[07:12] morning wrtp [07:12] fwereade__: hiya [07:20] morning [07:21] TheMue: mornin' [07:21] wrtp: heya [07:22] TheMue, heyhey [07:22] fwereade__: hi [08:43] wrtp, TheMue: I think https://codereview.appspot.com/6737050 is a trivial [08:45] fwereade__: LGTM [08:45] TheMue, thanks [08:46] fwereade__: looking [08:48] fwereade__: why can't we use the same log file for output and log messages? [08:49] fwereade__: in fact, we could just send log messages to stdout, maybe [08:50] wrtp, I dunno, I feel that it's a different class of message... anything going to out is evidence that we've royally screwed up somewhere [08:50] fwereade__: i'm somewhat -1 on adding another log file to monitor [08:51] wrtp, well, it was always there in python, and it saved us hours of debugging there, and would have saved it for us the other day too [08:51] fwereade__: i'm definitely not -1 on using Out [08:51] wrtp, in fact I'm pretty sure I LGTMed your container code only on the condition that you added it :/ [08:52] wrtp, but, meh, easy to miss :) [08:52] fwereade__: the nice thing about having output go to the same file as log messages is that you can see the output in context [08:53] fwereade__: given that, probably, the only output we're going to see is a panic stack trace, i think that's useful. [08:54] wrtp, or errors from before logging is set up, or failure to even launch the process [08:54] wrtp, it's a backstop for should-never-happen errors [08:54] fwereade__: true too. but it's all good to see in the context of the other logging messages (for instance the ones that were logged on the previous run) [08:56] wrtp, are you suggesting replacing --log-file with an Out then? [08:56] wrtp, or using both to write to the same place? [08:57] fwereade__: if we just remove the --log-file flag, will the Out field cause the stderr log msgs to go to the log file? [08:58] wrtp, think so, haven't tried [08:58] fwereade__: yeah, it looks like it will [08:58] wrtp, but does make it somewhat tricky to add log rotation in future, I think [08:58] fwereade__: that's an interesting point. [08:59] wrtp, and that does feel like something we will need sooner or later [08:59] fwereade__: we could potentially implement log rotation by restarting the agent, i suppose [09:00] fwereade__: if it's sufficiently infrequent, that's probably no problem [09:00] wrtp, that approach feels a bit lumpen to me, but maybe it's a matter of taste [09:01] fwereade__: yeah, i know what you mean [09:01] wrtp, anyway, I think we have passed the "this is not trivial" milestone in this discussion [09:01] fwereade__: sorry [09:01] wrtp, I'm fine punting that decision to niemeyer -- I don't *really* care how we get the panics, just that we do somehow getthem :) [09:01] wrtp, np at all [09:02] wrtp, whatever we implement should be pretty simple in the end [09:02] fwereade__: yeah [09:08] * wrtp wishes upstart had a reference manual [09:27] moin. [09:27] davecheney: any chance of recording from the gophers' meeting? [09:47] Aram: moin, moin [09:48] Aram: sorry, thre was no recording [09:48] but the slides are online [09:48] pity. [09:48] talks.golang.org [09:48] yeah, seen them. [09:48] sorry, very low fi [09:48] we didn't have it at the swanky google offices [09:59] hmm, how does anyone else feel about s/RelationEndpoint/Endpoint/ ? [10:05] fwereade__: in state, presumably? [10:05] wrtp, yeah [10:06] wrtp, Endpoint only means one thing, just like Unit, and we don't call Unit ServiceUnit [10:06] wrtp, (RelationUnit is, I think, different, because it really is a combination of a Relation and a Unit) [10:06] fwereade__: mixed feelings. on the one hand, i think "yeah, great idea - shorter name". one the other hand, the state name space is quite crowded. [10:07] wrtp, expand on the second bit -- don't see how a simple name change affects the crowdedness [10:07] fwereade__: it's just that it's not always obvious which names relate to which things. [10:07] fwereade__: but on balance i'm probably +1 [10:08] wrtp, ok, I'll probably run it past niemeyer when he arrives [10:08] wrtp, cheers :) [10:08] later all, lunchtime [10:57] http://codereview.appspot.com/6734043/ [11:04] hey, this is fwereade in the wrong place... am I meant to be in a meeting [11:05] (I went to fix cath's aunt's toilet over lunch, and it took a bit longer tan expected, but dimiter's house is on the way home...) [11:08] meeting ? [11:08] not that i know of [11:09] dimitern: neither me [11:09] wrtp: well, it's dimitern again, fwereade is on the way home now :) [11:10] dimitern: :-) [11:10] dimitern: nice meetin' ya :-) [11:10] (if this could actually be called "meeting"...) [11:13] wrtp: likewise :) [11:31] hey again all [11:32] I just had a sudden hey-wait-we-sometimes-meet-on-fridays-I-hope-it's-not-today moment [11:47] fwereade__: yeah, blah blah, airplanes, blah blah, uds [11:47] during this season, regular concerns are suspended [11:47] davecheney, haha :) [12:27] Hello all! [12:29] I'll step out to run some errands I have to sort before Copenhagen.. back later. [12:29] niemeyer: morning :-) [16:30] niemeyer_, g'morning had a question re the bzr revid that the store could serve up (namely how to access), pls ping me when your around [16:35] hazmat: Heya [16:35] hazmat: It's already implemented, but it's not yet deployed [16:35] hazmat: We have to sync up with mthaddon to get it out [16:35] hazmat: It's just another field on that usual charm doc info [16:36] niemeyer_, ah.. okay. i was wondering about that. will it just appear in the store data for a given charm alongside revision and sha? [16:36] cool [16:36] info doc [16:36] perfect, then the browser will just pick it up [16:36] hazmat: "digest", specifically [16:36] hazmat: The generic term is because it doesn't really matter for the store.. it's whatever the revision control labels it [16:37] sounds good, i just wanted to make sure it was in the info doc. [17:41] time to stop for the day. [17:42] have a great weekend everyone, see y'all monday or in copenhagen! [20:25] wrtp, cheers