[04:13] <wallyworld> hpidcock: this looks bigger than it is - small code changes but significant test boilerplate, deletion of unused code etc https://github.com/juju/juju/pull/10982
[04:14] <hpidcock> on it
[04:18] <wallyworld> ty, sorry in advance
[04:18] <wallyworld> it was a bitch to write
[05:37] <hpidcock> wallyworld: LGTM, tried it out on gce with no issues
[05:43] <wallyworld> hpidcock: awesome, ty
[05:43] <wallyworld> still go 3 pacgaes to go, won't be as hard
[05:44] <wallyworld> i tested with k8s as well
[05:44] <wallyworld> as they use resources etc
[09:04] <timClicks> could someone please take a look at a discourse answer I've just given and check that it makes sense? I should really log off https://discourse.jujucharms.com/t/how-to-deploy-charms-to-specific-clusters-on-vsphere-cloud/2396/8
[09:57] <manadart> stickupkid or nammn_de1: https://github.com/juju/juju/pull/10983
[10:34] <stickupkid> love it when this happens https://paste.ubuntu.com/p/ccsyVWTC5r/
[10:38] <achilleasa> stickupkid: you can just claim that one of the two contains a zero-width unicode character :D
[12:09] <manadart> achilleasa: Did you have any success with that test in your patch?
[12:17] <achilleasa> manadart: still chasing it down :-(
[12:18] <achilleasa> I am down to sprinkling fmt.Printf statements everywhere...
[12:18]  * achilleasa needs stronger coffee...
[14:13] <gnuoy> kwmonroe, hi there, a work aroundthat was put into layer basic sometime ago (https://github.com/juju-solutions/layer-basic/pull/51/files) seems to be no longer valid (https://github.com/juju-solutions/layer-basic/issues/149). I'd like to put up a PR to remove the workaround but am concerned I might set the world on fire. Do you have any thoughts on the issue ?
[14:14] <gnuoy> fwiw just removing the allow_hosts line works for me but am I guessing that was needed in the original work around
[14:15] <gnuoy> s/but am I/but I'm/
[16:16] <achilleasa> manadart: hmmm... I still need to add a "NewScopedProviderAddressInSpace" c-tor though
[16:16] <manadart> achilleasa: Yeah, that's OK for now.
[16:45] <nammn_de1> rick_h and anyone else regarding the "not returning an error if no controller is found". I would love to have some opinions on this one https://github.com/juju/juju/pull/10985/files#diff-bcdb02984a8543f1187ed21cb4812a8fL206 Just opened a draft to see how one could handle it. Wrote more on the gh PR
[16:46] <rick_h> nammn_de1:  cool otp but will take a peek after lunch ty
[16:47] <nammn_de1> rick_h: sure its just initial draft to discuss possible solutions, as this would require changes on multiple places
[16:53] <achilleasa> manadart: I think I found the issue that broke that test...
[17:15] <manadart> achilleasa: nice.
[17:29] <achilleasa> manadart: when I added the new shims for testing the link layer device update path I accidentally made the StateMachine interface in the instancepoller test code incompatible with what FindEntity returned...
[17:29] <achilleasa> obviously it worked with the mocked state in the instancepoller package but exploded for that particular agent test
[17:54] <kwmonroe> hey gnuoy, i've read the upstream issues once per hour since you pinged me, and i still don't have a good feeling about the ramifications of removing allow_hosts.  punting to cory_fu because he loves issues like https://github.com/juju-solutions/layer-basic/issues/149
[17:58] <gnuoy> haha, sounds good to me
[17:59] <cory_fu> kwmonroe, gnuoy: Since we have a pre-bootstrap step to update pip & setuptools, as long as those get the new enough version to have the "never use easy_install" behavior for packages using install_requires, I think it should probably be ok.  Definitely should get plenty of testing, though.
[18:00] <cory_fu> Oh, it probably depends more on the version of pip & setuptools used by the build process, actually, rather than at deploy time.
[18:00] <cory_fu> Should make it easier to test, at least.
[18:03] <cory_fu> Oh, nm, that bit is in fact to ensure that the bootstrap phase doesn't hit pypi for some packages.  So it does require a deploy test, and it requires one in a network-restricted environment (or at least blocking pypi, or maybe checking the logs for indications that pypi is being used)
[18:03] <cory_fu> gnuoy: So, testing will be a pain, but I'm cautiously optimistic that it will work fine.
[18:05] <cory_fu> I really really hope we can come up with a better method for dependency management for charms soon (new framework, then maybe backported to reactive).  I'm hopeful that the ESM stuff can inform us on better patterns
[22:40] <babbageclunk> wallyworld: I'm updating the CrossModelRelations client to be able to talk to either v2 or v1 of the server API (expanding the v1 events into v2 ones if needed). I don't think I need to do that for the RemoteRelations, since that's inside one controller - the API version will always be >= than the "client" (actually a worker in the controller agent anyway).
[22:40] <babbageclunk> wallyworld: does that make sense to you?