builders and Fault 8002.
[02:12] <wgrant> StevenK: When last night did you start it?
[02:12] <wgrant> I thought I fixed the 8002s in the early afternoon.
[02:12] <wgrant> Hum
[02:12] <wgrant> No, 21:06
[02:12] <StevenK> wgrant: It finished at 9pm, so 5ish
[02:12] <wgrant> Right, so you had the bad lpbuildd
[02:12] <wgrant> So that's fine
[02:12] <wgrant> I'll review after lunch
[02:12] <wgrant> :)
[02:13] <StevenK> wgrant: Than microservices!!
[02:13] <wgrant> I'm melting strawberry atm.
[02:16] <StevenK> Poor strawberry
[04:08] <StevenK> wgrant: It's after lunch; good news, bad news, no news?
[04:54]  * StevenK takes that as 'no news'
[05:09] <wgrant> Soon soon
[05:09] <stub> StevenK: You can review my branches if you are bored ;)
[05:36] <wgrant> StevenK: I've done a bit of testing of buildd-manager on DF.
[05:36] <wgrant> Have you looked at all?
[05:36] <StevenK> I have not.
[05:36] <wgrant> I think it's probably OK to deploy, particularly now that the ARM buildds don't suck; slightly overzealous error handling isn't going to be a huge problem any more.
[05:37] <StevenK> The pandas are still around, just manualed, right?
[05:37] <wgrant> I believe they're inactived.
[05:37] <wgrant> But probably not actually switched off yet.
[05:37] <wgrant> Yeah, still not there on /builders.
[05:37] <StevenK> They seem to be gone, indeed
[05:37] <StevenK> We probably need a fitz, though
[05:39] <StevenK> wgrant: If the new stuff works fine with un-upgraded buildds, then I agree about the error handling, and we should deploy.
[05:45] <wgrant> If something goes terribly wrong we can always undeploy :)
[07:09] <StevenK> wgrant: WRT lines 125-127, the method involved is one that deletes workitems that aren't referenced any more and is only called by updateWorkItems. I wanted to be certain it grabbed whatever was latest. But I can change it to use the cached version if you wish.
[07:13] <wgrant> StevenK: I thought the one I pointed out was in workitems_text
[07:13] <wgrant> The one after was updateWorkItems
[07:13] <wgrant> But I may be wrong
[07:14] <StevenK> The one after is lines 134-137
[07:15] <StevenK> That one doesn't used the cached version since we just deleted non-matching workitems, and the old code duplicated the method anyway.
[07:15] <StevenK> And we blow the cache away at the end of the method anyway
[07:16] <StevenK> wgrant: workitems_text is lines 88-89 and 97-102
[07:16] <StevenK> And does make use of the cached property, for obvious reasons.
[07:16] <wgrant> Ah, you're right
[07:16] <wgrant> I was looking at the wrong part of the file.
[07:17] <StevenK> wgrant: Your objection is withdrawn?
[07:17] <wgrant> Indeed
[07:18] <StevenK> wgrant: I'm looking into your suggestion about rewriting the horrible double loop
[07:20] <StevenK> wgrant: Blah, I forgot about a comment I rewrote to be less like a novel, but: http://pastebin.ubuntu.com/6005501/
[07:21] <wgrant> StevenK: That makes a lot less nonsense.
[07:22] <StevenK> I didn't want to stuff around with going back from specid in work_items_by_spec.keys() to spec so I could get a cache -- if the spec id isn't in work_items_by_spec it returns [] anyway since it's a defaultdict.
[07:23] <wgrant> StevenK: Oh, since you're going through rows this time you can remove the code that presets the work_items cache to []
[07:24] <StevenK> Good point, killed.
[07:24] <StevenK> I shall push and land