[04:03] <wgrant> cjwatson: BuildFarmJob mirrors relevant attributes of the implementations so there's one place to query for all types. The mixins handle that for you.
[12:06] <wgrant> cjwatson: Any other issues with the liveFS arc?
[12:08] <cjwatson> wgrant: Going pretty well at the moment.  I'm battling with traversal code.
[12:08] <cjwatson> Still need to write behaviour tests.
[12:08] <wgrant> Some complication with traversal, or just a part of the world you don't visit often?
[12:09] <cjwatson> Just the latter.
[12:09] <cjwatson> I think I should have something worth an initial review by the end of the week at this rate (I'm off this afternoon).
[12:10] <cjwatson> Also still need to write webservice-y methods for grabbing files out of the LiveFSBuild.
[12:12] <cjwatson> But (with the exception of whatever random traversal bug I have at the moment) I have a reasonable set of tests working, so over the hump.
[12:12] <wgrant> Great.
[12:12] <wgrant> Happy to help with the traversal bug, if you can't work it out.
[12:15] <cjwatson> Is there a way to get pdb to break in a traverse method called in the back end of an AppServerLayer webservice test?
[12:16] <cjwatson> It just hangs, so I'm obviously doing it wrong ...
[12:18] <wgrant> cjwatson: Heh, are you trying to view /builders?
[12:19] <cjwatson> No, /~owner/+livefs/blah
[12:19] <wgrant> That's probably affected too
[12:19] <wgrant> Hm.
[12:19] <wgrant> Not in the existing implementations.
[12:19] <cjwatson> Where blah is DISTRIBUTION/DISTROSERIES/LIVEFS so the traversal is a bit complicated to start with (I'm using a cut-down version of the branchnamespace thing)
[12:19] <wgrant> But the testbrowser respects meta refresh, so loading /builders or /builder/whatever will hang by default.
[12:20] <wgrant> Oh, you mean the pdb is hanging, not the test normally?
[12:20] <cjwatson> Right, the test normally is failing with a too-many-redirections error at the moment and I'm trying to debug it
[12:20] <wgrant> It's a subprocess in AppServerLayer, so that indeed won't work. It's probably easier to reproduce manually on the dev appserver.
[12:21] <cjwatson> hm, ok
[12:21] <cjwatson> or I guess I could do the FakeLaunchpadRequest thing and test just the traversal that way, come to think of it
[12:22] <cjwatson> but yeah, I'll bring up a dev appserver if that fails
[12:23] <wgrant> Right, you could do that, and I think we should have more tests like that, but for debugging it's easy to do it outside the test suite
[12:29] <cjwatson> If I can figure out how to log in ...
[12:29] <cjwatson> (a) make run only runs on port 8085, is that right? (b) +login says the openid provider is unavailable
[12:30] <wgrant> make run listens on several ports, but you should always access it through apache on 443
[12:30] <wgrant> If you don't have apache set up, +login will fail when it can't talk to https://testopenid.dev/
[12:30] <cjwatson> That gives me an ssl_error_rx_record_too_long error
[12:31] <wgrant> Sounds like you don't have mod_ssl enabled, or your LP vhost config is broken.
[12:32] <cjwatson> Oh, huh, no vhost
[12:32] <wgrant> If you're using LXC, make sure you've run "make LISTEN_ADDRESS=* install" and then restarted apache
[12:34] <cjwatson> Much better, thanks
[12:43] <cjwatson> No luck getting lp-shell to get an oauth token though.  Trying the more lightweight test approach
[12:44] <wgrant> Hm, should work fine with LP_DISABLE_SSL_CERTIFICATE_VALIDATIOn
[12:44] <cjwatson> (I copied and pasted the URL into firefox on my host, but the token doesn't seem to be recognised inside the LXC guest)
[12:44] <cjwatson> Yeah, I used that
[12:44] <wgrant> I usually run lp-shell on the host, but it shouldn't matter.
[12:47] <cjwatson> That seemed to get a bit further, but it got an ECONNREFUSED trying to send mail
[12:47] <wgrant> Ah yes
[12:47] <cjwatson> http://paste.ubuntu.com/7314405/
[12:47] <wgrant> That will break unless you have an MTA installed
[12:47] <wgrant> Install a local-only postfix in the container and restart lp-shell
[12:49] <cjwatson> Ah, brilliant, that works now
[12:55] <cjwatson> Just in time for me to EOD - will investigate more tomorrow
[12:56] <wgrant> :)