[01:53] jcsackett ahh thanks [01:54] huwshimi so any luck tackling it on your own? [01:54] hatch: I didn't mange to figure out anything yesterday :( [01:55] ok cool, I'll take a look [01:56] thanks heaps [01:56] huwshimi what if you include the juju-templates? [01:56] asically the error is because it can't find the element in the dom [01:58] I'll try [01:59] hatch: No change [02:00] blarg [02:01] ok I've pulled it down [02:01] will investimagate [02:02] hatch: Feel free to leave it to your work day, I have plenty to go on with. [02:12] huwshimi it's because it's rendering without the mv flag [02:12] it's rendering the old scaleup UI [02:12] if you put a debugger in the method that's failing and inspect the innerHTML of the container you can see this [02:13] Oh [02:14] huwshimi that's because the createServiceInspector method is no longer how we render the inspectors while under mv [02:14] so this is actually testing a path which is no longer valid [02:15] so really - the proper fix is to fix the way inspectors are rendered [02:15] for tests [02:16] but you may be able to hax around it by setting the mv flag and doing some other bits [02:27] yep, trying [02:59] huwshimi any luck? [03:02] hatch: Just having some lunch... [03:02] ok cool, I think I'm gona head and watch some tv, if you can't get it just leave the PR and I'll take a look in the am [03:04] have a good one === uru_ is now known as urulama === uru_ is now known as urulama [07:21] urulama: morning! [07:21] rogpeppe: hey there [07:22] rogpeppe: how's it going? [07:23] rogpeppe, urulama: Morning [07:23] urulama: not bad thanks [07:23] rogpeppe: i'm trying to finish the "sprint summary" document so that i can move forward with the implementation part, but it's taking tiiiiime :) [07:23] urulama: have you recovered from last week yet? [07:23] huwshimi: good evening ;) [07:23] :) [07:23] rogpeppe: em ... partly recovered. progress bar at around 33% :) [07:24] urulama: :-) [07:27] rogpeppe: we have a national holiday on Friday, might help to move to 50% :) and it's good to put things on paper as long as they are still in memory [07:27] urulama: definitely [07:29] rogpeppe: i plan to jump on PRs and current implementation around noon (in 2h) ... would that be ok? any open questions or anything until then? [07:29] urulama: no, all's good [07:30] rogpeppe: great to hear ... ok, back to tasks for the next 2 months [07:32] rogpeppe: oh, btw, how far have you and Francesco come regarding uploads? [07:32] urulama: uploads now work [07:33] rogpeppe: as in "zip properly formatted dir" and upload and new charm is in charmstore? [07:33] urulama: and you can now download individual archive files too [07:33] urulama: yup [07:33] \o/ [07:34] urulama: ah, the only thing that's not done is that we don't currently verify bundles against the existing charms [07:35] rogpeppe: as in when bundle is uploaded it is not checked wether the charms are in the CS? [07:35] urulama: yes [07:35] urulama: we deliberately left it for a later PR - but it does need doing [07:36] urulama: and there's also something i thought of in bed this morning [07:36] urulama: which is that we can't always start revision numbers from zero [08:00] morning, frankban [08:00] morning urulama [08:00] urulama: are you able to connect to canonical IRC servers? [08:00] yes [08:00] frankban: but i've seen others being disconnected [08:00] urulama: uhm, ok [08:01] * TheMue has troubles too, also with SSO [08:01] urulama: I see "Connection failed. Error: Connection refused" [08:01] will try again later [08:01] switched to zsh, let's see how it works [08:13] frankban: next step acme? :) [08:15] urulama: http://www.nooooooooooooooo.com/ [08:15] ;-) [08:15] :) :) :) [08:16] * urulama bookmarks the page for future use [08:38] frankban: ok, i have problems with cannonical irc as well, however, it manages to reconnect [08:39] urulama: ok [10:45] urulama, rogpeppe, jrwren: https://github.com/juju/charmstore/pull/69 [10:46] frankban: finishing a doc, then lunch, then PR. ok with you? can move PR first if it is blocking anything else [10:47] urulama: it's ok thanks [10:56] frankban: reviewed [10:57] rogpeppe: thanks [11:40] * urulama goes on short lunch [12:26] frankban: the tests currently check mostly if the time is there ... could we create a bit more complex test? starting with a charm revision at time 0, then upload new revision, one in the period T of time? Then, check if all revisions have proper n*T archive-upload time? [12:29] urulama: not sure if I understand. at the store level, we already check that the time is the expected one (in a specified timeframe) [12:34] frankban: suggestion is to have a "possible" reggression test, that adds a charm (addCharm), multiple times, and check proper revisions and timestamps [12:35] frankban: on the api level [12:36] urulama: so, we publish a charm with the same id two times and we check the id is incremented and t2> t1? [12:37] frankban: yes, maybe also with "sleep" of time T between publishing [12:37] urulama: why sleeping? [12:41] urulama: i don't really see what that test would be adding [12:41] urulama: given that we're testing that the upload time is within a very short period [12:42] urulama: (assuming frankban applies my suggested change) [12:42] rogpeppe: I did [13:03] * frankban lunches [13:10] * rogpeppe also lunches [13:36] ha, this is a nice tool: http://www.downforeveryoneorjustme.com/lastpass.com [14:17] bac: should be www.downfordownforeveryoneorjustme.com :-) [14:18] rogpeppe: i don't get it [14:18] bac: well, really, it would be an infinitely long name [14:18] bac: the point being that it's only that web site that makes the check [14:19] rogpeppe: true [14:19] bac: so it could be down for everyone except me and their server [14:20] rogpeppe: regardless of the inaccuracy of their naming, it is a nice check, especially when you have reason to believe it is just you. [14:20] as i often do [14:20] bac: absolutely. i've found it very useful in the past, although i'd forgotten about it of late [14:47] wb Makyo [14:47] rogpeppe: I am going to change BulkIncludeHandler to receive a request rather than a method and flags [14:47] frankban: i started off doing that [14:47] frankban: but it didn't work out [14:47] rogpeppe: we need access to req.Body e.g. for the extra info stuff [14:48] rogpeppe: why? [14:48] frankban: really? [14:48] rogpeppe: extra-info PUT [14:48] frankban: hmm, i guess i thought that stuff would be in the flags, but you're probably right [14:49] rogpeppe: anyway, I'd feel the framework to be more friendly if it gives the request to you anyway [14:49] frankban: the reason it didn't work out was because of Router.GetMetadata [14:50] frankban: and that it seemed wrong to be making up Requests for bulk metadata requests [14:50] jujugui call in 10 [14:50] frankban: are you planning to do extra-info ? [14:50] frankban: i was working towards that [14:51] rogpeppe: I was just taking a look at the framework and realized that. If you want to tackle it I'll live it to you [14:51] rogpeppe: I'll look at something else, like verifyWithCharms in publish API [14:51] frankban: i'd be happy to pair on it if you like. i'm wading through mud currently. [14:52] rogpeppe: sounds good [14:52] frankban: cool, let's do that. [14:52] frankban: after the standup? [14:52] rogpeppe: sure [14:56] frankban: BTW, if we used a standard "application/x-www-form-urlencoded" MIME type, then the PUT arguments will end up in the flags argument [14:56] frankban: we don't say much about how the PUT values are encoded in the spec [14:57] rogpeppe: good point, we dont [14:58] jujugui anyone else getting "party is over" error? [14:58] hatch: yup [14:58] worked now [14:58] just keep trying I guess [14:58] heh [14:59] urulama, frankban: we could standardise on using a single form attribute, say "value" containing the JSON-encoded value to PUT. [15:00] rogpeppe: It's doable, still not sure if I like it over just sending the json in the body [15:02] "It's taking too long to connect you to this video call. Try again in a few minutes." [15:02] jrwren: daily standup :) [15:02] "please wait..>" [15:15] rogpeppe: gogogo in 5 [15:15] frankban: ok [15:15] frankban: i'm there, BTW [16:01] man somethin is munchin on my internets [16:05] hatch: this is the second day, if we're going by your sign on/sign off. [16:05] yeah it is :/ [16:06] good thing I don't pay for QOS [16:06] oh wait.... [16:09] rogpeppe your card that's in landing....is it still landing or blocked? Maybe it should be moved into tracking? [16:10] hatch: i'm trying to land it but it's blocked [16:10] hatch: if tracking is a better place for that, then by all means move it there [16:11] rogpeppe yeah I think that lane is for actively landing....whereas yours might sit there for a week? :) [16:23] hatch: completely befuddled on this test I'm trying to fix. Ping me when you have a few moments. [16:23] kadams54 you're in luck I just finished a review [16:23] what's up? [16:25] this is the currently breaking code: https://github.com/juju/juju-gui/blob/develop/app/widgets/deployer-bar.js#L257 [16:25] It loops through a set of keys and checks to see if any of their values have a length. [16:26] ok [16:28] is there more to this story? [16:28] :) [16:28] The problem is that my code adds a new key who's property is an int [16:28] So I thought I'd just add an Array.isArray check before we look at length [16:28] It's false, but hasChange is still getting set to true [16:29] ok looking [16:30] Should I push my changes up so you can see them? [16:30] Or would you prefer something like a gist? [16:32] well I'm just trying to figure out what this is doing [16:32] This screenshot pretty much summarizes my confusion: http://cl.ly/image/20132A2e051Z [16:33] Both conditions for the if statement eval to false, yet we're inside and hasChange is set to true. [16:35] that might be an issue with how watches are evaluated [16:35] what if you run those in the console? [16:36] and you'll want to use Y.Array.isArray() I'm not sure if Array.isArray() is supported everywhere yet [16:36] what is this even testing for? What makes it a major change? If any of the objects have something in them then major change is true? [16:40] thanks for the review, Makyo. :) [16:41] hatch: we're using an object as a hash here. The original code makes an assumption that all the values in the hash are arrays of changes. If any of the arrays have a non-zero length, then we have major changes and need to display the summary panel in addition to the changelog. [16:42] ok but that basically means that if there is anything in changes then we need to show it...what tasks aren't shown? [16:42] what I'm getting at is this method isn't necessary [16:42] because if you do anything it will return true [16:42] if any of the changes keys have values it's true [16:44] jujugui: i need 2 reviews on this (very short) PR. https://github.com/juju/juju-gui/pull/491 [16:44] But not all of the keys in the hash are arrays of changes. [16:44] jcsackett no test? :) [16:45] hatch: nope--there's no meaningful way to get in and test the d3 logic. [16:45] kadams54 right - but what I'm saying is that this method is causing you issues.....I don't think this method is necessary because if there is anything in the changes object then it's shown [16:45] i can break out the logic about what to show into a separate method and test that, if you like. [16:45] hatch: not necessarily the case. [16:45] jcsackett well that's just not true :) we test the unit list somewhere [16:45] kadams54 ok what is the case where it isn't? [16:46] hatch: for example, if the user just adds a container, we only show the changelog. We don't show the summary panel. [16:46] hatch: we set up some stuff using the generateAndBind, but we don't inspect those contents. [16:46] In that case, all of the arrays in the changes hash would have a 0 length. [16:46] i.e., no "major" changes [16:46] hatch: which is reasonable, b/c it's just testing "does d3 set an a tag when i tell it to set an a tag. [16:47] hatch: as i said, if you like i can break that change out into a "chooseDisplayName" method for units, and test *that*. [16:47] jcsackett I'm pretty sure we have a test which renders the unit list and then checks the dom to make sure there are x number of units rendered [16:47] maybe inspector overview or something [16:48] kadams54 ok and you have added something to changes.addUnits? [16:48] Yeah… totalUnits… [16:48] hatch: that's where i looked. unless you [16:48] Which is an int [16:48] hatch: unless you're talking the scale out tests, which arne't qute the same kettle of fish. [16:50] hatch: sorry, I didn't add anything to changes.addUnits. I added it to changes. So changes now has changes.addUnits and changes.totalUnits. [16:50] kadams54 ok so you could have getChanges return an object where one property is the changes and the other is the total units [16:50] then you don't have to modify this method [16:51] which is what I would recommend because your putting stuff in the changes property that doesn't really fit with the rest of the dataset [16:51] OK, will do [16:51] sounds good? [16:51] ok cool [16:52] jcsackett I'm just saying it would be nice if we could test that when we pass an uncommitted model into the d3 stuff it renders the proper name [16:52] and the other way [16:53] it was what the bug was about afterall [16:53] hatch: i'm not disagreeing, but the test for that will be ridiculous. like i said, i can break out the logic part of it into another method, and test that it returns the right thing; after that it's a matter of believing d3 does the right thing or not. [16:53] I'll try and find the test [16:53] you may be able to copy it [16:53] one sec [16:53] but I agree if you have to write it all [16:53] no good :) [16:54] i'm reading through inspector_overview and not seeing anything that looks right. [16:56] jcsackett test_inspector_overview.js:477 [16:56] the test is massive but you can likely modify it to use one uncommitted unit model as well as the regular others [16:56] or add another unit to the list [16:57] hatch: i'll see--i don't think the return from that is quite right to see what's rendered in the lists, but you might be right. [16:57] oh, i see, it's dumping the whole damn mess. [16:57] ok. [16:58] so you should be able to just add an uncommitted unit to this and test that that unit is rendered properly and that another one is rendered property too [17:09] * rogpeppe is done for the day [17:09] g'night all [17:09] hatch: done. [17:10] thanks for helping sort out where to test it; that block is impenetrable when you're scanning without being certain what you're looking for. :p [17:11] haha yeah it is np [17:11] ok will qa [17:11] jcsackett lint error [17:11] lol [17:12] we should make a rule - if your tests fail with a lint error you got to buy everyone a coffee :P [17:13] i honestly am not sure of the utility of lint being run on every test--on merge, sure, but it's just noise to me. [17:14] I always run the full suite before pushing - and now that I'm in Ubuntu the tests don't fail randomly lol [17:15] hatch: i always run the tests; i don't always run check. [17:15] oh right cuz it's really slow for you isn't it? [17:15] hatch: sufficiently slow. [17:16] slow enough that i end up waiting a while, not long enough to do anything meaningful while it runs. [17:19] time for an ssd? [17:29] hatch: looks like huw needs second reviews on a couple PRs? [17:30] kadams54 the x uncommitted one I need to write the fix for it to get it to land [17:30] but the mv scale up one does [17:30] is yours all good to go? [17:31] looks good [17:31] once the tests pass you can land it [17:31] Yeah [17:31] I was noticing we had a build-up of cards in review [17:32] I'll dig into the mv scale up one. [17:34] Looks like the scale up one needs to be rebased and have some conflicts resolved === urulama is now known as urulama-afk [19:06] telcos should have a flag on peoples accounts which mark their level of expertise [19:07] so when you call with a problem they don't ask you if you have restarted your router and make you do it again for kicks [19:09] jcsackett can you rebase your branch and shippit? [19:10] #491 that is [19:19] I actually called Comcast once after I did some basic troubleshooting. I had forgot to reset my cable modem. [19:19] The support person saying "Turn it off and back on again" actually solved my problem. [19:20] lol [19:30] jrwren: config-changed error: http://pastebin.ubuntu.com/8029262/ [19:32] jrwren: relations in charms: http://pastebin.ubuntu.com/8029269/ [19:33] urulama-afk: that doesn't look like an error. :( [19:34] error: no relation id specified [19:34] maybe it's enough to trigger the hooks and to leave the charm in "installed" but not "started" state [19:35] urulama-afk: I'll try to reproduce. Thanks. [19:36] urulama-afk: I'm still new enough to charming that I don't remember all this stuff. I can't recall if that error: is valid or not. [19:36] jrwren: me and you both :) [19:36] jrwren: Still, and invalid error message is a bug, so I'll fix it. [19:36] s/and/an/ [19:37] jrwren: yes, i can exit the tmux and then the status is "started" [19:37] urulama-afk: can you curl that ip on the listen port? [19:41] jrwren: it gets stuck on most of the hook scripts, but running them manually and exiting tmux produces service in a "started" state, and curl works [19:42] hrm. it gets stuck and you have to resolved --retry? [19:42] jrwren: i mean, the CS is empty, but there are "error" codes to show that it works [19:42] urulama-afk: that is VERY weird. [19:43] urulama-afk: Empty and responsive is my goal. === urulama-afk is now known as urulama [19:43] jrwren: i haven't tried deploying without debug-hooks, let me add another unit, it might just as well work [19:49] jrwren: indeed ... seems like debug-hooks interpret your error messages as true errors ... without them, the charm starts and is in "started" mode [19:50] urulama: Good, that is what I thought. [19:50] you guys ok with all this? I've done some charm work, can maybe offer some input [19:51] jcsackett thanks [19:51] jrwren: ok, so fixing that is simple, just changing error to "warning" or something, as they are not true errors [19:51] hatch: thx [19:51] hatch: no worries; hadn't noticed i had all my reviews for it. :) [19:51] thanks for pinging me. [19:51] urulama: its charmhelpers. [19:52] jrwren: khm. then it might not be that innocent after all :) [19:52] hatch: What is the right way to use charmhelpers.core.hookenv.relation_get ? [19:53] what's this python mumbo jumbo...I just shell out to the real relation get [19:53] https://github.com/hatched/ghost-charm/blob/master/hooks/db-relation-changed [19:54] so sorry, I have never used charmhelpers :) [19:54] hatch: I'm pretty sure I was told to use this. [19:54] yeah you likely were [19:54] sometimes the abstraction is more work though [19:54] case-and-point :) [19:54] hatch: Are you sure it is only sometimes? :) [19:55] haha [19:55] jrwren you'll probably have some luck asking in #Juju-dev [19:56] hatch: do not lead jrwren into temptations! :) [19:57] It is likely as simple as if not relation_ids(): before those relation_get calls' [19:57] urulama haha sorry - I've written charms in node and bash, and fixed the gui one which is in python so never really used charmhelpers and the like [20:03] hatch: i deploy charms from CLI, that doesn't mean GUI is useless :D [20:04] urulama definitely not, after about the 1000th time of writing `juju set some-really-big-charm-name some-really-big-config-name="some really long value"` you'll just use quickstart and use the GUI exclusively [20:04] lol [20:05] :D :D [20:06] ok, time to really EOD ... see you all tomorrow, jujugui (and irc clients go 'ping' all over the globe) [20:06] later urulama [20:06] urulama: good night. TY. [20:06] haha cya urulama === urulama is now known as uru__ [20:42] jrwren: ask in #juju of marcoceppi or mbruzek around that. Maybe tvansteenburgh as well. They work on and maintain the charmhelpers [20:42] hatch: helps us use/make better our own tools :P [20:43] o/ [20:43] rick_h__ I'm not sure anything that's 3 levels deep for a simple shell-out call is a valid abstraction :D [20:43] hatch: time to make it better? [20:43] hatch: shell out with proper catching of return values, logging integration, etc can be worth it [20:43] however I have a lack of context in this matter so I wave my hands around in a general fashion [20:43] honestly I have no idea what it does under the hood, never used it :) [20:44] ask me what? [20:44] mbruzek: jrwren is having some issues with relation-get and charmhelpers or the like [20:44] mbruzek: just volunteering you for support of jrwren gets stuck :) [20:44] * mbruzek nods [20:47] maybe I should write a charm in python to learn about these helpers-of-the-charm [20:51] hatch: ALL HAIL THE HELPERS OF THE CHARM [20:51] hatch: If you are a cranky programmer like me, it might just annoy you that it isn't in bash. ;] [20:52] jrwren bash MAKES a cranky programmer [20:52] marcoceppi lol [20:52] rick_h__ you still around? [20:52] * jrwren uses /dev/tcp in bash instead of curl [20:52] lol [20:54] jcsackett Makyo kadams54 want to weigh in on https://github.com/juju/juju-gui/pull/486#issuecomment-51975411 ? [20:54] hatch: sure [20:55] hatch: packing up, flight home in the morn [20:55] rick_h__ want to weigh in on that? ^ [20:55] hatch: lgtm [20:55] hatch: looking [20:56] hatch: "update a template which doesn't exist" ? [20:57] rick_h__ the real issue is that it's not the template which doesn't exist, it's that it's using the old inspector rendering code [20:57] so we have to write a new test mock for generating the inspector to use the NEW mv style inspector rendering code [20:57] which has to be done anyways but will block this branch [20:58] hatch: that seems like something we'll need anyway? (a new helper for generating a new inspector) [20:58] hatch: ok, but this comment is about waiting until MV is released, not about a missing test helper? [20:58] nothing here is documenting/adding that we need the helper as the 'right' fix? [20:58] yeah sorry that wasn't clear [20:58] or am I mis-reading that? [20:59] * rick_h__ loads up kanban [20:59] so this branch either needs to wait until the new inspector mock test thingy is written or apply the included patch to land it now [21:00] I gather the new inspector mock thingy will take 1day and then moving all the tests to use it will be....2? [21:00] hatch: ok, I'm cool with the patch with a proper XXX, documenting the place the needs updating, and a follow up card at the top of the lane [21:01] ok I'll work on the inspector thingy, I added a friday card to chat about it [21:01] thx [21:01] hatch: coolio, thanks. [21:01] safe travels! [21:02] wheeeee [21:02] can't wait to have my own pillows again [21:02] lol [21:02] it's the small things, right? [21:03] you know it [21:03] like AC that works [21:03] getting sick of always sweating in EU [21:03] haha, I never would have thought that would be an issue [21:04] us norhern folks don't like it so hot, my basement office is calling me [21:05] haha [21:05] I'm with you on that one! [21:15] Makyo good choice in cards :) [21:15] well I'm out, have fun all. See you soon. [21:15] cya rick_h__ safe travels [22:06] jujugui lf a review (no qa) https://github.com/juju/juju-gui/pull/492 [23:03] Morning [23:21] morning huwshimi [23:24] huwshimi all reviews are done and comments made so you should be able to land both of your branches [23:24] today [23:24] and if you could also do a quick review on mine that would be great because mine provides the hooks you will need to mark up the UI with the 'deleted' attributes [23:24] ^ huwshimi [23:25] huwshimi: Good morning. [23:34] My wife and child just reminded me that today is my last day living at the age of 36 years old. [23:35] WHAAAAA [23:35] I thought you were like.......26 [23:35] lol [23:36] Whaaa??? [23:36] haha yeah [23:41] hatch: Yep no problems, just doing a little qa [23:52] hatch: So to use that flag would we have to update this to use lazyDestroyMachines? https://github.com/juju/juju-gui/blob/develop/app/widgets/machine-view-panel.js#L710