[01:00] yo thumper [01:11] hi bigjools [01:11] you wanted me earlier? [01:12] yeah, perhaps chat about maas [01:12] but I need to shower first [01:14] hmm... my magic change hasn't worked [01:14] I think I know why it failed... [01:14] but need more investigation [01:22] thumper: i'd like a catch up too [01:34] wallyworld_: give me 5 minutes to grab these log files. [01:35] ok [01:46] ok, ready [01:46] but confused [01:46] wallyworld_: shall I start a hangout? [01:46] ok [01:46] https://plus.google.com/hangouts/_/98238f3ad80fb6e5ea1e7c6e53442bef8dd68df6?hl=en [02:00] davecheney: "i can tell that you and I are going to get on like a house on fire in BNE" - is it ok if I laugh a lot? [02:01] yes [02:01] that would break the ice [02:01] :) [02:01] opp [02:01] changing host [02:01] two secs [02:03] aaaand, back [02:03] wifi, how does it even work / [02:03] ? [02:03] it uses the tears of developers [02:04] apparently not enough tears in my house today [02:04] also pixie dust [02:04] and unicorn vomit [02:04] -- that is what someone described the plague of frozen yoghut stores that are springing up in Sydney [02:05] * bigjools lives in the boonies so doesn't see that kind of nonsense [02:05] everyone here complained when McDonalds got planning permission [02:06] isn't that just what australian's do ? [02:06] whinge? oh yes [02:07] we don't want airports, freeways, desalination plants, or mcdonalds' in our backyards [02:08] hey wallyworld_, I found out that the big empty space next to Coles here is going to get a huge department store and a Woolies! [02:08] the locals will go nuts :) [02:09] great [02:09] wallyworld_: and the phuqers put speed humps on that incline so I can't take you over it in the air any more [02:10] :-( [02:10] * bigjools considers going monster truck stylee [02:11] oh you kids, with your monster trucks and airborne shenanigans [02:13] davecheney: you made a series called 1.12 from a tag called 1.11.4? [02:14] yessir i did [02:14] how do you plan to work now, are you pushing revisions from trunk back to 1.12 or something else? [02:14] select revisions [02:14] possibly none [02:15] you plan on branching a new series from trunk at some point? [02:15] the reason for a stable tag is to give something to james page so that we can try to push that into backports for all series [02:15] ok [02:15] at the moment the devel versions are only going into aucys [02:15] saucy [02:15] ffs lag [02:16] the next stable will be 1.14, and I understand that a better local provider is the headline feature for that release [02:16] and azure [02:16] bigjools: yeah, i didn't want to overstep in my email [02:17] it's in our best interests to get Azure done ASAP [02:17] as I wasn't sure of the plan, but azure in 1.14 sounds good to me [02:17] right-o [02:17] and I'll be doing a handover when you're all in BNE [02:17] bigjools: so you know something about this sprint ? [02:17] since Red are moving to maas pastures [02:18] davecheney: I know that it's planned, and that's about it :) Trying to get mramm to get it organised etc etc [02:18] ok, things are going about as smoothly as usual then [02:18] anyone know what date ? [02:19] I heard same week as IoM sprint [02:19] but I don't recall from whom that came [02:20] bigjools: that's arse backwards [02:20] thumper: must be there [02:20] otherwise itt's a waste of time [02:20] I don't entirely agree with that [02:20] * bigjools shrugs [02:21] as much as I'd like to spend time with thumper [02:21] * thumper deputises wallyworld_ [02:22] thumper: axw is your boss [02:22] he should get to meet you in person sometime this year :) [02:22] guffaw [02:22] wat [02:22] davecheney: you get that the right way round? [02:22] :) [02:22] davecheney: and I thought that it was mramm, not me [02:22] anyway... [02:22] on noes, not the not double negative [02:22] thumper: you need to study the org chart more closely [02:23] trololol [02:23] bigjools: https://plus.google.com/hangouts/_/98238f3ad80fb6e5ea1e7c6e53442bef8dd68df6?hl=en plz join [02:23] * thumper pokes the directory [02:23] thumper: eek, hang on [02:23] 2 mins [02:24] davecheney: look at axw in the directory, manager mramm [02:25] thumper: who do I report to ? [02:25] if it says mramm [02:25] it's wront [02:25] it's wrong [02:25] davecheney: arosales [02:25] ok, i guess it's correct [02:25] :) [02:25] this is what I expected anyway [02:26] * axw is not sure where he's meant to be placed [02:27] davecheney: did you see my comment on the warning log change, regarding transient errors? [02:28] axw: i sort of did [02:28] i threw some code at you [02:28] i tried to fix the problem from the other end [02:28] i hoped that there would be a middle group [02:28] ground [02:28] *looks* [02:30] same sort of thing [02:31] the problem I realised later is that there may be ERROR logs that don't result in a command failure [02:31] I was playing with the local provider yesterday, and got a bunch of ERROR logs about not being able to connect to port... then after a while it connects and succeeds [02:33] I suppose those could be changed to be not-errors though, except for the final one if it gives up retrying. [03:19] meeting agenda for tonight, https://docs.google.com/a/canonical.com/document/d/1eeHzbtyt_4dlKQMof-vRfplMWMrClBx32k6BFI-77MI/edit# [03:21] wallyworld_: now, about simplestreams... [03:23] jtv: otp to bigjools, give me a sec [03:40] jtv: you rang? [03:43] wallyworld_: otp to bigjools, give me aq sec [03:43] heh [03:43] lol [04:01] wallyworld_: off the phone now. I don't know who this bigjools thinks he is, keeping us all from getting work done like that. [04:01] indeed [04:01] it's a rare skill I have [04:03] jtv: so what bit of my limited knowledge did you want me to share? [04:03] wallyworld_: don't worry, he doesn't mean the skill is rare. He means it's rare for him to have a skill. [04:03] pwned [04:03] I'm trying to invite you to a hangout, but your name is just too common... === tim is now known as thumper [04:04] * bigjools bends over [04:05] * wallyworld_ unzips [04:05] /everyone screams [04:05] bigjools screams loudest [04:06] * bigjools loves the family channel atmosphere [04:06] Good thing you two are alone on a separate continent. Now all we need to do is seal the exits and let you fight it out. [04:07] bigjools: you started it [04:07] wallyworld_: and you're going to finish it... oh wait you already did [04:07] ;) [04:15] wallyworld_: hmm... don't really want to use host network [04:15] wallyworld_: weird shit happens [04:15] thumper: otp to jtv [04:15] wallyworld_: I had a container with that config, didn't really start properly, and worse than that, when I did lxc-shutdown, it shut down the host [04:16] as in, my laptop [04:17] haha [04:22] bigjools: I lost those links you pasted before [04:22] bigjools: got them handy? [04:23] thumper: http://bj0z.wordpress.com/2011/08/19/howto-build-a-base-lxc-container-in-ubuntu-11-04/ [04:23] thumper: http://unix.stackexchange.com/questions/50201/how-to-configure-external-ip-addresses-for-lxc-guests [04:36] thumper: I let wallyworld_ go, so you can have him now. [04:36] thumper: funny that your laptop shut down [04:36] well, funny for me :-D [05:18] thumper: well that sucks [05:19] wallyworld_: I'm sending an email out [05:19] end of day lack of progress email [05:19] thumper: ni, i have a new peroblem [05:19] which is? trouble spelling? [05:19] it seems you can't bridge to a wireless interface [05:19] typing [05:20] and when i tried doing it to my eth nic which is disconnected, it killed my connectivity [05:20] heh [05:20] and i don't have a wired connection that is easily accessible :-( [05:21] i mean, it's 2013 after all [05:23] :) [05:23] * thumper signs off until the meeting === tasdomas_afk is now known as tasdomas === ChanServ changed the topic of #juju-dev to: https://juju.ubuntu.com | On-call reviewer: dimitern | Bugs: 5 Critical, 76 High - https://bugs.launchpad.net/juju-core/ [07:34] dimitern: you've seen, I handled your review yesterday evening [07:34] dimitern: morning btw [07:38] TheMue: morning :) [07:38] TheMue: yeah, except one suggestion about table based tests [07:39] mornin' all [07:39] rogpeppe: moin [07:39] TheMue: https://codereview.appspot.com/11588043/diff/6001/environs/sync/sync_test.go#newcode85 [07:42] dimitern: yep, they haven't been table driven in the old original synctools code. this cl is intended to only move the logic with as less changes as possible. [07:43] rogpeppe: morning, reviewed your ...4045 yesterday [07:43] morning [07:43] TheMue: yup, saw that, thanks a lot! branch is landing now. [07:43] axw: hiya [07:44] TheMue: so don't you think converting them to a table-based test will improve the code, and reduce it a lot? [07:45] axw: hey [07:46] dimitern: maybe in another cl, but i dislike mixing two intentions in one cl [07:47] dimitern: but i'm also no real fan of table-driven tests. here you often need extra effort to follow failing tests, e.g. add the output of an index or description to see which of the tests is failing [07:48] TheMue: well, instead of like 6 tests which are almost identical, you can have a 6-element table and a single case testing all cases [07:48] TheMue: but anyway, please at least add a TODO about that [07:50] dimitern: shouldn't it be a CANDO? ;) [07:51] TheMue: well, you asked me for a review :P [07:52] TheMue: i think it should be done, not necessarily by you [07:52] TheMue: but having a TODO at the right place is helpful to remember [07:54] dimitern: will change it to TDT so you've got your peace. but I still have the opinion that they are worse readable and followable in case of a fail. the only positive part is writing less code. [07:55] TheMue: let's agree to disagree then :) [07:55] dimitern: and btw, SetUpTest and TearDownTest have to be moved inside the loop this way. instead using the mechanisms of all mature test frameworks [07:56] TheMue: when I have some time, I'll do it and show you how better it looks ;) [07:57] TheMue: FWIW i agree strongly with dimitern in this case [07:57] TheMue: with the non-table test I have to scan very carefully for subtle differences between the test cases [07:57] rogpeppe: I knew you would [07:57] rogpeppe: that block there is almost 1-1 ready for a table [07:58] TheMue: whereas if it was a table i could look at one single piece of code, and then look at the changing data independently [07:58] evening [07:58] davecheney: morning :) [07:59] TheMue: for the setup/teardown thing, you can call SetUpTest and TearDownTest within the test itself, which can work (i usually define Reset as {Teardown; Setup}) [07:59] davecheney: hey boss [07:59] rogpeppe: but i have to identify the test data that caused the fail. often enough found tests where i had to count tests myself instead just reading the source [07:59] TheMue: that's why we usually put an "about" field in the table [07:59] TheMue: and log it [08:00] +1 [08:00] rogpeppe: that sounds terrible [08:00] rogpeppe: why not just have an independent test? [08:00] thumper: +1 [08:00] :D [08:00] gents, meeting time [08:01] thumper: because in this case there are 110 almost-duplicated and non-trivial lines of test code [08:01] thumper: which could probably be reduced to a single function with a static data table [08:01] thumper: personally i think that's a win [08:04] axw: ping [08:04] axw: https://plus.google.com/hangouts/_/bf3f4cfe715017bf60521d59b0628e5873f2a1d3 [08:04] axw: https://plus.google.com/hangouts/_/bf3f4cfe715017bf60521d59b0628e5873f2a1d3 [08:05] ? [08:05] brt [08:05] axw: team meeting [08:05] ah [08:05] axw: yeah, but no one invited you :) [08:07] https://docs.google.com/a/canonical.com/document/d/1eeHzbtyt_4dlKQMof-vRfplMWMrClBx32k6BFI-77MI/edit# [08:13] mramm: hey there [08:14] we're having a meetiong [08:14] wanna come ? [08:14] hey [08:14] trying for the last bit [08:57] thumper: got 5? [08:57] wallyworld: for you, maybe even six [08:57] wallyworld: hangout or irc? [08:57] rejoin thehangout [08:58] i'm looking for a review of this, please: https://codereview.appspot.com/11800044/ [08:58] rogpeppe: on it [08:58] dimitern: thanks [09:14] rogpeppe: reviewed [09:15] dimitern: thanks [09:31] looking for a second review of https://codereview.appspot.com/11800044/ if anyone cares to take a look [09:42] rogpeppe, https://codereview.appspot.com/11800044/ reviewed, just one question [09:44] fwereade: no particular reason other than it doesn't seem like it's necessary for state to define every possible interface that a client might use [09:44] fwereade: but i'll move it the interface to state if you'd prefer [09:45] rogpeppe, I agree that it's not quite right as it stands, but I'd be happiest if we at least kept those bits all together [09:46] rogpeppe, LGTM with that then [09:46] fwereade: thanks [09:46] Damn, I was about to approve that one. :) [09:46] fwereade: you wanted a chat about instance-type selection in Azure? [10:07] night folks [10:11] jtv, yeah, sorry -- I was just wondering whether anything sprang to mind re possible unifications of the various don't-deploy-worthless-instances implementations [10:11] jtv, IIRC there's something in openstack and ec2, as well as azure now [10:12] jtv, (this is not a request for you to do more work, just a quick check to see if there's stuff in your brain that should maybe also be in mine) [10:12] What do you mean by worthless? [10:12] You mean, for setting a minimal baseline spec? [10:12] jtv, unlikely to be useful [10:12] jtv, yeah [10:12] Hmm... [10:12] That does look as if it ought to be the same story for every cloud, doesn't it? [10:13] jtv, it does rather [10:13] It seems like a simple problem technically... Just one function that everyone could call. [10:13] jtv, and I have a feeling that at least one of them tries with the baseline and then falls back to whatever-it-can-get rather than just refusing [10:14] jtv, yeah -- I don't intend to schedule anything now, just to remember it as something to bear in mind [10:14] jtv, would you add a tech-debt bug please? and link it in the CL? [10:21] OK [10:23] dimitern: the CL just landed again with table driven tests. ;) you can take a look? [10:23] TheMue: which one was it? [10:24] dimitern: https://codereview.appspot.com/11588043 [10:24] davecheney: hey man, you around? [10:27] TheMue: that's great, but my comment was not about cmd/juju/synctools_test.go, but about environs/sync/sync_test.go :) [10:27] fwereade: I filed bug 1204851 — now how do I link it to the CL? [10:27] <_mup_> Bug #1204851: Different providers set different baseline instance constraints

[10:27] jtv, sorry, I just mean to mention it in there somewhere [10:27] jtv: from LP's MP directly [10:27] dimitern: so next round ;) [10:28] dimitern: but why didn't mentioned you that for synctools_test? :P [10:28] TheMue: because they were less and not so obviously similar [10:28] dimitern, I don't think that's quite right, this CL doesn't actually address it -- it's just the thing that caused us to think "yeah, this should exist" [10:29] fwereade: sorry? [10:30] dimitern: they are even more similar than in sync_test [10:30] dimitern, I don't think it's very useful to have the bug linked to the CL, because the CL doesn't address the bug; but IME it *is* useful to be able to look back at a CL and see that it spawned a bug [10:30] TheMue: so good that you did them as well, no? [10:30] fwereade: ah, I don't really know, I was just answering jtv's question, I think [10:31] dimitern: please rephrase [10:31] dimitern, yeah, you were just missing a spot of context [10:31] fwereade: most likely [10:31] TheMue: i meant "it was good that you've seen the similarity there and found a way to refactor them into a table-based test" [10:32] dimitern: ah, ack [10:33] Argh. Can we get another gwacl update on the build machine? [10:43] jtv: on it [10:43] Thanks. I'll send the email. [10:44] jtv: thanks [10:44] jtv: oops :) [10:44] jtv: i meant "done" [10:44] Then thank you too. :)) [10:44] jtv: np [11:30] mgz, rogpeppe, TheMue, fwereade: standup [11:31] ta [11:34] rogpeppe: standup [11:41] hm, Addresses in state/open.go currently uses st.db.Session.LiveServers()... what's that do exactly? I'd want machine ids to get detailed address info [12:08] gary_poster: not that I'm aware of [12:09] pffe [12:14] fwereade: I cannot use StateInfo in apiserver [12:39] rogpeppe: any idea for a workaround getting the addresses? [12:39] dimitern: why can't you use StateInfo in apiserver? [12:40] rogpeppe: it needs a bootstrapped environment [12:40] dimitern: isn't the environment bootstrapped by the time the api server starts? [12:40] rogpeppe: provider-config file in the bucket [12:40] rogpeppe: and moreover, it's a client api [12:41] rogpeppe: it's not supposed to be run on the server [12:41] rogpeppe: how to test it? [12:41] dimitern: it'll still work though, as a temporary hack, won't it? [12:42] dimitern: how are any of the api server calls which use an Environ tested? [12:42] rogpeppe: http://paste.ubuntu.com/5911031/ [12:42] rogpeppe: none of apiserver calls use the environ, except for the upgrader [12:42] dimitern: i thought deploy did [12:43] rogpeppe: no [12:44] dimitern: Client.ServiceDeploy definitely does [12:44] rogpeppe: let me see [12:46] dimitern: hmm, i *think* it should work [12:46] rogpeppe: no, it doesn't [12:46] rogpeppe: not StateInfo or anything provider specific at least [12:46] rogpeppe: and that's what I need [12:47] dimitern: i don't see why [12:47] dimitern: can't you call NewConnFromState? [12:47] rogpeppe: why do I need that? [12:48] dimitern: because a Conn contains an Environ [12:48] rogpeppe: NewConnFromState essentially calls environs.New(cfg) [12:48] dimitern: yes [12:48] rogpeppe: I'm doing this: [12:49] dimitern: the api server doesn't have environ credentials when it first bootstraps, so the StateAddresses call will fail until the credentials are pushed, but i don't *think* that's a problem because we won't be using StateAddresses until the first connection has been made. this is my only concern. [12:49] rogpeppe: http://paste.ubuntu.com/5911043/ [12:49] rogpeppe: and StateInfo fails with "provider-config" not found [12:49] rogpeppe: which happens at bootstrap time [12:49] dimitern: oh god, StateInfo is non-provider-specific now [12:50] dimitern: that's broken [12:50] rogpeppe: why? [12:50] dimitern: because the local provider doesn't store anything in the private bucket [12:50] rogpeppe: I don't see how this matters [12:51] dimitern: you're using StateInfo wrong [12:51] dimitern: you need to call env.StateInfo() [12:51] rogpeppe: I'll try [12:51] dimitern: environs.StateInfo is just a way for providers to implement Environ.StateInfo [12:54] rogpeppe: now only APIAddresses test fails [12:54] rogpeppe: and each time the port is different, as reported by StateInfo [12:55] dimitern: that's expected, i think [12:55] rogpeppe: actually I think State.APIAddresses is too lazy [12:55] rogpeppe: it just replaces the port [12:55] dimitern: oh yes, that's true [12:55] rogpeppe: and returns the same host as the state address [12:55] rogpeppe: I can hack it a bit to get the actual port [12:55] dimitern: where the dummy provider starts the API server on a random port [12:58] dimitern: i think we should actually store the API addresses in the state [12:59] rogpeppe: yeah, perhaps, but until then I need a workaround [12:59] dimitern: i'm not sure there is one [12:59] dimitern: how can the state know what port number has been allocated by the dummy provider? [13:00] dimitern: isn't it quite trivial to add a State.SetAPIAddresses method? [13:01] rogpeppe: I did it [13:01] dimitern: cool! [13:01] rogpeppe: using JujuConn.APIInfo() in the test [13:01] rogpeppe: instead of s.State.APIAddresses [13:01] dimitern: i thought the test was testing APIAddresses... [13:01] rogpeppe: it was, but not anymore [13:02] rogpeppe: it's testing what's in the api info, the same as reported back [13:02] rogpeppe: will live test it now and see [13:02] dimitern: i'll reserve judgement until i see the CL :-) [13:21] dimitern: so, more table driven testing === teknico1 is now known as teknico [13:23] rogpeppe: live tests work! [13:23] TheMue: will take a look soon [13:24] dimitern: thx [13:25] dimitern: cool. am at lunch BTW [13:33] TheMue: reviewed [13:37] rogpeppe, fwereade: https://codereview.appspot.com/11800045/ [13:41] dimitern: thx === hatch_ is now known as hatch [14:20] fwereade: ping [14:21] dimitern, pong [14:21] fwereade: can you take a look please? ^^ [14:22] dimitern, ah, sorry [14:22] rogpeppe: you as well, after lunch? [14:23] dimitern: looking [14:23] thanks guys === tasdomas is now known as tasdomas_afk [14:30] ok, mu mouse officially acts weird lately - a single click is frequently a double click and it drops while dragging and still holding the left button... hmm need a screw driver [14:32] I find drag and drop harder with a screwdriver than a mouse [14:32] lol [14:41] * TheMue dcc's dimitern his BT mouse, isn't needed anymore [14:47] * dimitern bbi1h [14:51] (or so.. dentist) [14:54] dimitern, you'll have a review waiting when yuo return [14:57] does anyone recall the justification for CanDeploy? [15:07] dimitern: you have a review [15:08] fwereade: i remember the justification for CanDeploy [15:09] rogpeppe, oh yes? [15:09] fwereade: it's because if we implemented AssignedMachineId(unitTag) in API (which is what the deployer previously used) it's a bit of a security leak [15:09] rogpeppe, but CanDeploy itself doesn't need to exist AFAICS [15:09] fwereade: because it means any machine agent can ask for the assigned id of any unit [15:10] fwereade: well yes, i was puzzling over why we checked AssignedMachineId in the first place [15:10] rogpeppe, what does CanDeploy do that isn't already done with UnauthorizedErrors? [15:11] * rogpeppe checks [15:11] rogpeppe, it's just a big copy of the deployer's authfunc as far as I can see [15:11] rogpeppe, it's just really well obfuscated ;p [15:11] fwereade: the deployer's authfunc? [15:12] rogpeppe, deployer.go:53 [15:13] * rogpeppe looks in api*server*/deployer/deployer.go:53 :-) [15:13] rogpeppe, ha, sorry [15:13] fwereade: no, it's different to that [15:13] rogpeppe, ah, ok, which units will be handled differently? [15:14] fwereade: i *think* the case it's dealing with is when we get told about a unit by the watcher but that unit is no longer assigned to this machine [15:15] rogpeppe, we'll get an unauthorized out of Life in that case [15:15] fwereade: oh sorry, i didn't read far enough [15:17] fwereade: wow, that getAuthFunc does quite a lot of work [15:18] rogpeppe, the theory is that watchers are cheap [15:19] rogpeppe, if not, getAllUnits can be tuned on its own [15:22] fwereade: i suspect that making a set from the results of getAllUnits is vast premature optimisation, but i take your point [15:22] fwereade: yes, i don't see any point in CanDeploy [15:23] rogpeppe, I think it's more about clarity of expression than runtime optimization [15:23] rogpeppe, cool, thanks [15:26] fwereade: personally i think it's just as clear (+ shorter + more correct) this way, but YMMV. http://paste.ubuntu.com/5911521/ [15:29] rogpeppe, I think it's maybe implemented a little awkwardly -- would be nicer to build the set from units directly -- but I can't generally ever get very enthusiastic about replacing N map lookups with N linear searches [15:29] rogpeppe, not to say it's never justified [15:30] fwereade: the thing is currently N=1 [15:30] fwereade: so we lose out by building the set [15:31] rogpeppe, I'd rather pay the fixed costs now and have something that stays sane as N grows [15:31]