[00:15] curses!!!! test pass with .only, fail when run with everything else. [00:16] :-/ [00:19] it's around date formatting...how can that get corrupt? Man. [00:19] yuck [00:20] sorry, just ranting. [00:20] was hitting -cr thinking I was done for the night :P [00:20] I understand the feeling :-) [00:21] oh well, will look tomorrow with fresh eyes. party on [00:23] :-) ttyl === jcsackett_ is now known as jcsackett [01:29] fwiw, unexposed issue has gone away somehow, and I fixed Landscape issue. yay. [01:38] nytol [11:17] rogpeppe: hi, re CharmInfo error, from a first investigation, it seems to be a GUI error: the GUI calls CharmInfo before the charm is added to state, i.e. before the charm is deployed. makes sense? [11:18] frankban: that makes sense, but... what charm is it trying to get? [11:19] frankban: if the charm isn't added to state, why is the GUI trying to get info on it? [11:20] rogpeppe: because, I suspect, of a bug in the GUI. For example, https://ec2-50-17-53-143.compute-1.amazonaws.com [11:21] frankban: that all seems to work ok for me [11:22] frankban: BTW, unrelated remark, shouldn't those service icons show the number of deployed units for each service? [11:22] frankban: oh, i see, it only shows that when i hover the mouse over it [11:22] rogpeppe: if you click on top-right "Charms" and then, e.g. "cassandra", and then "Deploy", a charminfo is requested for cs:precise/cassandra-2 [11:22] (before the charm is deployed) [11:23] frankban: ah, i see! [11:23] frankban: that seems to argue that you might want a separate PutCharm command [11:24] frankban: then you can get the API to download a charm and add it to the state without necessarily deploying it [11:25] rogpeppe: I have to investigate further, who is doing that call, what;s the reason? AFAIK we also have the ability to retrieve charm info from the store [11:25] frankban: yeah, actually, i was just thinking that [11:25] frankban: well, i'm glad it's not a juju-core bug :-) [11:26] rogpeppe: well, me too :-) [11:26] frankban: nice to see the GUI working, BTW, and with double-clicking on services working too! [11:26] rogpeppe: yes, very cool [11:27] * frankban lunches [11:27] frankban: i've just pointed people to have a look at your gui instance [11:28] rogpeppe: ok, I will use that for a debugging session after lunch, so it will be around for a while [11:29] frankban: hmm, it's a bit broken. [12:04] * teknico lunches [12:04] rogpeppe, frankban if we are talking about a service from the Juju environment, we should get the charm from the Juju environment IMO. We have a variety of future stories about proxied charm stores [12:05] gary_poster: it sounds like this charm info call is happening before the service is deployed. currently there's no way to get charm info from the juju env unless an instance of that charm is deployed [12:06] gary_poster: I confirm the GUI calls CharmInfo before the charm is deployed (after "deploy" is clicked but before "confirm") [12:07] rogpeppe, ack; just saying that if it is supposed to be getting information about a service, we should ideally make sure that we are asking the environment about it. ok frankban, thanks [12:08] (rogpeppe, we actually have separate stores for the charm info we obtain from the store and the charm info we obtain from the environment) [12:08] gary_poster: interesting. [12:08] perhaps paranoid, or too future looking [12:08] my fault if so :-) [12:11] teknico, you probably know about the relation.ids that have spaces in them right? We may need to normalize them [12:12] well, perhaps we can do that in our d3 code [12:12] probably a better place [12:19] gary_poster: so why is deploy calling CharmInfo before the charm has been deployed? surely it should wait for the deploy to be successful, *then* get info on the charm from the environment? [12:20] rogpeppe, 1 sec, writing diagnosis for bug 1167295 :-) [12:20] <_mup_> Bug #1167295: graphical link not updated properly (juju core) < https://launchpad.net/bugs/1167295 > [12:22] gary_poster, I don't! I got nutin' to do with 'em! I don't even *like* them! frankban made them, I saw him! he's the culprit! take him, not me! [12:22] teknico, :-) [12:22] (and I'm lunching anyway, ttyl :-) ) [12:23] ttyl [12:23] gary_poster: aha! the relation keys used by juju-core have spaces on them,e.g. "haproxy:reverseproxy wordpress:website". if we use the key as is in the DOM classes this could be the problem [12:24] frankban, right :-) I just put that in the bug [12:24] (s) [12:25] I was unable to deploy dokuwiki [12:25] no helpful error in gui [12:26] gary_poster, rogpeppe: maybe an helpful one in the juju logs: http://pastebin.ubuntu.com/5695209/ [12:26] ooh, we still convert the ghost to a real service even on error [12:27] oh... [12:27] it actually is in the go environment [12:27] so not a problem on gui side [12:27] frankban: interesting [12:27] gary_poster: if you didn't see the "no options to set" growl notification, that's another bug [12:27] ah, I didn't [12:28] Well, I saw *a* growl notification [12:29] which probably correlates to that error [12:29] frankban: looks like a bug to me. i don't think "no options to set" should be an error [12:29] but I didn't see the message in growl or details [12:30] frankban, I can't dupe the CharmInfo thing. [12:30] w...t...f... [12:31] so my view module requires the datatype-date datatype-date-format. cool [12:31] but when running tests, it's not loading the en-US module [12:31] unless I require those modules also in my test's .use() [12:31] then it's peachy [12:32] * rick_h_ shakes fist at strange test/setup stuff [12:32] rick_h_, in devel and prod both? prod that might make sense. Well, no, not really. :-/ [12:32] gary_poster: I think you duped it: http://pastebin.ubuntu.com/5695226/ [12:33] gary_poster: that problem is not visible in the GUI [12:33] gary_poster: yea, both. I only noticed that a later test module loaded new date-format files and that the test suite had some date modules in .use() [12:33] frankban, ah! not even with a JS exception [12:34] rick_h_, gotcha. :-( dunno. I don't see how what we are doing is all that weird, except for instantiating a Y rather than running within the Y's function. I wonder if that would make a diff [12:35] gary_poster: sorry, just more mean the whole single file stuff I guess. Because when I did my tests with a .only all the modules loaded fine and it worked [12:35] but only when I removed the .only did I have Y problems then it seems. and they were pretty invisible because it just wasn't loading the locale stuff it needed to parse the dates, but the date format module did load so it ran but then bloew up mid-date-parse [12:36] just feels like it's half magic to get the right incantation to get bits to play together. [12:36] rick_h_, I agree, and I think we all would. Do you have any suggestions? The only one I have is one that I talked about a long time ago and that is probably too late to be practical: switch to the YUI test runner [12:37] all good, tests pass I can -cr and get this landed. Just lost a couple of hours so frustrated atm. [12:37] rick_h_, it's all good until the next time... :-) [12:37] gary_poster: yea, no. The only suggestion is to isolate tests into own files so that the tests stand alone, but that's probably a lot of work [12:38] while running tests in LP was a pita it was nice that you had a small known world while you were working. [12:39] gary_poster: heh no until next time. Sorry, I'm coming across as whiny to pulling up short. :) [12:40] rick_h_, not at all, you are trying to be nice about being frutrated :-) [12:40] frustrated [12:40] rick_h_, if we divided up all the tests into their own test runner (that's what you mean, right?) then how would you aggregate? [12:42] gary_poster: yea, in small projects I've just opened them all. I've always meant to look at yeti since it can do some agg and run in diff browsers I believe [12:43] rick_h_, ah ok. yeah, that would require us to be using the yui test runner/syntax I think :-/ [12:43] https://github.com/yui/yeti just uses globbing to file all the .html files and launch browsers to cycle through them [12:43] gary_poster: right, which would be a ton of work from where things are. [12:43] right :-/ [12:43] gary_poster: so yea, I'm complaining without a great solution from where things are atm [12:44] s'ok, doesn't mean problem isn't real or worth talking about [12:44] if I get a sec I might look out for something that works around the current mocha stuff. I'm not as familiar with the other tools around mocha off the top of my head [12:45] cool, thank you rick_h_ [12:58] jcsackett: hatch review call if you guys get a chance after some coffee and the like. https://codereview.appspot.com/8621043/ [13:04] rogpeppe: filed bug 1167344 [13:04] <_mup_> Bug #1167344: Deployment via the juju-core API fails with a "no options to set" error in some cases

< https://launchpad.net/bugs/1167344 > [13:05] frankban: thanks [13:05] frankban: and there's another associated bug i've just seen [13:05] frankban: the service is still created when that happens [13:05] frankban: i'm just doing a substantial cleanup in that area [13:06] rogpeppe: the bug includes the output of juju status for that reason [13:06] rogpeppe: cool, thanks [13:08] frankban, if I create a relation on your instance will that cramp your style? I want to try something [13:08] gary_poster: go ahead, no problem [13:09] thank you [13:17] teknico: swap reviews? [13:18] https://codereview.appspot.com/8623043 [13:18] bac, sure! [13:18] "cramp your style", now that's a new one :-) [13:23] rogpeppe, do you know offhand what the constraints are for valid names of services, interfaces and relations in juju? [13:31] teknico, want me to review your destroy_service, or wait till you add test Francesco mentioned? [13:32] either way is fine with me [13:32] gary_poster, probably better wait, also I think bac is reviewing it too [13:32] oh ok teknico [13:32] thx [13:32] gary_poster, I'll put the card back in coding [13:33] ok [13:35] teknico: done [13:36] bac, me too [13:37] teknico: hah on the misspelling. i meant to look it up. thanks. [13:38] bac, I'm actually kind of appalled by the usage stats :-) [13:39] gary_poster: i'll have a look; there are some regexps in state [13:40] gary_poster: a service name must match this regexp: [a-z][a-z0-9]*(-[a-z0-9]*[a-z][a-z0-9]*)* [13:40] frankban, did you mean that the successful service destroy test should check that the callback is actually invoked? [13:41] thanks rogpeppe! handling all the various allowable character sets in different parts of the system is...exciting (CSS + HTML + Juju) :-) [13:41] teknico: yes, with the arguments we expect [13:41] if interfaces and relations have similar constraints then juju will be the easiest [13:41] frankban, right, will do [13:42] gary_poster: i'm not sure that there are any current restrictions around interface and relation names [13:42] darn [13:42] ok thanks rogpeppe [13:42] rogpeppe, +1 on adding some, as restrictive as possible ;-) [13:42] gary_poster: i'll moot the possibility in juju-dev [13:42] service name regex would be great, for instance [13:42] thanks [13:43] excluding colons and spaces would nice at a minimum rogpeppe [13:44] gary_poster: if you're looking to do a review: https://codereview.appspot.com/8623043 [13:44] mostly looking to solve a bug, but I need room on the boatrd to actually work on it ;-) [13:44] so happy to review [13:45] gary_poster: that's the way it works... [13:45] gary_poster: +1 [13:45] cool rogpeppe [13:46] bac you reviewing frankban's bug 1166727? [13:46] <_mup_> Bug #1166727: Fix Status/StatusInfo handling in the go delta stream < https://launchpad.net/bugs/1166727 > [13:47] gary_poster: no [13:47] gary_poster: i can [13:47] bac cool please do, I'm not done with yours yet :-) [13:47] gary_poster: i will. ok, sure [13:47] thx [13:47] bac: thank you [13:51] frankban: done [13:53] bac: thanks [14:02] bac done. [14:02] thx [14:03] gary_poster: nice: http://jujugui.wordpress.com/2013/04/10/test-tip-whenever-you-instantiate-our-yui-app/ [14:03] thanks bac [14:07] gary_poster: any chance I could get a review on https://codereview.appspot.com/8578043 I'd like to land it as is then spin up another to finish off the final tasks [14:07] if possible that is [14:07] ok hatch, on it [14:07] thanks [14:08] hatch, you prefer typeof over Y.Lang.isString for speed? (fine by me, just learning your rationale) [14:09] gary_poster: just a preference really - all Y.Lang.isString does is do a typeof [14:09] http://yuilibrary.com/yui/docs/api/files/yui_js_yui-lang.js.html#l175 [14:09] ok cool hatch [14:10] rick_h_ and hatch: can i get a look at https://codereview.appspot.com/8600043/ [14:10] jcsackett: sure, otp but will look at in a sec [14:10] rick_h_: no rush. [14:11] rogpeppe, what's your expected schedule for turning on authentication, btw? [14:30] gary_poster: [sorry, was in a call] [14:30] gary_poster: i should have a branch proposed by eod [14:30] rogpeppe, yay :-) [14:30] gary_poster: BTW, do you actually need to see a unit's resolved mode in the API? [14:32] * gary_poster looking [14:34] rogpeppe, we allow resolving (though this doesn't appear to be working/showing up on go side) [14:34] jcsackett: comments replied to [14:34] gary_poster: ok, but do you need to see the resolved mode that's been set? [14:34] gary_poster: i suspect not [14:34] jcsackett: loading your review now [14:34] rogpeppe, I don't think so, no [14:34] gary_poster: in particular i'm thinking of removing the Resolved field from UnitInfo [14:35] looking, three deep in stack now ;-) [14:36] rogpeppe, no we do not need it AFAICT thank you [14:37] gary_poster: cool, thanks [14:49] rick_h_: when you asked me to add "new" to json and sidebar you meant "popular", right? b/c "new" is already there. [14:50] jcsackett: right, sorry the request was more "please look at the wireframes and have all three charm-containers supported" [14:50] jcsackett: so featured, popular, and new [14:50] rick_h_: that's what i figured, just double checking. [14:50] thanks [14:53] gary_poster: do you still need my GUI instance? if so, no problem, if not, I'll destroy it and create another one [14:53] frankban, destroy away, thank you [14:53] gary_poster: ok thanks [14:56] anyone else need any more reviews? [14:56] hatch: see you in a bit :) [14:56] hah [14:56] gary_poster: any closer to letting me know if I can merge what I have? :) [14:56] hatch, I just gave you a ton of comments [14:56] well, not really a ton [14:56] oh heh [14:57] but some ;-) [14:57] slow internets there [14:57] no it was essentially simultaneous to your comment, hatch ;-) [14:57] oh lol gotcha - ok will check again [14:57] so hatch, your goal is to land and then immediately follow up? I'm ok with that. will mark as such [14:59] hatch, "LGTM to land with immediate follow-ons for other changes as discussed." [15:02] ok thanks - yeah I want to get this landed because it's getting a little long in the tooth :) [15:03] I'll make your recommended changes in this branch and then followup with the inferred interface stuff [15:03] cool sounds great [15:04] anyone else available for a review on https://codereview.appspot.com/8578043 ? [15:05] hatch: I can peek at it in a few min if that works [15:05] yep it'll take me a bit to make the changes from Gary's review [15:05] hatch: k, hit me up after that then and I'll peek [15:07] yesterday I educated someone on Google hangouts for meetings - they were blown away by how simple it was [15:08] they said "with this, why do people need to go into an office?" haha [15:15] gary_poster: always keeping me on my toes with acronyms :) [15:16] YMMV - your mileage may vary ? [15:16] hatch, yes, sorry :-) [15:17] haha np - eventually we'll get english communications down to one big acronym EWGECDTOBA [15:17] :D [15:25] It appears console.log doesn't work at all in firefox. Is that a known issue? [15:26] hmm I haven't seen that before [15:26] in the console if you type `console.log` does it give you a fn? [15:26] benji: huh? using firebug or in the normal FF debugger? [15:27] hatch: it does (return a function) [15:27] ok so then console.log still exists [15:28] question now is why is it not outputting [15:28] rick_h_: normal debugger, we sometimes replace the console object with our own (to smooth over cross-browser issues and to disable logging in production builds) and that is what I suspect as the issue [15:28] ahhh [15:28] that could be it [15:30] hatch: nope, it turns out to be a PEBKAC [15:30] * benji tries to get hatch to learn all the acronyms in one day. [15:30] oh I know that one [15:30] I've used that one [15:30] lol [15:30] :) [15:31] it turns out that FF behavior has changed in say the last year since I tried to debug anything with it and the place that console.log messages go is defferent than what I remember, so I was looking in the wrong place [15:32] yea, they've been working hard on their dev tools stack [15:32] silly me, looking at the "console" to see the results of "console.log" [15:32] saw they finally released a network panel in the nightly today [15:32] rick_h_ yeah but I found out the other day I have NO idea how to open the full debugger lol [15:33] all I can get is a thick blue bar across the bottom [15:33] well the pain point is they're all different tools [15:33] * rick_h_ hugs chrome dev channel for best dev tools around [15:34] yeah I can't get the Safari one [15:34] it makes less sense to use than the IE one [15:34] if maybe just barely [15:34] lol [15:34] what's sad is the safari stuff is based on the same webkit stuff as chrome [15:35] they've managed to mess it up [15:36] Yeah it may be super powerful but if it's not intuitive it's a horrible tool [15:36] we have to learn so much new stuff every day, a new dev console is not high on the list [15:38] benji, is your "More picky yuidoc linter." in Slack Review really under review? or does it refer to the already merged lp:~benji/juju-gui/yuidoc-lint-comment-style branch? [15:38] teknico: it's merged, I'll move it; thanks [15:54] gary_poster, the build in jenkins after I landed my branch failed, apparently for unrelated reasons, do you want to restart it? [15:55] teknico, done (though you can too, you know :-) [15:56] gary_poster, yeah, I was reading up the docs and I don't think I yet have the whole VPN setup [15:56] jujugui: experiment: please try updating kanban board now to be up to date with your status it it is not already. I will review in two minutes. I will try to then only ask about things I don't think I understand or speciifically want to know more about [15:56] s/it it/if it/ [15:58] (done already, also bothered other people already about it :-) ) [15:58] :-) thx [15:58] hatch, duh, I meant map, thanks for correcting my thinko [15:58] rick_h_ https://codereview.appspot.com/8578043/ so this is an intermediary branch so you should see some TODO [15:58] jujugui call in 2 [15:58] gary_poster: haha np - I had to read it a few times before I assured myself it was a typo :) [15:59] yeah, sorry [16:14] frankban: have you got a gui instance up that i can check something on? [16:15] rogpeppe: https://ec2-54-224-91-223.compute-1.amazonaws.com/ [16:15] frankban: ta! [16:16] frankban: how does the YAML config file upload thing work? [16:17] frankban: i'm wondering if it's per-service or for the whole environment [16:17] gary_poster, teknico, bac: ^ [16:28] rogpeppe, on call will be with you soon [16:28] gary_poster: np [16:28] bcsaller, ready for you :-) [16:28] rick_h_ able to check out that review now? [16:28] jcsackett: can you peek at it again. Had to make some bigger changes but think it's all better now. https://codereview.appspot.com/8621043 [16:28] hatch: will do [16:29] thanks [16:29] rick_h_ oh looks like ben already did [16:30] rogpeppe, per service AFAIK [16:30] hatch: yea, just noticed that [16:30] I'm not sure what it would be for environment [16:30] woot! off the hook lol [16:30] rogpeppe, I know that we do upload service yaml [16:30] gary_poster: apparently the YAML file should contain a map of service names and a set of config attrs for each service [16:30] rick_h_ lol [16:31] yay landing that branch....finally [16:31] lol [16:31] even if it is bittersweet [16:31] rogpeppe, um. that doesn't sound familiar but on call again will return soon [16:31] :) [16:32] gary_poster: that's what i've just been told. i haven't looked at the python though. [16:32] oh darn there is a conflict === matsubara is now known as matsubara-lunch === deryck is now known as deryck[lunch] [16:42] ugh [16:46] hatch: can you look at https://codereview.appspot.com/8600043/ ? after sorting out your own woes, of course. :-P [16:46] certainly [16:46] almost done [16:46] me hopes [16:46] :) [16:47] I love it when merges put methods inside of methods [16:47] lol [16:50] jcsackett: all done, thanks for the cleanup [16:53] rogpeppe, the service name map does not make a huge amount of sense to me still given the usage--in the command line and the GUI you specify a config file for a service, at least in pyjuju--but yes. This is an example of something we were using last year, for instance: http://bazaar.launchpad.net/~yellow/charms/oneiric/buildbot-master/trunk/view/head:/examples/lpbuildbot.yaml [16:53] we would use that with [16:53] juju set --config=[FILE] buildbot-master IIRC [16:54] juju set --config=[FILE] buildbot-master [16:54] IIRC [16:54] which makes little sense [16:54] gary_poster: i agree [16:54] gary_poster: but there's the compatibility issue [16:54] certainly that's what the help suggests (juju set [options] ) [16:55] gary_poster: yeah, the help is no help :-) [16:55] :-) [16:55] rogpeppe: looking at the delta arriving to that ec2 instance, it seems that the status of the machines never do the transition from "pending" to something else [16:55] frankban: that's right [16:55] frankban: machine agents don't currently set status [16:55] rogpeppe, ah! we used it in deploy [16:56] where it made more sense perhaps [16:56] like this [16:56] frankban: just proposed: https://codereview.appspot.com/8630043/ [16:56] juju deploy --repository=./charms local:buildbot-master --config=./charms/oneiric/buildbot-master/examples/pyflakes.yaml [16:56] rogpeppe: great! [16:57] gary_poster: just had first successful live test pass with auth enabled, BTW [16:57] rogpeppe, so the config file format makes more sense there [16:57] gary_poster: kinda. i'm still pretty meh. [16:57] rogpeppe, awesome! no changes needed in GUI? [16:57] gary_poster: none [16:57] whee! [16:57] hatch: see 3.9.1 notes? [16:57] rogpeppe, with you on the meh [16:58] rogpeppe, but I say compatibility now, discussions later :-) [16:58] gary_poster: thanks for your sane suggestion yesterday re: passwords. it pulled the wool from my eyes. [16:58] gary_poster: i guess [16:58] rogpeppe, glad the suggestion made sense :-) [16:58] for config... [16:58] I dunno [16:59] the GUI doesn't have a lot to do with it [16:59] we just pass it to you [16:59] for a given deploy [16:59] I don't *think* we expose doing it for a set yet [17:00] so do what you think and let us know how it turns out, rogpeppe ;-) [17:00] gary_poster: ok [17:04] rick_h_: hmm? [17:04] hatch: release notes for 3.9.1, big handlebars bug [17:05] hatch: I guess we must not hit it much since our tests didn't go boom, but heads up [17:05] aren't we using 3.9.1? [17:05] hatch: hmm, just saw it on the rss feed so assumed it was 3.9.0 [17:06] 3.9.1 was released at the end of Match I think [17:06] hatch: heh, google reader fail...nvm :) [17:08] :) [17:08] who are you going to switch to once Reader is nomore? [17:08] I just use it as an API for gReader [17:08] hatch: no idea. Everything is fail so hoping they get better in the next couple of months [17:08] isn't Feedly the goto one? [17:08] :( [17:09] needs some major UX help [17:09] same with newsblur [17:09] ahh, have you seen the gReader app? I really like it (paid $5 for it lol) [17:10] hatch: no, web or bust [17:10] oh - I only ever read my feeds on my N7 [17:11] I do all three, phone, N10, and laptop [17:11] and it's UI is very functional [17:13] rick_h_ while we are on the topic of YUI updates...did you see the massive performance gains in 3.10? [17:13] rogpeppe: EOD, do you want me to leave the ec2 env alive? [17:14] frankban: no, you can take it down, thanks [17:14] rogpeppe: ok, cool, have a nice evening [17:14] frankban: and you [17:17] hatch: yea, looking forward to running the test suite with it :) [17:20] gary_poster: https://codereview.appspot.com/8626044/ === matsubara-lunch is now known as matsubara === deryck[lunch] is now known as deryck [17:48] rogpeppe, great :-) [17:49] bcsaller: so what's supposed to happen if the fakebackend has multiple possibilities for the inferred relation? [17:49] hatch: it has to throw an error [17:50] in Python we threw an ambigious relation error [17:50] * bcsaller looks at what python.js does [17:50] ok and that's before adding juju-info to the provides? [17:52] hatch: no, because I think that the charms won't define a requires juju-info as an example so it won't automatically match (its provided but not required) [17:52] i'm off now. g'night all [17:52] * hatch really wishes these rules were documented somewhere [17:52] night rogpeppe [17:53] hatch, on the bright side, arguably that's what we are doing. kinda. or I tell myself this. :-P [17:53] haha - I still think not requring an interface is wako [17:53] :P [17:56] requiring an implicit interface always would be bad [17:57] how do you figure? [18:02] everything would relate to everything [18:03] oh I misread - I was saying to always require an explicit interface [18:09] Makyo, do you have some time to pair with me on some d3 stuff? If you're in the middle of something, np [18:09] gary_poster, Sure. [18:09] oh awesome I found a new feature in Sublime.... [18:09] thanks Makyo guichat [18:09] 'New View Into File' [18:10] allows you to open up the same file in multiple windows [18:10] w00t I always missed that feature from Netbeans [18:10] *cough*vim*cough* [18:10] rick_h_: if it was in vim I never would have found it because it would have been under some obscure config property that I'd need to enable :P [18:11] s/obscure/obsure and undocumented [18:11] :P [18:11]