[12:07] <gary_poster> may be slightly late, sorry
[12:07] <gary_poster> aiming for :12
[12:12] <gary_poster> not yet :-( :15
[12:16] <gary_poster> gmb now sorry
[12:33]  * gmb -> lunch
[12:56] <benji> bac: your card now has room
[12:56] <bac> benji: thanks
[12:57] <benji> my pleasure
[13:06] <gary_poster> benji, I think I'm going to restart post update, even though not much changed, just for amusement, & then I'll be ready.  Are you reviewing or ready?
[13:07] <benji> gary_poster: reviewing, about ready
[13:07] <gary_poster> cool
[13:07] <gary_poster> brb
[13:12] <benji> gary_poster: done reviewing, getting coffee and preparing
[13:13] <gary_poster> cool benji, I'm in https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhorde
[13:22] <benji> gary_poster: I don't seeee youuu
[13:56] <bac> benji, gary_poster: if PGBouncerFixture is made to not do anything in start() then the subsequent error is indeed a 503
[13:56] <gary_poster> bac, huh
[13:56] <benji> good
[13:56] <gary_poster> ok, well, that means there's still a chance our hypothesis is on track, so that's good...maybe :-)
[13:56] <bac> yes, perhaps
[14:51] <gmb> Argh, I'm as think as two short planks, me.
[14:51] <gmb> *thick.
[14:51] <gmb> See, I can't even insult myself.
[14:56] <bac> gary_poster: i've grabbed the python-pgbouncer project and am trying to build it.  bin/buildout complains about not being able to find packages that are clearly available, e.g.
[14:56] <bac> Error: Couldn't find a distribution for 'z3c.recipe.scripts==1.0.1'.
[14:57] <bac> what's the fix?
[14:58] <gary_poster> bac, not enough info, but at a guess, project's buildbot.cfg is constraining you somehow.  Maybe try linking lp's download-cache to the local dir?
[14:58] <gary_poster> bac, if not I'll get a branch and look
[14:59] <bac> gary_poster: ok.  i'd just created a local download-cache
[14:59] <gary_poster> cool.
[15:00] <bac> gary_poster: gah, install-from-cache in buildout.cfg
[15:00] <gary_poster> bac, so clear what to do now?
[15:01] <bac> yep
[15:10] <gary_poster> cool
[15:26] <frankban> gary_poster: I was taking a look at serge's lxc tests (here: http://bazaar.launchpad.net/~serge-hallyn/+junk/lxc-test/view/head:/lxctest.py)
[15:27] <frankban> gary_poster: basically it creates, starts and then destroys a container during a test run
[15:27] <gary_poster> frankban, should I look at them in preparation for our discussion?  I'll assume no until it seems otherwise
[15:27] <gary_poster> :-)
[15:28] <frankban> gary_poster: no :-)
[15:28] <gary_poster> :-) k thanks
[15:29]  * gary_poster listens
[15:30] <frankban> gary_poster: I am wondering if I can go on the same path, maybe creating a zope test layer that creates and starts an lxc in setUp and destroys it in tearDown: this way you can also be able to run normal (fast) tests excluding that layer... (if I understand correctly zope layers)...
[15:36] <gary_poster> frankban, lifeless will tell you to use testresources.  jml might tell you that testresources are designed much better but used and hardened much less.  Promoting zope layers anywhere in Canonical will probably get you unhappy people; promoting testresources (https://launchpad.net/testresources) will at least get robert and maybe jml on your side
[15:43] <frankban> gary_poster: thanks, taking a look at testresources now, I considered zope layers just because zope.testrunner is already used in the project
[15:43] <gary_poster> frankban, zope.testrunner already used by lxc?
[15:44] <frankban> gary_poster: no, by lpsetup
[15:44] <gary_poster> frankban, OIC
[15:44] <gary_poster> frankban, +1 to switch to nose fwiw
[15:45] <gary_poster> but frankban that's on the basis of statstics not personal experience :-P
[15:46] <frankban> gary_poster: yes I see that nose is very trendy lately ;-)
[15:49] <gary_poster> :-)
[15:49] <gary_poster> zopr.* is the opposite of trendy :-)
[15:49] <gary_poster> zope.*
[15:55] <bac> frankban: in setuplxc we have "repository = LP_REPOSITORIES[1] if valid_ssh_keys else LP_REPOSITORIES[0]" to chose between the http url for the lp branch and lp:launchpad.  why not just use http regardless?
[15:55] <gary_poster> bac, because it was going to be a developer tool
[15:55] <bac> i don't grok
[15:55] <gary_poster> now lpsetup will need to be both
[15:56] <gary_poster> bac, having a writable branch can be convenient for a developer
[15:56] <gary_poster> buildbot does not care
[15:56] <bac> gary_poster: you mean a pushable branch?
[15:56] <gary_poster> benji, I updated https://docs.google.com/a/canonical.com/document/d/1msejam1-AliaJ55lVEYayfWlmHTetjGheF3Tzjm9evA/edit with a story for how to get the build working, and an alternate easier but worse long-term story at the end
[15:56] <gary_poster> bac y
[15:57]  * gary_poster lunches
[16:52]  * benji wonders if gary_poster is eating lunch now.
[17:05] <gary_poster> benji, no :-)
[17:05] <benji> gary_poster: as-in "just finished" or "not even started"?
[17:07] <gary_poster> benji, as in finished a bit ago.  :-) ready to hang out if you are
[17:07] <benji> gary_poster: sure, after I finish a review
[17:07] <gary_poster> cool
[17:14] <benji> gary_poster: golden horde?
[17:14] <gary_poster> benji, y, I await you there
[18:57] <bac> hi gary_poster, i've got a buildbot launched and happy.  i'd like for it to test my branch not devel.  i thought i could just change the URL in master.cfg for schedulers to point to my branch on LP but that didn't work.  ideas?
[18:59] <gary_poster> bac, yeah that's the right idea.  You have to change it twice in master.cfg IIRC (I might not RC); moreover, you do need to stop and start the master for the changes to take effect.
[18:59] <bac> oh, well i did stop it
[18:59] <gary_poster> (they have a reconfigure command but it was buggy when we tried it, at least for the version of buildbot we run)
[18:59] <gary_poster> ah ha, but did you start it?! ;-)
[19:00] <bac> gary_poster: so you think i must change CHANGESOURCES and SCHEDULERS
[19:00] <bac> and then stop/start master and stop/start slave?
[19:00] <gary_poster> bac, all of that except the slave part
[19:04] <gary_poster> benji, exact same failure :-( https://launchpadlibrarian.net/102920431/buildlog.txt.gz
[19:05]  * gary_poster takes a break for a sec
[19:06] <benji> hmmmm
[20:02] <bac> benji: my buildbot test is showing the 'bzr locale' errors again.  your fix from earlier this month should've put those to bed, no?  http://ec2-23-22-21-234.compute-1.amazonaws.com:8010/builders/lucid_lp/builds/4/steps/shell_9/logs/stdio
[20:04] <benji> bac: I think it was frankban's fix actually, but I think they only applied to one or two commands.  I'm not sure which ones.
[20:05] <bac> benji: i was referring to https://code.launchpad.net/~benji/launchpad/bug-972456/+merge/100645
[20:05]  * benji looks
[20:15] <gary_poster> The byword for the parallel test runs is "stability."  That, and "reliability."  "Predictability" comes to mind as well.
[20:17] <gary_poster> benji, subunit failed again, yes?
[20:17]  * benji looks.
[20:17] <gary_poster> I think I'd like a build expert to help. :-/
[20:17] <benji> yes
[20:18] <gary_poster> benji, yes, it failed; yes, a build expert would be nice; or both? :-)
[20:18] <benji> both (retroactively)
[20:20] <benji> another thought, where does python3-testtools come from?  do we have it?
[20:20] <benji> in fact, are we sure which of the long list of dependencies wasn't met?
[20:26] <benji> gary_poster: are recruiting someone or shall I?  (and look above for non-gary_poster-containing remarks)
[20:32] <gary_poster> benji, I'll try recruiting
[20:32] <gary_poster> no idea for rest
[20:33] <gary_poster> benji, unless you will recruit in after-hours?
[20:33] <benji> I can't make heads or tails of it
[20:33] <benji> gary_poster: I didn't get that last bit about after-hours
[20:34] <gary_poster> benji, you were able to talk to wgrant after your EoD--on your weekend I think?  My intent would be to try and recruit someone within our normal working hours--maybe wgrant in our morning, or...something
[20:34] <benji> ah! gotcha! :)
[20:36] <benji> if we don't have anyone by the start of our morning tomorrow, I'll hit up wgrant and see what he says
[20:37] <benji> hmm, we /could/ try to build it locally so we can debug in a more hands-on way
[20:38] <gary_poster> maybe so
[20:38] <gary_poster> in a container or vm
[20:41] <benji> actually we can build it from the recipe, I'm trying now (on my shared term if you're feeling voyeuristic)
[20:42] <benji> gary_poster: it worked locally, so that suggests to me that it is a PPA issue; let me try adding the yellow PPA to my machine and trying again
[20:42] <gary_poster> ok
[20:46] <benji> gary_poster: no dice, it failed, but in a way that suggests to me that it should succeed in the PPA (at least my other build succeded in PPA when it failed locally like this one did)
[20:46] <gary_poster> heh
[21:03] <gary_poster> benji, https://code.launchpad.net/~yellow/+recipe/subunit-daily ...have we gotten as far as "Successful build" before?
[21:03]  * benji looks
[21:03] <benji> gary_poster: I don't think so.
[21:04] <gary_poster> I added the testing-cabal's ppa as a dependency of the yellow ppa, benji
[21:04] <benji> ah!
[21:04] <gary_poster> that was an idea jelmer had had awhile ago, but other things we did were supposed to make that unnecessary
[21:04] <benji> I wonder if it was python3-testtools that did it
[21:04] <gary_poster> :-/ maybe
[21:05] <benji> gary_poster: I'll see you tomorrow, it's off to the farrier for some new van shoes.
[21:05] <gary_poster> :-) ok bye benji
[21:23] <gary_poster> benji built
[21:23] <gary_poster> night
[21:29] <benji> gary_poster: cool!  (no van farrier for me tonight, my old blue car has a dead battery in retaliation for not being driven in two weeks)