[05:42] hello [05:42] https://bugs.launchpad.net/juju-core/+bug/1038296/comments/5 [05:42] ^ tricky one to chase down [05:57] davecheney, heyhey [05:58] davecheney, ahh, well caught [05:59] fixing it might be harder [05:59] i've just proposed a small patch that adds the changes to the test to get that debugging [06:00] this will have to be fixed sharpish, because while it happens infreqently in dummy [06:00] the race will happen almost 100% of the time on ec2 [06:03] ouch! [06:04] actually I shold test that before throwing off wild assertions [06:04] morning [06:04] fscking launchpad, why can't I add a bug from the milestone screen !! [06:06] davecheney: *lol* e'body loves this tool [06:06] https://bugs.launchpad.net/juju-core/+bug/1042066 [06:06] ^ thoughts [06:11] davecheney: sounds reasonable. but i would like it configurable [06:11] davecheney: one mode workes like today, to test functionality [06:12] davecheney: and the other one more randomly to behave like in the wild [06:12] TheMue: sure, we can do that [06:15] shit, https://bugs.launchpad.net/juju-core/+bug/1038296/comments/6 [06:15] turns out, our recovery logic works, but this isn't awesome [06:44] Well, it's almost 3am here, so I'm going to crash for a bit, see you all in the morning [06:46] cu, and have a good night === TheMue_ is now known as TheMue [08:12] right - that is enough bugs logged for tonight [08:12] gotta get some sleep so I can fix 'em tomorrowq [08:13] davecheney: enjoy your evening [12:44] Hello all [12:58] niemeyer: hiya [13:01] TheMue: Yo [13:08] fwereade_: ping [13:08] niemeyer, pong [13:08] fwereade_: Yo [13:09] niemeyer, sorry I missed you on friday [13:09] fwereade_: Oh, n[p [13:29] rogpeppe: Heya [13:29] yo! [13:29] public holiday today BTW, i'm not here :-) [13:29] rogpeppe: Just reviewed the test refactoring branch.. nice stuff! [13:30] rogpeppe: Ah, cool :-) [13:30] niemeyer: cool, thanks [13:31] niemeyer: did you think of a neater way to do what i was trying to do with gocheck? [13:31] niemeyer: i think the reassurance it gives is very useful - it's so easy to break things in subtle ways. [13:32] rogpeppe: Agreed re. reassurance, and no I haven't looked yet [13:33] niemeyer: np [13:46] fwereade_: Is it a holiday for you as well? [13:46] mramm: Morning [13:56] niemeyer: Morning [14:28] niemeyer: am spending the afternoon putting all our wedding photos online, and wondering about using juju to host it, for the crack - do you know of any decent open source photo website s/w? [14:30] rogpeppe: Neat! [14:30] rogpeppe: I used to use Gallery for that [14:31] rogpeppe: http://gallery.menalto.com/ [14:32] niemeyer: cheers! not entirely sure whether using juju is a good idea, as no storage management yet... we'll see. [14:32] rogpeppe: It's somewhat easy to workaround that [14:33] rogpeppe: It's actually neat as it forces you to think through, rather than just forgetting the data in a machine somewhere [14:33] niemeyer: what i don't want to do is spend hours uploading GB of photos and then lose 'em... [14:33] rogpeppe: Right [14:37] rogpeppe: FWIW, we use EBS for the disk on all instances, so this shouldn't happen in generla [14:37] rogpeppe: Then, I'd suggest using S3 as a back-and-forth media for quick re-deployments [14:38] niemeyer: ah, makes sense. [14:38] niemeyer: so have a big tar file in S3 and suck that down when starting a new unit. [14:38] rogpeppe: Yeah [14:39] rogpeppe: and via config you can easily send it back [14:39] niemeyer: interesting. how would that work? [14:39] niemeyer: set a config attr saying "please archive now" ? [14:39] rogpeppe: Once we have the Go port nailed down, the stop hook should work too [14:40] niemeyer: yeah, *that* would be good. [14:40] rogpeppe: Yeah, you can use a named config value [14:40] rogpeppe: Whenever it changes, save it again [14:41] niemeyer: a bit of a hack, but yeah, that'd work ok. [14:42] rogpeppe: Doesn't feel like a hack to me.. it's a straightforward convention with named backups [14:44] niemeyer: ah, if you don't overwrite backups, yeah. [15:27] TheMue: go-mstate-life-file reviewed [15:46] TheMue: ping [15:57] niemeyer: seen it, cheers [16:00] niemeyer: just tested google two step authentication and got troubles with all those apps who need an individual pw [16:05] * TheMue has an LGTM overflow ;) [16:22] TheMue: Good stuff indeed [16:22] niemeyer: thx [16:22] niemeyer: and mstate is real fun [16:23] I'm mostly out of things to review, which would be a feat were not for the fact that the pending one is a massive 1k+ branch [16:23] I'll head to lunch now, though [16:23] TheMue: Glad to hear it [16:23] niemeyer: enjoy your lunch, i'll go to archery now ;) === Beret- is now known as Beret [17:26] fwereade_: it-lives has final review! [17:26] a final [18:28] Back in ~30mins [18:49] fwereade_: i wonder if you could do me a favour and check that the goamz live tests work ok for you. i get SignatureDoesNotMatch errors and i'm not sure if it because i've just mucked up my aws account or because there's actually something wrong. [18:50] fwereade_: hmm, you're probably done for the day, of course! [19:38] niemeyer: ping [19:58] rogpeppe: Hi [19:59] Doc appointment took a bit longer than expected [21:08] * niemeyer feels sorry for mramm [21:23] niemeyer, tyvm for reviews :) [21:23] fwereade_: My pleasure [21:24] niemeyer, the ErrDying thing has caused rogpeppe's dislike for ErrorContextf to click, though, by ErrorContextf-ing the modes I can't (easily) actually return ErrDying, and *that* was what I was having a problem with [21:25] niemeyer, I'll figure something out, though :) [21:25] niemeyer, `x.tomb.Kill(nil); x.tomb.Kill(ErrDying)` still gives Error() of nil, right? [21:25] fwereade_: Aha, that makes more sense, thanks [21:26] fwereade_: Yeah [21:27] niemeyer, ok, I'll come up with something clean, but maybe tomorrow [21:27] fwereade_: Of course, thanks a lot [21:27] niemeyer, cheers, gn [21:27] davecheney, gm and gn :) [21:27] fwereade_: Have a good night [21:31] niemeyer, oh, just one question on the other review -- part of me wants to say that rather than "dropping good information on the floor" I should maybe actually explicitly require nil when setting charm state to Deployed [21:31] fwereade_: What's wrong with preserving the URL? [21:31] niemeyer, it's not so much that the information is *bad* but it is a not a canonical source of truth [21:31] fwereade_: There is an actual URL, right? [21:32] fwereade_: Sure, but it still feels like we know what we've last put there [21:32] fwereade_: THat's what the info tells [21:32] niemeyer, there's deployedURL, which is definitely correct, and then there's this; if we have and trust this, then we don;t need the one inside the charm dir, do we? [21:32] fwereade_: While this isn't a big deal, I don't see a reason to not have it either [21:33] fwereade_: Makes sense, or am I missing something? [21:33] niemeyer, either this, or the .juju-charm file, seems to be redundant [21:33] niemeyer, I would be comfortable with either as the single source of truth [21:33] niemeyer, having both just makes me nervous ;) [21:34] fwereade_: It's not about having a single source of truth in this case.. they have different meanings [21:34] fwereade_: We're not trusting on both [21:34] niemeyer, hmm, true [21:34] fwereade_: We know they mean different things and may be different [21:34] fwereade_: and we know why [21:34] fwereade_: We're also not even exposing that info ATM [21:34] fwereade_: I'm suggesting that mainly because it enables us to change our minds [21:35] niemeyer, ha, yeah, fair enough :) [21:35] niemeyer, ok, sgtm [21:35] niemeyer, cheers [21:35] fwereade_: Cheers [21:36] niemeyer, oh, and JujuConnSuite.Reset is so that I can write the uniter tests as a table, in which each starts from pristine state -- am I missing some obvious better way to do that? [21:36] fwereade_: Hmm, interesting [21:37] fwereade_: I guess it sounds reasonable.. [21:38] niemeyer, it's wanting to do stuff like this that underlies my half-belief that Fixture is a useful concept that usefully encompasses Suite and Test... but sadly *doesn't* handle both Suiteness and Testness at the same time, so the idea clearly needs more thought [21:41] niemeyer, oh, and I'm not sure I explicitly thanked you: I forget which review it was where you suggested replacing table comments with summary fields and embracing zero values, but it really made table-driven testing click -- ie it now seems *better*, not just different, IYSWIM [21:41] fwereade_: That's awesome! [21:42] niemeyer, it's one of those things that now seems ridiculously obvious, it takes a bit of effort to remember the previous perspective :) [21:43] fwereade_: Yeah, I totally know what you mean.. quite a few things take a while to click [21:45] * fwereade_ threatens himself with a hammer, and reluctantly slinks off to bed; nn all :) [21:48] fwereade_: Have a good night man :)