[01:16] <wgrant> StevenK: Ah
[01:16] <wgrant> StevenK: You broke the build twice.
[01:50] <StevenK> wgrant: I did notice, yes.
[01:50] <StevenK> One of them I'm annoyed about, since that test failed in ec2 and then locally
[01:51] <StevenK> However, it also seems we have a smoking gun for the postgres connection leak. Maybe.
[02:04] <wgrant> StevenK: The postgres leak first showed up in r15502, 68 revs before this reliable failure.
[02:05] <wgrant> But the immediate issue is fixed.
[02:06] <StevenK> wgrant: Ah, nice.
[02:06] <StevenK> How does that deal with ClassAlias? It doesn't?
[02:07] <wgrant> StevenK: __storm_table__ on a ClassAlias is correct.
[02:07] <wgrant> So it should be fine.
[02:07] <wgrant> We'll find out for sure in 5.5 hours
[02:10] <StevenK> wgrant: The Armor of Al'tair is mine, btw
[05:18] <rigel_> Hello, I saw the hiring notice on the Launchpad blog, and I noticed it mentioned having an interest in Haskell. I was wondering if Canonical have plans to use Haskell for any projects in the future. Thanks!
[08:44] <cjwatson> hm, qastaging looks unhappy
[08:45] <cjwatson> (503 for the last hour or so)
[08:46] <StevenK> That's not good, because that is roughly the time the latest revno would have hit it.
[09:14] <wgrant> cjwatson, StevenK: It's probably due to a bug in the new staging restore procedure.
[09:14] <wgrant> It's meant to take staging down for a day or so, but not qastaging.
[09:14] <wgrant> fg
[09:14] <wgrant> fg
[09:22] <cjwatson> StevenK: I should've said "at least the last hour or so" - I have no information about before that
[10:59] <czajkowski> aloha
[11:01] <jelmer> kai ora
[11:02] <nigelb> jelmer: Isn't it kia ora? :)
[11:02] <jelmer> nigelb: euhm, yes, thanks :)
[11:03] <nigelb> :D
[11:05] <czajkowski> nigelb: pedantic :)
[11:05] <czajkowski> jelmer: ello hows you
[11:05] <nigelb> Hah :)
[11:09] <jelmer> czajkowski: all's well :) how's your weekend?
[11:13] <czajkowski> not bad, got into a tidying mode yesterday and redid the spare room so all the shuttle boxes are now labeled and I can now see the drobo box and router are off the ground
[11:14] <czajkowski> am going to attempt to tackle the wiring later on that is currently glaring at me once I come back from lunch
[11:40] <jelmer> czajkowski: sounds awfully well organized
[11:42] <czajkowski> jelmer: the spare room was a dumping ground for new gadgets, boxes servers and bf bike gear, there needed to be some order :)
[12:33] <StevenK> czajkowski: Ew, Drobo
[23:23] <cjwatson> Anyone fancy reviewing https://code.launchpad.net/~cjwatson/launchpad/queue-api-fix-urls/+merge/113776?  Last piece needed to make the API queue client basically usable.
[23:24] <cjwatson> (There's one bug remaining that shows up if you try to override just the section of a source upload; workaroundable in the client and easy to fix anyway.)
[23:30] <wgrant> cjwatson: That won't work
[23:30] <wgrant> cjwatson: It'll break the security team, since it won't work for private source.s
[23:31] <wgrant> sourceFileUrls was originally introduced for generating USNs.
[23:32] <cjwatson> I introduced sourceFileUrls.
[23:33] <wgrant> Ah yeah, different object.
[23:33] <cjwatson> What's a working way to do this?  Proxying through Archive definitely doesn't work ...
[23:34] <cjwatson> And it looked to me as though .getURL() worked for private URLs, but perhaps I was misreading
[23:35] <wgrant> Oh, getURL will work, yes. Thought you were still using http_url
[23:35] <cjwatson> (It starts "if not self.restricted:")
[23:35] <wgrant> But that's fairly evil
[23:36] <cjwatson> Because it hands out private URLs that can be passed on to anyone?
[23:38] <wgrant> Sort of
[23:38] <wgrant> We generally try to avoid issuing TLTs unless someone explicitly asks for them.
[23:38] <lifeless> (A TLT requires a DB commit)
[23:39] <lifeless> TLT's are the right thing to do to hand out urls into the librarian for private objects.
[23:39] <cjwatson> I guess an alternative would be to invent a thing to proxy URLs through the PackageUpload object
[23:39] <lifeless> But - unless you're very sure every script calling this will download the files, better to do what e.g. bugs do
[23:39] <wgrant> lifeless: In every single other case we redirect through the webapp.
[23:40] <lifeless> which is what wgrant just said :P
[23:40] <lifeless> wgrant: I thought, perhaps wrongly, that the API has a codepath to invoke the same thing
[23:40] <wgrant> No.
[23:40] <lifeless> huh
[23:41] <lifeless> so its uglier than even I appreciated ;)
[23:46] <cjwatson> So near and yet so far.