[00:03] <jelmer_> thumper: I'm surprised you're seeing errors on subversion.conf rather than Sqlite locking errors
[00:04] <thumper> jelmer_: yeah, no idea really
[00:05] <jelmer_> thumper: anyway, it shouldn't be all that hard to fix, somebody just needs to take a couple of hours and JFDI
[00:05]  * thumper nods
[00:05]  * jelmer_ -> sleep
[06:03] <wgrant> thumper: Why can't you expose IBranchTarget?
[06:03] <wgrant> You can expose an interface without a URL -- the object's URL is used.
[06:11] <wgrant> Oh, unless objects don't really implement IBranchTarget, but adapt to it...
[06:41] <thumper> wgrant: branch targets are adapted to the branch target objects
[09:12] <jtv> danilos: yes, about permissions and flag-stealing policy...
[09:12] <danilos> jtv, go ahead
[09:13] <jtv> Did I remember correctly that the flag-stealing policy (how do I type that "O/+"!?) allows flag-stealing if (1) you're translating upstream or (2) you're translating in Ubuntu but with upstream privileges?
[09:13] <jtv> We documented what the policy does, but didn't go into much detail (tired as we both were after the sprint I guess!) about where it came from.
[09:14] <jtv> So I thought I'd better check.
[09:15] <jtv> The reason I'm checking is that translating Ubuntu with upstream privileges will probably AIUI still produce different effects from translating directly upstream, and I haven't given much thought at all to whether that's what we want.
[09:15] <danilos> jtv, so
[09:15] <danilos> jtv, it comes from "external policy flag" which we assume is set to "prefer upstream" right now
[09:16] <jtv> Plus your privileges, right?
[09:16] <danilos> jtv, plus privileges, of course
[09:17] <jtv> Ah yes, and then as a sort of optional extra in the design we talked about kyleN's scenario where upstream prefers Ubuntu rather than the other way around.
[09:17] <danilos> jtv, right, that bit was kind of symmetric as well except that we don't have a way to set that option yet
[09:17] <jtv> Someday we'll get around to thinking about how close we get to symmetry when you set both, but not today I hope :-)
[09:18] <danilos> "set both"? what do you mean?
[09:18] <jtv> When you make the Ubuntu translations follow upstream and vice versa.
[09:18] <danilos> jtv, so, privileges here have a bigger "weight" if the option to prefer "current context" is set
[09:19] <danilos> jtv, well, that's not really an option because well, you can't prefer ubuntu over upstream and upstream over ubuntu at the same time :)
[09:20] <jtv> Not at the same time, but you could prefer the current context when you're editing Ubuntu and then again prefer the current context when you're working upstream.  :)
[09:20] <jtv> But anyway
[09:21] <danilos> jtv, it'd be nice to investigate if this policy can include translation focus handling, but I'd definitely be a bit worried about that without going through the same exercise we went through in Recife
[09:21] <jtv> The sketch API I pushed Friday expresses the "weight" by passing a boolean for "translator is privileged upstream" to the Ubuntu translation method but not the converse to the upstream one.
[09:21] <jtv> Ah, translation focus, that's the other thing I wanted to ask about.
[09:23] <jtv> I've got some XXX's up on the wiki page.  Today I'm planning to outline how the share-upstream-focus-imports change affects things.
[09:24] <jtv> Unless you have it all figured out already and just need some time to write it down, I'll just update the wiki page and ask for feedback.
[09:25] <bigjools> wgrant: do you have any thoughts on extending the file copy check to binaries?
[09:26] <bigjools> oh and I just spotted where you're sitting :)
[09:26] <jtv> wgrant: duck!
[09:26] <wgrant> bigjools: Where are you? I've tracked down StevenK.
[09:26] <bigjools> you just looked at me
[09:26] <wgrant> Bah.
[09:26] <bigjools> haha
[09:26] <wgrant> It's too dark back there.
[09:27] <StevenK> Bwahaha
[09:27] <bigjools> I can almost hear you laughing StevenK
[09:27] <wgrant> Heh.
[09:27]  * StevenK resolves to not laugh
[09:28] <wgrant> bigjools: I don't think the file copy check is useful there.
[09:28] <bigjools> I mostly agree
[09:28] <wgrant> bigjools: Except for epochs, binaries don't conflict.
[09:28] <wgrant> Only (component)orig tarballs do.
[09:28] <wgrant> (epochs are a concern, but they're evil, so screw them)
[09:28] <bigjools> heh
[09:29] <StevenK> ARGH!
[09:29] <wgrant> Uhoh.
[09:29] <StevenK> HE SAID THE NAUGHTY WORD
[09:29] <wgrant> Heh.
[09:31] <wgrant> bigjools: So, our file data model sucks, and we have lots of bad data in production. What do we do?
[09:31] <bigjools> fix it (tm)
[09:32]  * StevenK tries to make a big destination archive to break copying
[09:32] <wgrant> mozilla-daily?
[09:32] <StevenK> I can't copy into that, though
[09:32] <StevenK> Well, not without hand-waving
[09:32] <bigjools> I just read: StevenK tries to ... break copying
[09:32] <wgrant> INSERT INTO TeamParticipation [...]...
[09:33] <bigjools> or go admin on DF :)
[09:33] <wgrant> That was my intention, yeah.
[09:33] <StevenK> I am wondering if it's a large source archive, or destination archive, or both
[09:34] <wgrant> Destination.
[09:34] <wgrant> The source archive doesn't matter at all.
[09:34] <wgrant> The key metric for the old code is number of SPPHs with the target name in the target archive.
[09:34] <wgrant> For the new code, it's the number of SPPHs in the target archive, but it's encapsulated in a DB query so it's not comparable.
[09:36]  * wgrant vanishes.
[10:04] <mwhudson> so ... why am i getting bucketloads of mail aimed at ~registry
[10:04] <thumper> blame someone on bugs I guess
[10:04] <thumper> I've had to create a mail filter
[10:04] <thumper> X-Launchpad-Message-Rationale contains @registry -> delete
[10:05]  * mwhudson tries to remember how thunderbirds message filters work
[10:16] <bigjools> jml: are you free by any chance?
[10:16] <jml> bigjools, I'm available.
[10:16] <jml> bigjools, as in speech
[10:16] <jml> bigjools, in Pomegranite downstairs.
[10:17] <bigjools> jml: could you spare me some time in Kawi or are you ensconsced down there?
[10:17] <jml> bigjools, sure. in a couple of minutes, poolie is laying out the bzr sprint
[10:17] <bigjools> it's about the sftp server
[10:17] <bigjools> jml: great, thanks
[11:12] <jml> bigjools, https://lpstats.canonical.com/graphs/CodehostingCrowberryConnections/ -- this graph shows the connections to the codehosting ssh server
[11:12] <jml> bigjools, you'll easily be able to see where we added the timeout :)
[11:16] <bigjools> jml: it's not that obvious actually!
[11:16] <jml> bigjools, https://lpstats.canonical.com/graphs/CodehostingCrowberryConnections/20081101/20100511/
[11:17] <bigjools> jml: ok that one is :D
[11:17] <jml> StevenK, what's your email address?
[11:17] <jml> :(
[11:17] <jml> mail server fail
[11:19] <StevenK> jml: stevenk@{ubuntu,canonical}.com ?
[11:20] <jml> bigjools, anyway, one hour timeout seems to work out just fine. I don't remember why it was so big though -- mwhudson might
[13:01] <wgrant> StevenK: Have you actually pushed your latest changes?
[13:01] <wgrant> Also, you should check out how it works on the primary archive .
[13:01] <wgrant> With multiple SHA1s.
[13:01] <StevenK> wgrant: Yes, I have.
[13:02] <StevenK> wgrant: Oh?
[13:02] <wgrant> StevenK: Because there is bad data.
[13:03] <wgrant> And you need to be able to deal with it.
[13:03] <wgrant> (or we need to fix it, but that may be impossible.)
[13:03] <StevenK> Personally, I think we should fix it.
[13:07] <StevenK> wgrant: Right, so we key off the filename, so one of the SHA1s will win.
[13:07] <StevenK> Just thinking about it
[13:08] <wgrant> Yes.
[13:08] <wgrant> I guess we should run a query and find out how much bad data there is.
[13:08] <wgrant> Someone with dogfood access could do that easily.
[14:19] <allenap> sinzui: Hi there. I see that there are 4 identical download links on https://edge.launchpad.net/lazr.config. Is that a known bug?
[14:20] <wgrant> allenap: That's correct. The file has been added four times.
[14:20]  * sinzui looks
[14:21] <allenap> wgrant: Ah, okay. Is that a bug in LP, or just a mistake?
[14:21] <sinzui> allenap, that is not a bug, just incompance
[14:21] <sinzui> ence
[14:21] <wgrant> allenap: A mistake.
[14:21] <sinzui> allenap, I deleted the duplicates
[14:22] <allenap> sinzui, wgrant: Cool :)
[15:29] <mwhudson> oh the irony
[15:29] <mwhudson> jelmer: did you see that the dulwich import is failing?
[15:30] <jelmer> mwhudson: yeah, bzr-git bug I need to fix
[15:42] <jelmer> mwhudson: the wonders of dogfooding ;-)
[15:48] <deryck> gmb, I've volunteered you for something at UDS. :-)
[15:48] <gmb> deryck, Oh gawd.
[15:48] <gmb> deryck, It's not inline upstream dupe searchign is it?
[15:48] <deryck> gmb, it's not, but nothing major.  Just scope and estimate a way to show what external trackers use our plugin(s).
[15:49] <gmb> deryck, Ah, cool.
[16:29] <mars> sinzui, ping, is Edwin around this week?
[16:29] <sinzui> mars, Edwin will be available tomorrow
[16:29] <mars> sinzui, ok, thanks.
[17:44] <wgrant> StevenK: Once you've finished fighting Twisted... did you end up checking what will happen if there are multiple SHA1s? It's probably going to non-deterministically fail. That is probably bad.
[17:45] <jml> StevenK, https://code.edge.launchpad.net/~jml/launchpad/poppy-sftp