[12:06] buildbot-slave is working once you change the sources to the standard precise ones (from the ec2 ones) and update/upgrade so you get a newer lxc than 45, but buildbot-master fails because of a syntax error in lpbuildbot that Brad and I missed ("descriptionDone['cleaned','previous','lxc','containers']))" should be "descriptionDone=['cleaned','previous','lxc','containers']))". Fixing and then continuing to see how we are doi [12:06] ng. [12:08] bac benji gmb call in 2 https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhorde [12:11] bac benji yoo hoo [12:28] gmb, confirmed that I just approved next Mon & Tues [12:28] gary_poster: Thankyou kindly. [12:29] of course [12:29] notifications work a lot better on one screen. sound would be nice too though; I wonder why XChat stopped doing what I had configured...I'll look [12:39] well, sound is working again...who knows for how long [12:49] gary_poster: i've got a potential mentor for frankban but he doesn't want to start until after the 12.04 release...which is not that far off [12:50] bac, great! fine with me. If we have someone coming along sooner, great, but as you say, the release is not far off [12:51] yep. it is raphael and i think he'd be great [13:02] oh, agree, cool [13:18] juju expose didn't work on master. had to do it manually. Mildly concerning. [14:20] gary_poster: thanks for the details of your setup [14:20] bac, welcome [14:26] benji, bac, gmb, TESTS PASSED, GREEN BUILDBOT! *\o/* http://ec2-50-16-120-159.compute-1.amazonaws.com:8010/waterfall [14:27] will run again [14:27] gary_poster: great! [14:27] A++++, will run again [14:27] Woot. [14:27] 34 minutes? [14:28] bac: pretty much: "Elapsed38 mins, 36 secs" [14:28] Wow. That seems... short. [14:29] benji: oh, i saw this "elapsedTime=2075.273179" which must omit some startup [14:31] actually, I've been using the elapsed time, from http://ec2-50-16-120-159.compute-1.amazonaws.com:8010/builders/lucid_lp/builds/0/steps/shell_9 in this case, so 34 minutes is true. that is a bit suspicious, I must admit--it cuts down 20 minutes from our previous runs [14:32] the build is already gone so I can't see what happened last time, but I'll look after this next run [14:53] call off the parties...next run took 14 minutes and the testr directory looks...wrong somehow [14:54] * benji takes down the disco ball. [14:59] :-) [15:00] testr's reporting leaves a lot to be desired [15:17] gary_poster: It's not something to do with testr's splitting up of the tests per-core, is it? [15:17] Or are we futzing that somehow? [15:23] gmb, you mean, the reporting has something to do with testr's splitting of tests? I don't think so. We aren't futzing with that part of things. What I thought I understood before about what testr does seems to show nothing now. [15:23] the test run file now only seems to show a start time and a stop time [15:23] gary_poster: No, I mean the short runtimes. [15:24] Actually, that makes no sense. Ignore me gary_poster . [15:25] gmb, oh, well, I had that analysis earlier from runs that both frankban and I made that I was getting test run times of 4 hours/number of cores plus 20 minutes [15:25] so eight cores was 50-ish minutes [15:25] not 38 and not 14 [15:25] Right. Like I said, my thinking didn't actually make sense :) [15:26] :-) gotcha, np [15:53] Something has gone seriously wrong in our setup :-( runs are not producing output, and now lxc are stopping because they are unable to access dhclient.leases in the container...and maybe for other reasons? [15:54] ugh :-( [16:02] I'm going to kill this instance and try a newer version of precise. Maybe the update hosed things somehow? :-/ [17:45] our entropy hack is not installed? [17:46] ah. I don't have the newest version of the setup file. :-( [17:53] gary_poster: have time for a quick chat in a bit? [17:57] bac, sure [17:58] bac, https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhorde when you are ready [18:00] gary_poster: i'm trying but google is caught in an oauth redirection loop and i can't get in. let me try firefox [18:00] ok [19:53] gary_poster: i'm trying to test my change for testr init. i've not run a launchpad buildbot on ec2 in a really long time, if ever (my testing originally was just using pyflakes). does the slave typically show pending for the length of time setuplxc runs? [19:55] bac, yes [19:56] bac, I've discovered the problem for test runs btw [19:56] gary_poster: so, a couple of hours of churning? [19:56] bac, on the 8 core, about an hour IIRC [19:56] gary_poster: ugh, i did a single core [19:56] mm, probably 2 or 3 then yeah [19:56] i figured i didn't need to see the tests run to completion but just get started [19:57] gary_poster: what was the bug? [19:58] bac, the new zope.testing egg we made no longer gives the right output for --subunit --list-tests (cue my previous email saying how forking means you need to write tests...) [19:58] If you switch to the p4 version then it works [19:58] ergh [19:58] but without the p5 fixes of course [19:58] yeah [19:58] so we need a p6 :-) [20:36] If anyone would like to review https://code.launchpad.net/~benji/launchpad/bug-963463/+merge/100514 I would appreciate it. [20:47] benji: i will [20:48] bac: thanks [20:50] benji: only question is whether a helper function would be warranted since you repeat yourself thrice [20:51] bac: good question; I could go either way. I think I'll add one. [20:52] you are Dr Dry [20:53] benji: done [20:55] heh [21:02] * bac walks the dog