[00:00] <cjwatson> I may be able to make it more robust, but it seems to more or less work.
[00:01] <cjwatson> Do you use -c then?
[00:07] <wgrant> cjwatson: I haven't used parallel tests locally in a while, but when I did I always used aufs2 to isolate the working dirs.
[00:08] <wgrant> I don't know that there's much state in them, but I'd be surprised if there were none.
[00:26] <wgrant> cjwatson: My other concern with the UUID change is that I can currently easily get into a running test's DB just by checking its PID.
[00:26] <wgrant> It may be worth the tradeoff if we don't need COW on the working tree, though.
[00:38] <cjwatson> wgrant: We could prepend the PID as well, presuming there's no particularly pressing length limit and that you could tab-complete the DB name.
[00:39] <cjwatson> (I'm not in a rush if it's controversial, can always apply it locally as needed)
[06:59] <wgrant> cjwatson: I wonder if Archive:+edit should hide checkboxes with impossible virtness.
[07:00] <wgrant> cjwatson: Currently it lets my unprivileged account enable ia64 and sparc, which won't do anything.
[07:46] <wgrant> Or I guess we could just restrict them, now, actually.
[08:23] <cjwatson> wgrant: I thought I did hide impossible architectures.  Are you testing this on prod?
[08:25] <wgrant> cjwatson: This was on dogfood.
[08:25] <wgrant> Running devel tip.
[08:26] <wgrant> Archive.available_processors doesn't check virtness at all.
[08:26] <wgrant> (if the widget still existed on +admin it would make sense for that variant to show them all, but it doesn't so that's sort of moot)
[08:28] <cjwatson> wgrant: Well, sure, but you won't see ia64 or sparc on prod because we hide arches that don't exist for any non-obsolete series.
[08:28] <wgrant> cjwatson: Oh, of course.
[08:28] <wgrant> That particular case indeed doesn't exist on prod.
[08:33] <cjwatson> It's true that we show non-virt-capable arches, but I thought it was overcomplicated to try to handle that - would've needed either a separate view on +admin, or JS on +edit to show the hypothetical unrestricted non-virt-only non-obsolete arches when you uncheck require_virtualized
[08:33] <wgrant> cjwatson: Well, +edit doesn't have the virt checkbox in the first place, does it?
[08:33] <wgrant> If the arch widget were still on +admin, it would need to be fancy or just skip the check.
[08:33] <cjwatson> oh, true, so  perhaps we could.
[08:33] <wgrant> But on +edit the virt status can't change.
[12:03]  * wgrant takes a hatchet to DatabaseLayer
[12:13] <cjwatson> Oh?
[12:14] <wgrant> It has ridiculous overhead.
[12:14] <wgrant> Per-test overhead, that is.
[12:14] <wgrant> Even in the fast path.
[12:16] <wgrant> 10 tests that just each create a person take a total of 1200ms to execute
[12:17] <StevenK> Welcome to Launchpad, I believe you've been here before?
[12:17] <wgrant> I haven't looked at this problem in more than 18 months.
[12:17] <wgrant> But I have enough wine that I can probably get through the test fallout this time.
[12:18] <StevenK> wgrant: Aiming for the Ballmer Peak?
[12:18] <wgrant> It's more that fixing tests that whine about ID differences is INCREDIBLY TEDIOUS.
[12:18] <StevenK> Ah yes
[13:40] <cjwatson> wgrant: I'm wondering what the path-like entry in the payload of merge proposal webhooks should be, and this makes me wonder whether we should consistently include a webservice link in other payloads.  (Perhaps as an addition; I can see how it's useful to have just the path sometimes.)
[13:41] <wgrant> cjwatson: We can't include a full link, as the version and root may vary.
[13:41] <wgrant> But we could include the force_local_path version, indeed.
[13:42] <cjwatson> Right.