[10:31] * tomwardill mutters at lxc and testr [10:31] getting there, it'll at least start a lxc now, but testr fails to find any tests [10:49] cjwatson: any idea how to debug testr failing to produce a list of tests? [10:55] Do you have a .testr.conf ? [10:56] I'd drop down a level at a time - next level down is probably "sudo /usr/local/bin/lp-setup-lxc-test --list" I think? [11:00] yep, that seems to work [11:00] (I've hacked it to use lxc-copy and lxc-attach rather than lxc-start-ephemeral) [11:00] which I suspect is the problem, but I can't work out why not [11:00] testr appears to be executing the command successfully, so I assume it's either not receiving or failing to parse the output somehow [11:02] * tomwardill -> lunch [15:02] okay, can someone `apt-get install python-testrepository` and then do `testr list-tests` in a LP checkout and see if it produces anything? [15:36] testr doesn' [15:36] doesn't appear to work in xenial [15:36] or at least, doesn't list any tests [15:57] Well, /bin/sh: 1: xvfb-run: not found [15:58] * SpecialK|Canon spins up a container [16:00] oh, yes [16:00] you will need to do that in a launchpad/rocketfuel container [16:05] pappacena or cjwatson might be in a positiont o do that faster :) [16:06] * pappacena doing [16:06] also empty output, let's see if I can remember how this works [16:07] empty for me too https://www.irccloud.com/pastebin/2PM4CxJJ/ [16:07] the docs imply that should produce a list, but don't actually state it [16:08] (or indeed how/where/what form the list would be) [16:08] You do get one on stdout if you run the underlying command [16:08] Let me poke a bit [16:09] yeah, that's the symptoms I'm getting [16:09] I want to say something like 'subunit V1 vs v2', but I've got nothing to back that up with [16:09] It is possible [16:09] Could try | subunit-1to2 or some such [16:10] Or actually [16:10] --subunit-v2 maybe [16:10] cjwatson: aha! [16:11] that did it [16:11] testrepository imports subunit.v2 if it exists [16:11] and changes behaviour [16:11] and I don't think it existed on precise [16:11] yep, that would do it [16:11] right, lets see if I can make everything else work now [16:11] * tomwardill undoes all the hacks [16:12] We'll need to be a little careful because the buildbot master will still be precise (for the time being) and thus won't understand v2 [16:13] For test subunit output [16:13] Feeding the output through subunit-2to1 for the time being will probably deal with that [16:13] right [16:13] that's the config I've got locally atm (precise master, xenial slave), so should be able to experiment with it [16:13] Ultimately moving to subunit v2 will probably be a good thing though; should fix some encapsulation bugs [16:14] s/slave/worker, sorry. Buildbot still has old references. [16:14] subunit v2 was invented exactly because v1 got out of sync really easily [16:15] And I introduced subunit v2 support to zope.testrunner when I did a bunch of work on it in 2018 [16:15] Mainly to massively simplify the LP fork [16:15] ah, nice [16:15] So it *should* work, but it's basically untried in an LP context [16:15] well, my cpu is definitely doing something [16:15] Because we couldn't really use it until we upgraded buildbot :) [16:16] yep, it's running tests, with terrible output and weird formatting [16:16] testrepository should hopefully be ingesting it all [16:16] looks like it's failing to start postgres correctly [16:17] but that's progress at least [16:30] it also never finishes the step, so that's a thing. [16:30] * tomwardill will look at that tomorrow, slightly early EOD for bicycle club [16:45] You know I think I'll take TypeError: 'object() takes no parameters' as a signal that I should EOD and try again tomorrow [16:54] lol, have a good evening :)