frankbanhi hatch 13:52
frankbanrogpeppe: I am encountering 8 failures in juju-core trunk tests, all related to uniter_test.go13:54
frankbanrogpeppe: and hi :-)13:55
rogpeppefrankban: hiya13:55
rogpeppefrankban: you're best raising this in juju-dev, i think. i'll just check that i see the same thing first though.13:55
frankbanrogpeppe: thanks13:55
frankbanMakyo: good morning, re annotations branch, we are almost there :-)13:56
Makyofrankban, \o/13:57
MakyoI will need at least one more coffee before I can understand Go.13:57
rogpeppefrankban: uniter tests pass for me in trunk13:58
rogpeppefrankban: could you paste me your errors?13:58
frankbanrogpeppe: trying, re-running the whole suite, failures include lots of debug messages13:59
frankbanrogpeppe: http://pastebin.ubuntu.com/5625282/14:04
rogpeppefrankban: interesting. i'll pass it over to dimiter and william, who've been working on the uniter stuff14:05
frankbanrogpeppe: thanks14:05
gary_posterrogpeppe, hi!  Exciting seeing the allwatcher branches move through.  On a mostly unrelated topic, we will want the GUI to expose upgrade charm from both pyjuju and juju-core.  AFAIK it is not yet implemented at all in juju-core, even as a command.  Am I right?  If so, do you know who will be responsible for that implementation?  I have some questions from the designers that I'd like to get answers for (http://paste14:09
rogpeppegary_poster: in a meeting currently14:10
gary_posterrogpeppe, ack np 14:10
gary_posterjujugui call in 2 in guichatr14:27
benjigary_poster: my browser doesn't want to connect to the hangout; I'll be there as soon as possible14:30
gary_posterack benji14:31
rogpeppegary_poster: dimitern is implementing charm upgrading. i think it's going pretty well14:35
gary_posterrogpeppe, great will ping him later14:35
gary_posterbenji,  how goes set_constraints14:35
rogpeppegary_poster: i'd still be interested in any comments you have on the allwatcher reviews, BTW. i'm going to be submitting https://codereview.appspot.com/7815044/ shortly, but if you want to take a look, i'll hold off for a while.14:36
benjigary_poster: pretty good; once I figure out this import loop it should be smooth sailing14:36
gary_posterok great thanks both14:36
* benji reboots while giving the Google Hangouts plugin the evil eye.14:37
rogpeppefrankban: i'm presuming you saw those uniter test failures on the latest version of trunk (1012) ?14:41
frankbanrogpeppe: yes, running the suite again using 101314:42
rogpeppefrankban: perhaps you could liaise with dimitern on #juju-dev to try to resolve this, as you seem to be able to reliable reproduce the issue.14:43
rogpeppefrankban: i'm going for lunch now14:43
frankbanrogpeppe: I will, thanks14:43
andreas__hi, is this a known bug? http://i.imagebanana.com/img/2sphe691/Selection_001.png14:46
andreas__the clutter, and the fact that there are no relations drawn14:46
=== andreas__ is now known as ahasenack
ahasenackif I drill down on a box (double click), the relation is displayed14:47
Makyoahasenack, Not seen that before.  What browser and version of the gui?14:51
ahasenackMakyo: chrome, gui "stable" according to the charm14:52
ahasenackjust deployed14:52
Makyoguihelp ^^^14:52
ahasenackGoogle Chrome25.0.1364.172 (Official Build 187217) 14:52
Makyoahasenack, Alright.  Give me a sec take a look on my side real quick14:52
gary_posterahasenack, never seen.  Makyo thanks for looking.  A URL for Makyo would help if you can share it ahasenack 14:54
ahasenackgary_poster: I don't think I can, it's deep within a private network14:55
gary_posterahasenack, ack14:55
Makyoahasenack, Is there anything in the JS console? Ctrl+Shift+J14:55
ahasenackMakyo: hm, yes14:55
ahasenackMakyo: http://pastebin.ubuntu.com/5625372/14:56
ahasenackhttp://pastebin.ubuntu.com/5625378/ when I reload the page14:58
Makyoahasenack, Ah, thanks.14:58
hatchMakyo: I am thinking that might be caused by views/landscape.js ln 1914:59
* hatch grumbles because he can't get the IE tests to fail again14:59
Makyohatch, Yeah, that's the exception.  Looks like  we're passing undefined to it from getLandscapeURL, though.  Digging into that.15:00
ahasenacklandscape itself might be having problems in sending down some annotations, we have a bug open to track a traceback15:02
Makyoahasenack, Yeah.  From the looks of it, the services do not have a landscape-computers annotation coming from juju (app/views/landscape.js:123 for those keeping track) which is causing the drawing of services to fail part-way through.  There should be a check on that annotation in there as well.15:04
ahasenackMakyo: ok, so that is breaking the rendering15:06
Makyoahasenack, Yep.  The services being cluttered is a separate issue, FWIW, but that is what's stopping the services from rendering completely with the health graphs, proper name placement, etc.15:06
MakyoFiled #1156662, will do a quick branch to at least add the check.15:10
_mup_Bug #1156662: getLandscapeURL is not checking annotations on models other than environment <juju-gui:New for makyo> < https://launchpad.net/bugs/1156662 >15:10
hatchMakyo: I wonder if we shouldn't notify the user that there is a problem?15:11
ahasenackMakyo: thanks15:12
gary_posterahasenack, the services being cluttered is something that the user is supposed to be able to address by dragging them around.  Not ideal, but workable, and only painful initially (the service locations are shared and persistent)15:13
Makyohatch, Yeah, we don't have any requirements yet around that story.  Any suggestions, guihelp, ahasenack, goodspud?15:13
hatchalert('Landscape broked - tell boss');15:13
ahasenackwell, I will have to see how it looks when the rendering is complete15:14
bacgary_poster: lp2kanban is having buildout issues.  can you quickly diagnose it?  https://pastebin.canonical.com/87040/15:20
hatchMakyo: do we have a story for warnings in general? something akin to http://twitter.github.com/bootstrap/components.html#alerts15:20
bacgary_poster: if not i'll dig into it in a bit, after getting my real branch re-proposed15:21
gary_posterbac afraid completely booked15:21
bacgary_poster: ok, p15:21
Makyohatch, two: notifications and alert widgets (though the latter are generally used in a confirmation sense).15:21
Makyohatch, You can try removing a subordinate relation to see one used in a warning sense, though.15:22
hatchahh ok - I wonder if it makes sense to use that type of warning?15:22
Makyohatch, Perhaps.  While it's a good indicator that something's fairly wrong, we've been using notifications for something wrong with the data we've received.15:23
hatchahhh yeah I suppose it's not an issue with the gui itself but with the data being received15:24
Makyoahasenack, Is a screenshot helpful? http://ubuntuone.com/3KzgiaR1qBqjWpv2Sj4D3r  If you need more, I can see what I can do.15:26
hatchjcsackett: are you around?15:39
benjirogpeppe: I have an import loop that I'm having problems solving; here is the diff that generates it: http://paste.ubuntu.com/5625502/15:42
ahasenackMakyo: what is that screenshot about? That 1 alert?15:42
rogpeppebenji: it might be worth compiling against go tip - the go tool should tell you what the import cycle is.15:43
Makyoahasenack, The completed rendering.15:43
rogpeppebenji: if you don't want to, paste us the branch and i'll do it.15:43
benjirogpeppe: any tips on how to install go tip?15:43
gary_posterahasenack, I think he was replying to your "I will have to see how it looks when the rendering is complete."  You'll want to see it in full QA of course15:43
ahasenackMakyo: ah, sure, it looks fine. I meant that I would have to see how mine looks, with so many units15:43
Makyoahasenack, Ah, okay.15:44
gary_posterahasenack, how many services is that?15:44
benjiis installing tip reversable?  will it lead to other problems working on go code if the others on the team aren't using tip?15:44
ahasenackand how hard it will/would be to unclutter it15:44
ahasenackgary_poster: 29 or so15:44
rogpeppebenji: i install go tip inside $HOME/go; and yeah it's reversible by simply deleting that directory.15:44
gary_posterahasenack, that's at the top limit of what we say we support.  You should still be able to get it to work.  Does that collection represent some real-world-ish use case?15:44
benjirogpeppe: cool, is there a web page that describes how do do that?15:45
rogpeppebenji: assuming you've got mercurial installed, hg clone https://code.google.com/p/go/; cd go/src; ./all.bash15:45
ahasenackgary_poster: yes15:45
rogpeppebenji: should do the trick15:45
rogpeppebenji: one mo, i'll look15:45
benjirogpeppe: cool, I'll give it a try15:45
gary_posterahasenack, cool.  Do you mind telling me what the use case is?  It would be good for us to document a real world use case of that size15:45
ahasenackgary_poster: openstack deployment15:46
rogpeppebenji: you *might* have to do 'hg update tip' before building15:46
rogpeppebenji: oh yes, you'll need to set $GOROOT to the root of the go source tree15:46
gary_posterahasenack, wow!  I thought that was under 10.  Maybe once you have that running with proper rendering I can ask you for a screen shot so I can see how they all interconnect.15:47
rogpeppebenji: i regularly use two copies of go root - one running 1.0.2, one running tip15:47
ahasenackgary_poster: yeah, I have more than one unit per service here15:47
rogpeppebenji: here's the shell script i use for switching to use 1.0.2 http://paste.ubuntu.com/5625516/15:48
gary_posterahasenack, you use multiple units within a service15:48
ahasenackgary_poster: like 4 compute nodes, for example15:48
rogpeppebenji: it's worth having a separate GOPATH to use with the separate go version, i think.15:48
gary_posterahasenack, so unless there's something unusual with what you are doing, the compute node should be a single service15:48
gary_posterahasenack, with 4 units15:48
ahasenackgary_poster: and each unit is one computer15:49
gary_posterahasenack, yes15:49
ahasenackopenstack has many services15:49
rogpeppebenji: i generally develop against tip and test against 1.0.2 before proposing (or when i want to run tests while i'm switched to another branch)15:49
gary_posterahasenack, right but < 10 yeah?15:50
frankbanMakyo: ping15:50
rogpeppebenji: ah, here are the official installation instructions: http://golang.org/doc/install/source15:50
rogpeppebenji: they basically amount to what i said, i think15:50
Makyofrankban, Hey15:50
gary_posterthen if you expand them out, you expand them *within* the services.  Am I missing something basic about what you are doing?  Should we have a call instead?15:50
benjirogpeppe: thanks, that looks helpful15:50
frankbanMakyo: could you please branch lp:~frankban/juju-core/ann-doc/ and check all tests pass?15:50
Makyofrankban, yep, on it.15:51
ahasenacklet's see, compute, quantum, dashboard, keystone, rabbit, ceph, cinder, glance, juju-gui, mysql, cloud-controller, swift-proxy, swift-storage15:51
gary_posterahasenack, ok cool, 13 is still in my rough mental model :-)15:52
ahasenackjuju-gui not strictly being openstack, I was just going over juju status15:52
gary_posterahasenack, ah, cool! so you have 13 services, using 29 machines using some distribution or other?15:53
frankbanMakyo: thanks15:53
gary_posterI mean, a distribution so that compute nodes have 4 machines, and dashboard gets 1, and so on15:53
ahasenackgary_poster: yes, most have more than one unit15:54
ahasenackotherwise it wouldn't be fun, that's the power of juju15:54
ahasenackI also have landscape-client on each unit as a subordinate15:54
gary_posterahasenack, cool, ok.   agreed.  13 services--ok, right, 14 with Landscape-- shouldn't be too bad to arrange at all.  We have something like that for our own demos.  If you arrange it before the demo it can even look pretty when the audience sees it. :-)15:55
gary_posterthanks ahasenack.  so your work is done until Landscape engineers add the missing annotation, right?15:56
gary_posterI mean the work for QAing this, of course :-)15:56
ahasenackgary_poster: yes, in fact I have a workaround to try in landscape that someone just gave me, I'm about to try that15:56
gary_posterahasenack, excellent!15:56
ahasenackgary_poster: thanks for the help15:58
gary_posterahasenack, np, thank you15:58
hatchso our city had a massive watermain break Friday which caused the city to issue a 'boil water' warning - no big deal except that all of the coffee shops are shut down because the water in their machines is piped right in from the city water16:06
Makyofrankban, Everything but uniter_tests.go passes.16:13
frankbanMakyo: could you please try the same in the master branch (most recent trunk)?16:13
Makyofrankban, yep.16:13
frankbanMakyo: thanks again.16:14
hatchw00t w00t - found the CI issue16:15
* hatch does the happy dance16:15
frankbanhatch: cool! what was the problem?16:20
Makyofrankban, same failures in trunk.16:20
frankbanMakyo: :-/ 8 failures?16:20
hatchfrankban: missing dependencies which would be auto resolved on our local machines16:21
hatchben and I are looking into the solution now16:21
Makyofrankban, Yep.16:21
benjirogpeppe: I have go version devel +43eb97ed849a but am still getting the standard terse import cycle error.  Is there a switch I have to throw to get more info?16:21
rogpeppebenji: hmm. it might depend where in the import cycle you're compiling :-|16:22
* benji laughs an acidic laugh.16:23
rogpeppebenji: beforehand i remember sometimes not being shown the cycle, but i can't remember when or why16:23
rogpeppebenji: could you push your branch and i'll take a look?16:23
rogpeppebenji: when i add the state import to api/params, i see this:16:25
rogpeppecan't load package: package launchpad.net/juju-core/state/api/params16:25
rogpeppeimports launchpad.net/juju-core/state16:25
rogpeppeimports launchpad.net/juju-core/state/api/params: import cycle not allowed16:25
rogpeppebenji: so i see where your problem is now16:26
frankbanMakyo: can you confirm tests in state/. pass?16:27
frankbanMakyo: (in the ann branch)16:27
rogpeppebenji: ok, i have a solution for you16:28
Makyofrankban, Yeah, they pass.16:29
frankbanok, Makyo: maybe it's time for some pairing, let's use termbeamer, what's your jabber/gtalk account? could you add me please?16:30
frankbanMakyo: mine is frankban@gmail.com16:31
Makyofrankban, drab.makyo@gmail.com16:32
rogpeppebenji: suggestion: from the juju-core root: mkdir constraints; bzr mv state/constraints*.go constraints; fix everything in state to import the new package and use constraints.Constraints not state.Constraints.16:35
rogpeppebenji: modification: as above, but rename Constraints to Value along the way.16:38
rogpeppebenji: ping17:04
=== matsubara is now known as matsubara-lunch
benjirogpeppe: (I was eating lunch while you guys figured that out.) cool, thanks17:15
rogpeppebenji: ah, i wondered17:16
rogpeppebenji: np. does that seem reasonable?17:16
benjirogpeppe: yep I think it's reasonable; if I find out otherwise I'll be sure to let you know ;)17:17
rogpeppebenji: :-)17:17
hatchjujugui if anyone has a moment for a 5 line diff review :) https://codereview.appspot.com/7621045/17:25
benjihatch: I do17:26
bacgary_poster: http://stackoverflow.com/questions/14801416/zc-buildout-stopped-working-importerror-no-module-named-apport-fileutils  << this was the problem17:28
gary_posterack bac.  sorry it took time17:29
hatchgary_poster: remember when I was saying guys can get huge air kiteboarding? This takes it to a whole other level http://www.youtube.com/watch?v=3JWi78hk8oQ :)17:39
gary_posterhatch, :-) cool will look in a few17:39
* benji enjoys the very exciting thunderstorm going on here.17:42
ahasenackgary_poster: the workaround worked, I see the relations drawn in the juju-gui "picture" now. I'll look for the landscape stuff now17:47
gary_postergreat ahasenack!  If you don't see "Landscape" logo in the bottom right then it's not showing up.  It's pretty obvious.17:48
ahasenackgary_poster: I see it17:48
gary_posteroh, great17:48
ahasenackgary_poster: aren't the words reversed?17:48
ahasenackgary_poster: I read "environment\nlandscape"17:48
gary_posterahasenack, not clear what you mean yet, sorry.  screenshot?  Or maybe the one Matt gave earlier could be used to make your point?17:49
* gary_poster has to go back to other computer to find the image17:49
ahasenackgary_poster: I mean, I would say "Landscape Environment", not "Environment Landscape"17:49
=== deryck is now known as deryck[lunch]
gary_posterahasenack, Landscape is not the environment...I think I know what you mean (from memory; not looking atm).  That is the link to the environment information within landscape--the information about the environment that landscape knows about.  If you double click on a service (or single click -> "View") then you will see that show as "Service"17:52
gary_posterahasenack, this is good feedback for our design people though.  Would it be too much to ask for an email about the things that confused you?17:52
gary_posterThe "Landscape" text is just indicating that this is Landscape-specific information17:53
ahasenackgary_poster: I know, it's a link to the juju environment in landscape17:53
gary_posterhatch, that's pretty insane :-)17:53
hatchHaha especially considering those kites are designed to fly lol17:54
gary_posterhatch, not designed to, you mean,right?17:54
ahasenackgary_poster: isn't it better to file a bug? It can be closed if they don't agree17:55
ahasenackI do find that wording incorrect, though: "environment LANDSCAPE"17:55
ahasenackmakes no sense :)17:55
hatchgary_poster: oh yeah, definitely -not-17:55
gary_posterahasenack, +1.  I would fill the bug out for you, but I don't understand the position yet--perhaps stared at this too much :-)17:56
gary_posterso your explanation--why it seems obviously incorrect--would be valuable17:56
ahasenackgary_poster: bottom-right corner: http://i.imagebanana.com/img/wq182eee/Selection_001.png17:56
gary_postercool ahasenack, right, thx.17:57
ahasenackgary_poster: bug coming up17:57
gary_posterhatch did that IE fix make everything better, or just was incremental improvement?17:58
hatchBen is still working on the timeout bug but this should fix the IE bugs - I'm just about to trigger another build to test17:59
benjirogpeppe: Moving the Constraint.Value methods won't work because they take state.State arguments, inducing a new import cycle.  Should I just move the data type only and leave the methods in the state package?18:10
rogpeppebenji: looking18:10
rogpeppebenji: i think those functions (create, read and writeConstraints) should remain in state, as they're directly to do with mongo interactions18:11
rogpeppebenji: the Constraints type and its methods can (i think!) move ok18:12
benjirogpeppe: makes sense, thanks18:12
rogpeppebenji: in fact, everything from 'type constraintsDoc' onwards should remain in state18:12
=== matsubara-lunch is now known as matsubara
hatchbcsaller__: it looks like the sauce labs IE instance is caching stuff - this last run had the same issue we see if we don't clear IE's cache18:18
bcsaller__hatch: I saw that on Thursday as well. It didn't make sense as it should be a new VM but they might go through a caching proxy 18:19
hatchyep - well I'm going to grab some lunch and then I'll investigate this when I get back18:19
hatchjust wanted to let you know where it left off18:19
hatchbefore I go - anyone know how to make a newline on the wiki?18:21
gary_posterhatch, new lines, I think? :-)18:22
benjithe go import system leaves something to be desired; a name like "constraints.Value" makes sense for all packages except the one in which it is defined (where it is simply named "Value" which is somewhat informationless).18:24
benjiI guess that's the same problem that the Python import system has; I wonder why I don't remember tripping over it there; different naming conventions perhaps?18:25
hatchgary_poster: hah I tried that, I ended up just putting a line between the other lines....double space :)18:26
rogpeppegary_poster: next branch here - in the next branch we'll actually hook it up to the API! https://codereview.appspot.com/765004818:29
gary_posterhatch, heh.  I18:33
gary_poster'll look at it if you want.18:33
gary_posterrogpeppe, awesome!18:33
bacguihelp: anyone have a moment for a review?  https://codereview.appspot.com/787104418:59
gary_posterbac, on it18:59
Makyobac, On it.18:59
MakyoToo late!18:59
bacwow!  such enthusiasm19:00
baci love that we had a bug where we try to beautify prettify.js19:00
bacor, better that prettify needs beautifying19:01
gary_posterbac LGTM with trivial19:04
bacthanks gary_poster19:04
gary_posterbenji did you get unblocked with circular import?19:06
benjigary_poster: yep; I just had a 100% passing full test run; now I just have to write the code to do the actual thing the branch is supposed to do ;)19:06
gary_posterbenji, heh, cool-ish ;-)19:07
benjiI'm contemplating posting the branch as-is for review.  (The combination of cobzr and lbox make me shudder at the thought though.)19:07
bachatch: i've got to pass a context, and since it isn't used we decided a week or so ago that null was a pretty good choice.19:08
bachatch: is there a better option?19:08
hatchsorry - I meant to say, why do you need to pass a context at all19:08
hatchwhy can't you just call the method?19:08
gary_posterhe's making a partial without making a closure19:09
baci'm using Y.bind to curry the values of the other params19:09
bacyeah, that19:09
gary_posteryeah, that19:09
hatchoh ok then - could you comment as such? That's not a very common pattern so someone may come by and remove that in the future :)19:10
gary_posterhatch, what's the way you would have done that?19:10
hatchyeah I would have done a closure - but I'm going to guess that this method does need to be called externally at some point?19:11
gary_posterjust trying to keep code divided up nicely19:12
gary_posteryou could still do that with a closure19:12
gary_posterwrapping the other function19:12
=== deryck[lunch] is now known as deryck
gary_posterthis._send_rpc(..., function(data) {this.handleRemoveRelation(callback, endpoint_a, endpoint_b, data); })19:14
bachatch: that function is commented as: // Curry the endpoints.  No context is passed.19:14
bacand, it is a common practice.  i did it 8 times in that one module.  :)19:14
* hatch bows his head in shame19:14
gary_posterhatch, why?19:15
bacgary_poster: are you proposing that change?19:15
gary_posterbac, no.  you'd have to push the if in there too19:15
gary_posteror something19:15
hatchI read the comment and it didn't click :)19:15
gary_posteroh :-)19:15
hatchok then just change the names to camelCase :P19:15
gary_posterthis is legacy support19:15
bachatch: as to the camel case, the existing call sites use endpoint_a and endpoint_b.  i'd rather not make the change and have a mismatch19:16
gary_posterwe need to do a mass remove_relation -> removerRelation19:16
gary_posterwhich we can do19:16
hatchOK FINE!!!!19:16
gary_posterbut separately :-)19:16
bacremove_relation -> destroyRelation19:16
hatchI'll put LGTM on it19:16
bacdestroy is the canonical name.  remove is an alias.19:17
gary_postersorry for weighing in.  I think I might be procrastinating.19:17
hatchbcsaller__: ok now that all the tests pass in the CI it's all good but for some reason it's still timing out even though it hits 100% then just sits there19:18
hatchI think we need a 'complete' flag or something19:19
bcsaller__hatch: you're not seeing a request to login in the video? Thats why the timeout19:19
hatchnope - it's just sitting here until the timeout hits19:20
hatchgotcha - that's the part you were mentioning about earlier19:20
hatchso is there anything else that I should be doing on this? Or is the rest in your wheelhouse?19:21
bcsaller__hatch: I think I can handle the last step19:23
hatchok great - I'll finish up these CI docs then - lemme know if you want to bounce any ideas around19:23
hatchjujugui I need one more review for the CI documentation https://codereview.appspot.com/7678048/ anyone have a few minutes?20:00
* gary_poster did first one :-)20:00
Makyohatch, sure.20:01
* BradCrittenden dang20:01
=== BradCrittenden is now known as bac
gary_posterhatch, I'm going to try to give you a patch to hook up with docs, gimme a bit20:03
bacinteractive testing of the go api with firefox is a real pain...but very gratifying when it works20:06
gary_posterhatch if you apply http://paste.ubuntu.com/5626260/ (in particular the docs/index.rst change) then you will be able to run make view-docs and go to the new doc from the (non-YUI) index page20:12
* benji wades through 2,154 lines of test runner output.20:17
* gary_poster , knowing benji, knows that this is not hyperbole20:17
hatchgary_poster: sure I'll make these changes then run the make20:21
* bac dogwalk20:21
gary_postercool hatch.  The only point of running the make is to show you that it works generally, works specifically with your new file, and might be useful. :-)20:22
* benji wishes it were.20:22
benjiThe whole "hey one of your tests failed, here is the unrelated output of everything that has ever been logged in the history of the universe; you're welcome" thing is not so great20:23
hatchgary_poster: that worked well - the docs actually look pretty good *brushes shoulder like a boss*20:25
gary_posterhatch, they do :-)20:26
hatchok changes have been pushed20:28

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!