niemeyer | wallyworld: No, unfortunately I can't do a release today. This needs to be prepared and announced as usual, and it's not a good idea to do that on a Friday before leaving for a sprint | 00:06 |
---|---|---|
wallyworld | niemeyer: ok. we have have to release beta12 today/tomorrow to we'll need to deal with it | 00:08 |
wallyworld | s/to/so | 00:08 |
niemeyer | wallyworld: Sorry about that, but I really can't risk compromising everybody else with a broken release when I'm not going to be around for days to fix it | 00:10 |
wallyworld | niemeyer: totally understand. i think the CI guys are looking at a workaround | 00:10 |
marcoceppi | is there a limit on the size of a local resource I can deploy with beta11? | 00:16 |
marcoceppi | nevermin,d just took a hot min | 00:18 |
niemeyer | wallyworld: Note the code can actually import from mgo.v2-unstable | 00:29 |
niemeyer | wallyworld: That may be the easiest | 00:29 |
wallyworld | niemeyer: sure, but that means churn just to change all the import paths | 00:29 |
niemeyer | wallyworld: ... which is a trivial mechanical change, right? | 00:30 |
wallyworld | it is. and then we change back again next week or whenever. that bit about Go is annoying to say the least | 00:31 |
axw | wallyworld: http://reviews.vapour.ws/r/5244/ - trivial fix for windows failure | 01:02 |
wallyworld | ok | 01:02 |
wallyworld | axw: lgtm | 01:04 |
thumper | niemeyer: except it is a "trivial mechanical change" across 10 different repos | 01:19 |
alexisb__ | so thumper, wallyworld are we all good for release except for dep updates for the mgo fix? | 01:43 |
alexisb__ | can one of you please send a summary to balloons and myself | 01:44 |
thumper | alexisb__: I'm not sure if we even need to do the dep update | 01:44 |
thumper | if the release process can pull over the branch | 01:44 |
alexisb__ | thumper, understood | 01:44 |
wallyworld | alexisb__: as per my email i am expecting what tim said | 01:44 |
alexisb__ | I want to make sure that is clear to all | 01:44 |
alexisb__ | what is being requested | 01:44 |
wallyworld | i asked nic/martin to confirm in the email | 01:45 |
alexisb__ | wallyworld, thumper I sent a clarification email just in case :) | 01:53 |
alexisb__ | please take a look and add comentary/corrections if neccisary | 01:54 |
wallyworld | alexisb__: thanks. let's hope we get the answer we are looking for. we need a jedi mind trick | 01:55 |
niemeyer | thumper: A trivial mechanical change across "10 repositories" (doubt it) is still a trivial mechanical change. It'd be done by now. | 02:31 |
thumper | you doubt that we have that many that depend on mgo? | 02:32 |
thumper | really? | 02:32 |
niemeyer | Yeah, I doubt that a juju release has to build 10 independent repositories that all depend on mgo.. will be surprised, honestly. | 02:33 |
thumper | not build 10, but there are ~10 external repos that juju/juju depends on that also bring in mgo | 02:34 |
thumper | all those would also need to change, no? | 02:34 |
thumper | luckily we have a smart releast team that have worked a way for us not to bother | 02:34 |
thumper | but yes, changing imports is invasive | 02:35 |
niemeyer | I don't know.. depends on how you organize your code I guess. I'm just saying I'm surprised if that's the case, as I try not to do that. | 02:35 |
thumper | sure... whatever | 02:35 |
niemeyer | It still a mechanical change though, which I bet takes less typing than we've done thus far. | 02:35 |
niemeyer | I bet your release team very far surpasses the necessary skills to do that. | 02:36 |
thumper | no... we are working around it so we don't have to | 02:36 |
thumper | yes, their skill far surpasses what is needed to rename shit | 02:36 |
niemeyer | That's good to hear .. nobody wants stuff named shit. | 02:37 |
natefinch | wallyworld, thumper: btw, you can cheat with godeps and still be importing gopkg.in/mgo.v2 but pin a revision from v2-unstable. That way, only the dependencies.tsv file needs to change. It's a hack, but it works and doesn't cause a ton of churn. | 02:51 |
wallyworld | natefinch: interesting, i didn't realise we could do that | 02:53 |
* thumper wonders about the nested packages | 02:53 | |
natefinch | yeah, godeps is a handy little hack... it doesn't care about branches at all, so as long as the revision is in the git repo, godeps can point the local code at it. | 02:56 |
thumper | natefinch: https://github.com/go-mgo/mgo/blob/v2-unstable/session.go | 02:56 |
thumper | natefinch: still impots from v2-unstable | 02:57 |
thumper | so not sufficient | 02:57 |
natefinch | thumper: oh, yeah, that won't work | 02:57 |
natefinch | crud | 02:57 |
natefinch | I've done it with feature branches, but they weren't hard coded in the import statement. That does make it basically impossible | 02:58 |
mup | Bug #1570594 changed: read access to admin model allows grant <docteam> <juju-release-support> <juju-core:Fix Released by thumper> <https://launchpad.net/bugs/1570594> | 03:14 |
thumper | natefinch: however we can specify the revision that was the tip of babbleclunk's branch | 03:30 |
natefinch | thumper: sounds like it could work | 03:37 |
thumper | jam: ping | 03:48 |
thumper | jam: unping | 03:52 |
=== thumper is now known as thumper-afk | ||
wallyworld | axw: i've updated the mgo dep, will fix mgo issue in beta12 http://reviews.vapour.ws/r/5245/ | 05:15 |
wallyworld | jam: so as axw pointed out, the above PR won't work because code in gopkg.in/mgo.v2 imports from gopkg.in/mgo.v2-unstable. wtf | 05:20 |
axw | wallyworld: gopkg.in/mgo.v2-unstable/txn imports gopkg.in/mgo.v2-unstable | 05:21 |
axw | the commit you're pointing at is from the unstable branch | 05:21 |
wallyworld | yes, but i thought i was following the advise that we just hack the commit sha in deps.tsv | 05:21 |
wallyworld | which won't work will it | 05:22 |
axw | nope | 05:23 |
wallyworld | so no choice other than to vendor if we want beta12 out | 05:23 |
axw | wallyworld: we build with go 1.6 now, so that would be the least amount of work I think | 05:24 |
axw | wallyworld: take it or leave it, but this works: https://github.com/juju/juju/pull/5805 | 06:25 |
wallyworld | looking | 06:39 |
wallyworld | axw: hmmm, seems like now we will just be double handling everything. commit to the upstream repo and copy all the files to the vendored package. i'm -1 on it, so best you get jam or someone to +1 | 06:46 |
wallyworld | unless there's something i don't get | 06:46 |
axw | wallyworld: how's that worse than committing to upstream, copying to juju namespace? | 06:47 |
wallyworld | now we commit to upstream and update dependencies.tsv | 06:48 |
wallyworld | one small change to juju/juju | 06:48 |
axw | wallyworld: in other words, what are you going to do if you don't want to merge this | 06:48 |
axw | wallyworld: this doesn't have to live beyond beta12 | 06:48 |
wallyworld | seems like a huge change just for one beta, but i guess it's just copying files | 06:49 |
wallyworld | the other alternative is just to update the import paths to v2-unstable | 06:50 |
axw | wallyworld: copying a bunch of files into one temporary location, vs. changing files over all our repos | 06:50 |
wallyworld | fair point. it can be reversed after beta ships | 06:51 |
wallyworld | although we'd just need to change import paths in juju/juju | 06:51 |
axw | wallyworld: nope. blobstore takes mgo.v2, we pass into it from juju | 06:51 |
axw | likewise for juju/txn I think | 06:52 |
wallyworld | sure, but blobstore has no depencencies.tsv | 06:52 |
wallyworld | ah | 06:52 |
wallyworld | but yeah import paths | 06:52 |
wallyworld | either way it's a mess | 06:52 |
wallyworld | axw: so maybe it's the only viable option, but cut you get a +1 from william or john just to be sure? | 06:53 |
axw | wallyworld: sure | 06:54 |
axw | I think will is out | 06:54 |
axw | jam: are you about? | 06:54 |
axw | away on IRC at any rate | 06:54 |
wallyworld | axw: here's a quickee fix for a critical restore bug http://reviews.vapour.ws/r/5247/ | 07:03 |
wallyworld | the main issue is now gone since rog deleted failback connect | 07:03 |
axw | wallyworld: why change it to use ControllerByName first? | 07:06 |
wallyworld | axw: i wanted to reuse existing cloud / region | 07:07 |
axw | wallyworld: if you rebootstrap with a mismatched cloud/region, it'll be terribly broken anyway | 07:07 |
wallyworld | ah yeah, that is true | 07:07 |
wallyworld | i'll revert that bit | 07:08 |
jam | axw: wallyworld: I thought we could point to babbageclunk's revision as well | 07:21 |
jam | which was written against mgo.v2 so shouldn't be importing mgo.v2-unstable | 07:21 |
jam | thumper mentioned a revision | 07:21 |
axw | jam: dependencies.tsv does not separate repo from package path | 07:22 |
axw | so you can't say get babbageclunk's repo, and call it gopkg.in/mgo.v2 | 07:22 |
axw | wallyworld: in your PR, I meant a test for the thing that broke in the first place | 07:23 |
axw | wallyworld: I didn't see a test added | 07:23 |
wallyworld | axw: the thing that broke was in api connect code that was deleted | 07:24 |
axw | wallyworld: that was the trigger, but you still fixed a bug? | 07:24 |
wallyworld | yeah, i could add a check for the order of updating controller info | 07:25 |
wallyworld | i'll do that | 07:26 |
jam | axw: because the revision is in the repository that also holds gopkg.in/mgo.v2 godeps will let you use it | 07:29 |
jam | axw: anything revision that is in the upstream repository is a potential target for godeps | 07:29 |
jam | I tried it with aee6a64 to confirm | 07:30 |
jam | though that would suffer from the import problem | 07:30 |
axw | jam: babbageclunk's most recent commit has the -unstable import paths | 07:32 |
axw | it wasn't done as a separate merge step | 07:32 |
wallyworld | axw: test done, will propose once merge finishes | 07:33 |
axw | wallyworld: thanks | 07:34 |
wallyworld | i should have done it in the first place | 07:34 |
jam | axw: well that sucks | 07:38 |
wallyworld | axw: and here's the test http://reviews.vapour.ws/r/5248/ | 07:55 |
wallyworld | jam: so we either (temporarily) vendor as axw has proposed or we update import paths in juju/blobstore, juju/txn and juju/juju | 07:57 |
axw | wallyworld: LGTM, thanks | 07:57 |
wallyworld | ta | 07:57 |
jam | wallyworld: if we vendor we have to update everything don't we? the discussion with mgz yesterday was that we might be able to add an "apply this patch" as part of the 'make-release-tarball.sh' script | 08:08 |
wallyworld | jam: yeah, that's my issue with vendoring. axw has done the PR - it's huge. but the current thinking is this will be just for this beta. the changes to core are trivial. 99.99% of the PR is copying in the files of all the deps | 08:13 |
wallyworld | the alternative is to churn the necessary dependent repos and then update core also | 08:13 |
wallyworld | with the new import paths | 08:14 |
axw | wallyworld jam: to be up front, I would like it to be there for good, and I'll make my argument for that later. but we don't *have* to have it for good | 08:14 |
wallyworld | axw: if it is there for good, my issue is the double handing of every dependent repo update | 08:14 |
axw | wallyworld: please let's talk about that later :) | 08:14 |
wallyworld | :-) | 08:15 |
axw | (wallyworld: I don't disagree that that's a problem, but there are benefits that I'd like us to weigh up.) | 08:16 |
wallyworld | sure, fair point | 08:16 |
axw | I intended to bring it up at the sprint, but there was never really a good time | 08:16 |
wallyworld | yeah :-( | 08:16 |
anastasiamac | axw: sounds like we need another sprint :) | 08:25 |
axw | heh | 08:26 |
axw | I'm all sprunt out | 08:26 |
anastasiamac | the same. m glad my voice is back at least \o/ | 08:26 |
babbageclunk | hey wallyworld - what happened with this PR? https://github.com/juju/juju/pull/5804 | 08:29 |
wallyworld | babbageclunk: i had to not do it cause it would have broken everything due to mixed import paths | 08:30 |
wallyworld | code in that rev imports from v2-unstable | 08:30 |
babbageclunk | wallyworld: ah, gotcha. | 08:30 |
urulama|____ | does openstack provider in juju 1.25 support v3 keystone auth? | 08:31 |
=== urulama|____ is now known as urulama | ||
wallyworld | no | 08:31 |
urulama | but it does in juju 2.0, right? | 08:31 |
urulama | (to a very limited default region degree) | 08:32 |
wallyworld | babbageclunk: jam and axw and myself have been discussing. there's a PR to vendor all the things http://reviews.vapour.ws/r/5246/; or we have to update all the necessary repos with new import paths | 08:32 |
wallyworld | babbageclunk: either way, pain and suffering | 08:32 |
babbageclunk | wallyworld: :( | 08:33 |
wallyworld | urulama: you can specify domain and endpoint url | 08:33 |
wallyworld | urulama: bbiab, dinner time | 08:33 |
urulama | wallyworld: bon appétit | 08:33 |
babbageclunk | How many reviewers does a 19k-line diff need? | 08:42 |
urulama | :) | 08:43 |
babbageclunk | Prob'ly a few. | 08:44 |
axw | wallyworld: I'll hold off until the beta is out, but I have a PR ready to delete some code: http://reviews.vapour.ws/r/5249/ | 08:45 |
axw | no rush on a review, can wait till monday | 08:46 |
* axw goes to make dinner | 08:46 | |
babbageclunk | axw, wallyworld, jam - is there anything I can do to help with the mgo update before the release? With vendoring the package, are we then going to apply the minimal patch in place? (Sorry for interrupting dinners.) | 09:22 |
axw | babbageclunk: what's in my PR is actually the v2-unstable branch, just renamed to v2 with import paths fixed | 09:37 |
axw | babbageclunk: so it has your changes | 09:37 |
* axw goes away again | 09:39 | |
babbageclunk | axw: Awesome, thanks. | 10:07 |
urulama | axw: ERROR storing charm for URL "cs:ubuntu-0": delegatable macaroon cannot be obtained for public entities | 10:22 |
urulama | axw, wallyworld: ^ is this client side? | 10:22 |
axw | urulama: I don't know which part you mean, but that error message is printed out by the client: https://github.com/juju/juju/blob/master/cmd/juju/application/deploy.go#L460 | 10:24 |
axw | urulama: sorry, gtg | 10:24 |
urulama | axw: all good, default grant is read only, that's returned instead of saying you don't have write access | 10:25 |
babbageclunk | does anyone know if there's a reason not to replace common.SortStringsNaturally with the version from utils throughout juju/juju? | 10:37 |
babbageclunk | The utils version handles values with multiple embedded numbers. | 10:38 |
babbageclunk | (like 3/lxd/4) | 10:38 |
axw | jam: I've just weeded out a few issues with my vendor branch, should be good now. if you're OK with it, can you please merge it? otherwise someone will need to do the dep changes | 10:43 |
axw | I need to go help with the kids now | 10:44 |
babbageclunk | (answering myself) gah, this change was already done (https://github.com/juju/juju/pull/5666), but on the model-migration branch. | 10:50 |
babbageclunk | That seems obtuse. | 10:51 |
voidspace | babbageclunk: I bought the Microsoft LX6000 headset. Works great and doesn't look so ridiculous. :-) | 11:45 |
anastasiamac | babbageclunk: as far as I know, we intend to merge model-migration branch next week (right after shipping beta12) | 11:47 |
anastasiamac | babbageclunk: mayb hangon on updating the bug and mark as fix committed once the merge occurs?.. :D | 11:47 |
babbageclunk | voidspace: yay, headset buddies! | 12:31 |
babbageclunk | anastasiamac: ok, thanks - I'll do that. | 12:31 |
anastasiamac | babbageclunk: \o/ thank you for ur patience :D there are plenty ofother bitesizes that can be nibbled on if u r keen :P | 12:32 |
babbageclunk | anastasiamac: :) Actually yeah - I find myself at a bit of a loose end this afternoon, so I'll have a look. | 12:33 |
anastasiamac | babbageclunk: wow! i forgot what it feels like :) u r great for tackling these! | 12:33 |
babbageclunk | :) | 12:36 |
katco | natefinch: standup time | 15:02 |
mup | Bug #1603473 opened: Relation fails as untis/machines are on different subnet on a multi NIC setup, juju 2.0 beta11 <11> <2.0> <beta> <juju> <juju-core:New> <https://launchpad.net/bugs/1603473> | 15:33 |
alexisb__ | voidspace, ping | 16:18 |
=== redelmann is now known as rudi|comida | ||
alexisb__ | dooferlad, ping | 16:22 |
dooferlad | alexisb__: I am both on holiday and cooking, but what's up? | 16:26 |
alexisb__ | dooferlad, nevermind, it is not that urgent | 16:26 |
alexisb__ | sorry didnt know you were out today | 16:26 |
alexisb__ | enjoy your holiday | 16:26 |
perrito666 | dooferlad: also dont cook and irc | 16:27 |
dooferlad | perrito666: it's OK, that's what irccloud is for. I am done with knives and am just waiting for things to cook. | 16:28 |
=== natefinch is now known as natefinch-afk | ||
natefinch-afk | out for a bit taking a kid to a checkup | 17:17 |
perrito666 | natefinch-afk: gluck | 17:25 |
frobware | alexisb__: will take a look over bug #1603473 again | 17:31 |
mup | Bug #1603473: Relation fails as untis/machines are on different subnet on a multi NIC setup, juju 2.0 beta11 <11> <2.0> <beta> <juju> <juju-core:New> <https://launchpad.net/bugs/1603473> | 17:31 |
alexisb__ | thanks frobware | 17:32 |
alexisb__ | frobware, if we cna just get an idea of next steps that would be awesome | 17:32 |
frobware | alexisb__: just asked for their MAAS network setup - with that I can try and reproduce | 17:33 |
=== rudi|comida is now known as redelmann | ||
katco | need to run out for a late lunch meeting. bbiab | 18:41 |
=== natefinch-afk is now known as natefinch | ||
mup | Bug # changed: 1567635, 1568101, 1583789, 1584616, 1586512, 1590605, 1591387, 1594924, 1595686, 1596493, 1596597, 1597342, 1597704, 1598049, 1598063, 1598113, 1598118, 1598164, 1598286, 1598289, 1598293, 1598897, 1598964, 1599402, 1599779, 1599905, 1602732, 1602895, 1603208 | 19:43 |
mup | Bug #1593828 changed: cannot assign unit E11000 duplicate key error collection: juju.txns.stash <ci> <conjure> <deploy> <intermittent-failure> <oil> <oil-2.0> <juju-core:Fix Released by 2-xtian> <https://launchpad.net/bugs/1593828> | 20:13 |
natefinch | if anyone has some time, moderate sized review (+631, -20) http://reviews.vapour.ws/r/5238/ | 20:50 |
=== natefinch is now known as natefinch-afk | ||
mup | Bug #1603577 opened: backup-restore: panic: empty value for "api-port" found in configuration <backup-restore> <blocker> <ci> <regression> <juju-core:Triaged> <https://launchpad.net/bugs/1603577> | 22:41 |
mup | Bug #1603577 changed: backup-restore: panic: empty value for "api-port" found in configuration <backup-restore> <blocker> <ci> <regression> <juju-core:Triaged by wallyworld> <https://launchpad.net/bugs/1603577> | 22:47 |
katco | wallyworld: mup is telling on you :) | 22:52 |
wallyworld | alexisb: that bug is a slightly different manefestation of the same know restore issue, so the beta should still go ahread | 22:54 |
wallyworld | alexisb: what was rhe final decison with regards to mgo - vendoring was considered and ruled out, looks like a patch approach was adopted | 22:57 |
mup | Bug #1603577 opened: backup-restore: panic: empty value for "api-port" found in configuration <backup-restore> <blocker> <ci> <regression> <juju-core:Triaged> <https://launchpad.net/bugs/1603577> | 22:59 |
mup | Bug #1603584 opened: juju-uitest calls obsolete --show-passwords <ci> <juju-gui> <regression> <juju-core:Triaged> <https://launchpad.net/bugs/1603584> | 23:11 |
mup | Bug #1603585 opened: modelManagerStateSuite.SetUpTest no reachable servers on windows <ci> <intermittent-failure> <test-failure> <windows> <juju-core:Triaged> <https://launchpad.net/bugs/1603585> | 23:41 |
alexisb | wallyworld, it is not blocking beta | 23:45 |
alexisb | but I am assigning it to you as it was a new symtom of the same restore issue you are working on | 23:45 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!