[12:08] bac benji frankban gmb call in 2 [12:09] rt [12:19] benji, Thank you kindly: https://code.launchpad.net/~gmb/launchpad/ebt-logging-bug-974586/+merge/101720 [12:21] sorry i got booted. trying to rejoin. [12:46] gary_poster: starting hangout [12:46] * bac tries again [12:50] gary_poster: g+ is having fits and won't let me join goldenhorde. i've sent you an invite [12:51] k bac [12:53] bac, "this hangout no longer exists" [12:53] that's a lie, a damned lie. [12:54] i'm in it. existing. [12:54] argh [12:54] :-) [12:54] gary_poster: ok, the normal hangout now let me in [12:55] bac, don't see you in https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhorde [12:55] google needs to spend more resources making hangouts work instead of making new whitespace [13:34] heh [13:35] frankban, https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhordeoneonone whenever you are ready [14:09] gary_poster: the newest version of juju solve the bootstrap --constraint problem [14:09] great bac [14:31] gmb, ready in 10 or less; still on with fb [14:32] k [14:36] gmb, https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhordeoneonone whenever ready [14:45] gary_poster: this works: http://pastebin.ubuntu.com/926506/ it includes at line 13 a work-around for a new issue with deploy & --constraints [14:45] cool bac will look soon [15:08] bac, so you already have everything up, or are you waiting on slave [15:08] & is the workaround "68.4"? [15:09] gary_poster: the workaround was to do it separately as shown at line 13 [15:10] oic [15:10] gary_poster: waiting on slave [15:10] cool bac. are you looking at 974617? if so will move card [15:11] gary_poster: not yet [15:11] btw frankban, when you are ready to work on a parallel testing card, please check with gmb to see if you can pair with him on it [15:11] sure gary_poster [15:12] thank you [15:12] bac, cool [15:15] gary_poster: working the ppa card first [15:15] gary_poster, I think bug 974608 comes under "AAAAAAAAH! Sampledata!" [15:15] <_mup_> Bug #974608: bugtask-find-similar.txt fails intemittently on paralleltest runs < https://launchpad.net/bugs/974608 > [15:17] gary_poster: I created a card for my incremental branch (lp:~benji/testtools/modernize-tsfr) and marked the 974622 card as blocked on my branch being reviewed (jml will do so today, UK time) [15:18] great benji thanks [15:18] gary_poster: and now the multi-branch work lane is over limit [15:18] benji, s'ok. it's a reasonable view on reality [15:18] gary_poster: should I speculatively continue on the 974622 work assuming the branch survives review or do something else? [15:19] benji, +1 on speculation IMO [15:19] k [15:19] agree? [15:19] I guess it shouldn't be "blocked" then [15:19] yeah. It will be blocked when you can't do anything on it [15:20] gmb: I am ready to pair on your nice bug, when you are [15:21] frankban, Ah, I think that the one I'm on now isn't suitable - it's a relatively quick fix, AFAICT. We might be better off picking one of the other cards and pairing tomorrow morning instead. [15:21] gmb: coll [15:21] hem cool [15:21] frankban, look, slack time! :-P [15:25] bac, I need to take lunch now (cat to vet). I'll ping when back [15:25] gary_poster: coincidence. i was about to go eat chinese food [15:26] (speaking of cats) [15:26] lol [15:26] heh [15:26] this past trip to VN was the first time i saw a restaurant specializing in "Thịt meo". [15:27] I have a suggestion for everyone here who want to create a programming language: please call it in a way you can easily find info using google: a neologism with len >= 4 is strongly appreciated [15:27] frankban: +1 [15:29] G+ needs a 'translate' button for comments [15:33] heh [15:33] frankban, I find "golang" works ok [15:34] gary_poster: coll didn't know, I was using "go language" [16:04] * gmb -> very late lunch; back later [16:57] benji, I saw discussion in -dev with jml. What he said about the purpose of the forwarding result makes sense (where each test needs to be isolated, not just each call), but I thought that the semaphore was handled explicitly so that would work fine. As I thought about it more, though, I realized I can't imagine a truly sane story for global tags within a merged set of test streams [16:58] the expectations for the meaning of global tags would have to be drastically reduced [16:58] though, I'm not really sure what the use cases are for global tags anyway [16:58] stepping away; back again soon [17:25] gary_poster: (I was eating lunch.) right, I don't think there is a general story that will work with tags without a great deal of development effort, however if we are specific about what TSFR needs to do and what it doesn't do, I think we can get something that continues to work. The funny part is that it is less and less important to us for our worker ID tagging result wrapper. [17:27] benji, less and less important: it was only importat initially as it helped us understand what was going on, and you were using it as a template of sorts, right? [17:27] bac, lemme know when you are around and we can pair [17:27] gary_poster: i'm here. [17:27] right; at this point we could just abandon any work on it and go ahead with what we want [17:28] well, not quite: I'm pretty sure as-is TSFW is broken with respect to tags, so we'll have to do *something* about it, but we can more clearly see what that something is once it bites us [17:30] benji, if we are sure of a fix, I'd be in favor of applying it. Otherwise (for instance, the undoing of the global tags) I'm fine with putting it aside. [17:31] https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhorde [17:31] bac ^ [17:32] gary_poster: I'm not sure of the fix, but I'm relativly sure there is a problem. I'm afraid this is one of those things that cross package boundries and will be hard to get just right. [17:37] gary_poster: I'm trying to figure out how to move the cards around to represent the current state of the world: I intend on 1) writing an add-worker-id-tag result wrapper, 2) integrating it into ConcurrentTestSuite, 3) fixing the tag bugs I believe we will uncover in TSFR. [17:38] 1 and 2 seem to be covered under the card for 974622 and 3 is the other card I created earlier; should they both be in Active/Coding? [17:41] benji, have you started work on 3? [17:41] gary_poster: I thought I had, but now I see that I haven't really. Take that for what it is. :) [17:42] benji, heh [17:42] s/it is/you will. [17:42] so I guess it's reasonable to move the card for 3 into the started-but-threw-it-away bucket and make a new card if we need it [17:43] ok, benji, I moved that card (#3) back to "Tasks" and changed the title to be less confident ("ThreadSafeForwardingResult doesn't handle tags as well as it could. We suspect.") [17:43] sounds good [17:44] benji is it also appropriate to mark 974622 s unblocked [17:44] as [17:44] gary_poster: oh, on a different note, a nurse is going to be coming by around 2 (they are habitually late) to teach me how to change Katie's catheter dressing for when we're in TN, so that may spoil our one-on-one time [17:44] yep, it is unblocked [17:45] benji, ok, let me know when/if you are available [17:45] and benji, cool thanks, i marked card as unblocked [18:02] gary_poster: i now have master and slave coming up. slave is verified to be m2.4xlarge [18:03] bac, excellent [18:25] gary_poster: the nurse is already 25 minutes late, what say we start our call? [18:26] benji, oh ok! I was just going to reboot. I am in https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhordeoneonone [19:10] gary_poster: master and slave up and ready after 65 minutes. i've manually done work-around for bug 974584 and have forced a build [19:10] <_mup_> Bug #974584: Semaphores cannot be created in lxc container < https://launchpad.net/bugs/974584 > [19:10] yay, bac! [19:30] gary_poster: chat? [19:31] bac, still talking :-P ready in 5 or so ok? [19:31] ok [19:31] thx [19:37] bac, 4 [19:37] bah [19:37] bac, https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhordeoneonone [20:17] benji, do you agree that the following buildbot waterfall indicates that the "teach buildbot to understand subunit in test results to properly report failure numbers in waterfall" card has a bug or problem of some sort? http://ec2-50-19-152-175.compute-1.amazonaws.com:8010/waterfall [20:18] * benji looks [20:18] benji, would it be worth asking bac to verify that [something or other that you describe] is in the buildbot.cfg of the master? [20:20] gary_poster: on the first point, the "no test results" looks discouraging, as I understand it there should be a summary line there [20:20] right, so we're on the same page there [20:21] gary_poster: on the second part, if I recall correctly, no .cfg change should be required, I just changed the bzrbuildbot Test class to report in the new way. Checking that the bzrbuildbot/subunittest.py module exists would be a good smoke test [20:22] benji, ok cool. bac, could you look for that on the master when you get a chance? [20:24] benji: -rw-r--r-- 1 buildbot nogroup 3255 Apr 12 18:05 /var/lib/buildbot/masters/master/bzrbuildbot/subunittest.py [20:24] that what you want to see? [20:24] bac: yep, looks good [20:25] why we're not seeing a meaningful summary is a mystery to me [20:30] benji, dragged that card back to active coding but also pulled your head off it. We need to get it resolved but it doesn't have to be you [20:30] k [20:30] I'll send an email to yellow inviting gmb and frankban to look at it tomorrow if they want [20:31] my inclination, for whoever looks at it... well, I'll reply to your email with my inclination [20:31] :-) k [20:32] bac, is it appropriate for me to drag card for bug 974617 to done done? [20:32] <_mup_> Bug #974617: test_operational_view_integration fails intermittently in parallel tests < https://launchpad.net/bugs/974617 > [20:32] yes [20:32] cool thx [20:37] bac, could you write an email to the yellow list with a description of how you successfully started up your instances today, please? [20:37] benji, I just sent the email, but who knows when it will actually make it through the mailing list. [20:37] ok [20:38] thank you [20:38] gary_poster: inside the container: [20:38] buildbot@lptests:~$ bzr whoami [20:38] Launchpad PQM [20:38] cool [20:38] this is *before* i did anything...so not so cool [20:38] ...oh... [20:38] even so, the celery test fails [20:38] so that's pretty confusing [20:38] with the same error? [20:38] yes [20:38] huh [20:38] ok [20:39] bac, I'll try now to dupe in a local lucid lxc container... [20:39] when you ran the test locally you were not in an lxc, right? [20:43] correct [20:46] this is odd, gary_poster [20:46] http://paste.ubuntu.com/927024/ [20:46] this is inside an lxc container on the slave on ec2 [20:47] so the cmdline 'bzr whoami' gives results but bzrlib.config doesn't [20:48] and if i manually set 'bzr whoami', then bzrlib.config *still* returns None, None [20:48] bac, what about the thing that calls _auto_user_id, though [20:49] bac config.username() [20:51] gary_poster: harder to test as you have to instantiate a Config subclass object [20:51] bac, ah ok. but bac, this fails for me on lxc [20:51] so some kind of interaction [20:51] running the test again [20:52] same same [20:53] bac, doing a pdb... [20:54] well at least it is easy to reproduce... [20:54] y [21:00] bac, could you set a pdb in your local environment and see what self.file_name is in the username method when you run that test? [21:01] can do it tomorrow [21:01] EoD [21:01] ok [21:01] I get /tmp/tmpOmHAkg/.bazaar/bazaar.conf [21:02] which looks like it is a nice test isolation thing [21:02] and then... [21:03] _get_default_mail_domain() returns None in _auto_user_id [21:03] which causes everything to fall over [21:04] gary_poster: so it smells like an lxc isolation problem [21:04] /etc/mailname is not installed [21:04] mm [21:04] smells like an "unspoken dependency" problem to me with that [21:04] I don't know who is supposed to define /etc/mailname [21:05] but I bet it is from a package that we don't install [21:05] because we shouldn't have to [21:05] could be wrong [21:05] ok running [21:05] bye all [21:13] setting /etc/mailname to 'example.com' allows the tests to fail differently with a TimeoutError