[03:33] <wgrant> StevenK: https://oops.canonical.com/oops.py/?oopsid=OOPS-95fa6dd923c7434f893c96dbc692e5da
[03:36] <StevenK> It that going to make me very sad?
[03:39] <wgrant> Possibly :)
[03:41] <StevenK> wgrant: I was pondering what to grab next, since that checkwatches trace doesn't help much.
[03:49] <StevenK> wgrant: Reproduced on dev
[03:50] <wgrant> Great
[03:50] <wgrant> Should be an easy fix :)
[03:50]  * wgrant is stabbing rosetta
[03:51] <StevenK> Which part?
[03:51] <wgrant> +translate timeouts
[03:51] <wgrant> At least one variety of them
[03:51] <StevenK> Braver man than me
[03:51] <StevenK> I'm just concerned that Rosetta is a hydra
[03:51] <StevenK> Cut one head off and 30 grow back
[03:57] <wgrant> I know almost enough about it now to approach it soonish
[03:58] <wgrant> Where by "approach" I probably mean "spend two days trying to learn how it works"
[03:58] <StevenK> wgrant: "Badly" and "slowly"
[03:58] <wgrant> True
[03:59] <wgrant> But I think a few extra cols on TranslationMessage (and backfilling its 70M rows) should fix a few things
[03:59] <StevenK> Ah, denorming or some other plan?
[03:59] <wgrant> Yeah
[03:59] <wgrant> Mostly from potmsgset to translationmessage
[04:44]  * StevenK stabs the branch scanner
[04:56] <StevenK> wgrant: https://code.launchpad.net/~stevenk/launchpad/productseries-preload-for-merges/+merge/153052
[04:57] <wgrant> StevenK: Why'd you replace the _known_viewers set with a list?
[04:58] <StevenK> Why does it have to be a set, everywhere else just sets it to a list
[04:59] <wgrant> Nothing uses .add?
[04:59] <wgrant> eg. bugs certainly used to
[04:59]  * StevenK reverts that bit
[05:00] <wgrant> One day we will be on 2.7 and there'll be no excuse to use lists.
[05:01] <StevenK> wgrant: Is that your only objection?
[05:02] <StevenK> Not quite a revert, I've removed a spurious set of brackets
[05:02] <wgrant> StevenK: I'm wondering if you're preloading too late or something, since it apparently ends up accessing preloaded attributes before they're preloaded.
[05:03] <StevenK> wgrant: Such as?
[05:04] <wgrant> StevenK: That would only crash if the attribute it was attempting to precache had already been accessed *before* the preloading.
[05:04] <wgrant> Which is a potential flaw in the preloading, depending on how prevalent it is.
[05:04] <StevenK> wgrant: It would have been no such attribute if it hadn't been?
[05:05] <wgrant> StevenK: Right
[05:09] <StevenK> wgrant: Maybe DecoratedBranch's cleverness is biting us
[05:09] <StevenK> It has a cachedproperty for the same thing