| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2019-03-16 | |||
| 01:05:18 | openstackgerrit | Merged openstack/python-openstackclient master: Fix: incorrect check when no shared/private input https://review.openstack.org/642708 | |
| #openstack-sdks - 2019-03-17 | |||
| 00:08:49 | openstackgerrit | Merged openstack/openstacksdk master: Add proxy API reference to baremetal user guide https://review.openstack.org/643683 | |
| 18:38:29 | openstackgerrit | Eric Fried proposed openstack/openstacksdk master: WIP Make factory for a CloudRegion from CONF objects https://review.openstack.org/643601 | |
| 18:40:47 | efried | mordred: Updated ^ (please refresh if you had local edits) | |
| 20:20:42 | openstackgerrit | Eric Fried proposed openstack/openstacksdk master: WIP Make factory for a CloudRegion from CONF objects https://review.openstack.org/643601 | |
| 20:20:55 | efried | mordred: and again (added oslo.config to requirements) | |
| 22:09:48 | adriant | mordred: I guess at some point I should add Adjutant to the SDK now that I have a reasonable service type | |
| 23:42:25 | mordred | efried: sweet. left a comment about adding oslo.config. | |
| 23:42:27 | mordred | adriant: yes! | |
| 23:50:19 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: WIP Make factory for a CloudRegion from CONF objects https://review.openstack.org/643601 | |
| 23:50:27 | mordred | efried: ^^ took a stab at it | |
| #openstack-sdks - 2019-03-18 | |||
| 00:22:35 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: WIP Make factory for a CloudRegion from CONF objects https://review.openstack.org/643601 | |
| 00:22:39 | mordred | I'm stupid | |
| 00:31:08 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: WIP Make factory for a CloudRegion from CONF objects https://review.openstack.org/643601 | |
| 10:47:44 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Add unit tests for connection.add_service https://review.openstack.org/643604 | |
| 11:03:33 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Create runtime descriptor for new service https://review.openstack.org/643918 | |
| 12:37:39 | gtema | mordred, dtantsur, Shrews: I would like to ask you to have a look at https://review.openstack.org/#/c/642218/ I am currently not very happy how it feels and want to know your opinions | |
| 12:49:18 | mordred | gtema: I think it looks great! | |
| 12:50:15 | gtema | ok, I just dislike usage of stuff, which is not belonging to the object and is only present in runtime | |
| 12:50:53 | gtema | but if you guys are ok with that we can push it and do probably some followup | |
| 12:51:38 | gtema | thanks mordred | |
| 12:56:27 | mordred | gtema: well, what do you mean - maybe I'm missing the thing that's bothering you | |
| 12:56:52 | mordred | gtema: the mixin classes do make the things part of the object | |
| 12:57:38 | gtema | sure, and that is exactly the "problem" for me. It is ok when the class use stuff from Mixin, but I do no like when one Mixin use stuff from another Mixin | |
| 12:58:35 | gtema | so I am more a fan of a regular inheritance, which is clear from the object description | |
| 12:59:51 | gtema | something what I have done with _network_common. The most "clear" way would be if _compute, _network, _floating_ip would inherit from it, and not indirectly use stuff, which is added on a higher level | |
| 13:01:01 | gtema | this could have also solved initialization of mixins in the connection - how it is currently done is not nice | |
| 13:04:33 | gtema | ok, letme quickly remove WIP from the title | |
| 13:05:01 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Split OpenStackCloud into reasonable pieces https://review.openstack.org/642218 | |
| 13:07:59 | mordred | gtema: ah - yes. I understand your concern there | |
| 13:08:45 | gtema | mordred: great | |
| 13:09:44 | mordred | gtema: I think we'll be fighting to balance a desire to organize stuff into files with the mixin mess just because python doesn't have a better way to organize implementation | |
| 13:10:00 | gtema | agreed | |
| 13:10:41 | gtema | ok, will continue "thinking" about that issue | |
| 13:11:05 | mordred | all that said - I was originally thinking that once we got things moved into the proxy objects that potentially most of the remaining methods could just turn in to very small wrapper methods - and we could just shove all of OpenStackCloud directly on to Connection | |
| 13:11:18 | mordred | but it would still be thousands of lines long even in the best case | |
| 13:12:40 | gtema | yes, that's clear. With this change you have only "basic" logic left in the OpenStackCloud and it can be further merged into Connection. But here it is exactly the most prominent that mixins access methods, which they do not really belong | |
| 13:12:59 | mordred | yeah | |
| 13:15:24 | gtema | mordred: any plans for SDK on PTG? | |
| 13:17:19 | dtantsur | I feel like Forum may be a bit better for SDK | |
| 13:17:59 | gtema | I mean both, I will be hopefully going to summit (tomorrow appointment in embassy) including PTG | |
| 13:20:06 | dtantsur | I'll probably spend the whole PTG in the ironic room | |
| 13:20:22 | gtema | ok | |
| 13:20:33 | mordred | I don't think we have any current plans to - in the past most of us have either been absent or too consumed with other rooms | |
| 13:20:56 | gtema | got it | |
| 13:20:59 | mordred | dtantsur: was there a forum session about the migrate-to-osc goal? | |
| 13:21:14 | gtema | hmm, I haven't seen | |
| 13:21:29 | gtema | shame on me | |
| 13:21:45 | dtantsur | I haven't seen one either | |
| 13:21:49 | mordred | gtema: that said - I'd be happy to sit down in either location and figure stuff out | |
| 13:22:08 | dtantsur | s/either location/a bar/ | |
| 13:22:26 | gtema | ok, great. If I'm there - definitely we will find a place :D | |
| 13:40:24 | dtantsur | gtema: good luck with the interview! | |
| 13:54:37 | gtema | dtantsur: thanks | |
| 13:59:49 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Split OpenStackCloud into reasonable pieces https://review.openstack.org/642218 | |
| 14:14:19 | mordred | gtema: maybe read http://www.qtrac.eu/pyclassmulti.html and then the Steven D'Aprano reply on https://groups.google.com/forum/?hl=en#!topic/comp.lang.python/goLBrqcozNY | |
| 14:14:55 | mordred | gtema: there are maybe some ideas there that you might dislike less | |
| 14:15:19 | mordred | the class-level import-from is a *fascinating* idea I'd never considered | |
| 14:15:59 | dtantsur | yeah, it's quite creative. I'm not sure if it's a good thing or not ;) | |
| 14:16:03 | mordred | me either | |
| 14:16:27 | mordred | like - I'm not sure it would be better than the mixin approach - but thought I'd drop it in here for consideration | |
| 14:18:05 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Fix wait_for_server docstring https://review.openstack.org/643955 | |
| 14:21:09 | Shrews | hrm. my first impression of that is "ewwww" | |
| 14:22:20 | dtantsur | I know a simple solution: let's write a metaclass to inject base classes in runtime | |
| 14:23:42 | Shrews | fwiw, we did somewhat split up that code in the past. but then put it back together again for some reasons lost to me | |
| 14:31:45 | mordred | Shrews: I'm not sure we ever actually landed the patch to do the split | |
| 14:31:55 | mordred | Shrews: I feel like we started working on it then gave up | |
| 14:32:34 | dtantsur | there was a user cloud/admin cloud split, I think | |
| 14:34:55 | mordred | brtknr: I am also wondering about that - since we have gate jobs for 3.6 and 3.7 | |
| 14:36:31 | brtknr | mordred: Sorry, just realised that this is a problem with python-magnumclient so it might not apply to this | |
| 14:37:16 | Shrews | mordred: oh we landed it. messed git history up tremendously well | |
| 14:37:57 | mordred | Shrews: neat! | |
| 14:38:26 | mordred | brtknr: awesome. well - for that I can't help much :) | |
| 14:39:46 | openstackgerrit | OpenStack Release Bot proposed openstack/keystoneauth master: Update master for stable/stein https://review.openstack.org/643996 | |
| 14:41:19 | openstackgerrit | OpenStack Release Bot proposed openstack/openstacksdk master: Update master for stable/stein https://review.openstack.org/644016 | |
| 14:41:57 | openstackgerrit | OpenStack Release Bot proposed openstack/os-client-config master: Update master for stable/stein https://review.openstack.org/644024 | |
| 14:42:34 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: WIP Make factory for a CloudRegion from CONF objects https://review.openstack.org/643601 | |
| 14:42:57 | mordred | efried: ^^ took a stab at your suggestion from PS8 | |
| 14:43:18 | openstackgerrit | OpenStack Release Bot proposed openstack/osc-lib master: Update master for stable/stein https://review.openstack.org/644042 | |
| 14:43:43 | efried | mordred: yeah, nice, guess we'll see if it works, I'll recheck the dependents. | |
| 14:43:49 | mordred | cool | |
| 14:44:18 | mordred | efried: once we get a version of it that we're happy with for the downstream patches, I'll work on some unittests for it | |
| 14:45:46 | efried | mordred: any thoughts on the `valid_interfaces` debacle? | |
| 14:46:27 | mordred | efried: you mean the need to support valid_interfaces as an alternate way of saying interface? | |
| 14:47:10 | efried | mordred: Well, ksa adapter conf options have valid_interfaces and deprecate interface. And the thing we're doing here is accepting ksa adapter conf opts. | |
| 14:47:49 | efried | mordred: But the logic that converts the valid_interfaces conf opt into the internal `interface` thingy - which accepts a list IIUC - only lives in the load-from-conf ksa code | |
| 14:47:57 | mordred | heh. well - openstacksdk accepts a list or a scalar for interface | |
| 14:48:12 | efried | which is currently embedded in a factory method that can't easily be invoked from sdk code, from what I can see. | |
| 14:48:13 | mordred | yeah - valid_interfaces is really only a thing in oslo.config files - it's not the actual interface anywhere else | |
| 14:48:36 | efried | Proposal: factor the logic in ksa loading into a helper method that we can call. | |
| 14:49:06 | mordred | we could - except I don't want to deprecate interface in openstacksdk - so we'd have to make sure not to throw any warnings | |
| 14:49:49 | mordred | (interface is the better option name - I think the issue is that oslo.config doesn't support defining a single thing as list-or-single-value and that's why we had to add valid-interfaces right? | |
| 14:49:52 | efried | mordred: I refer to: https://github.com/openstack/keystoneauth/blob/master/keystoneauth1/loading/adapter.py#L223-L232 | |
| 14:50:12 | mordred | AH | |
| 14:50:14 | efried | I think it's because we had already defined 'interface' as not a list opt? | |
| 14:50:15 | mordred | that | |
| 14:50:36 | mordred | yeah. and oslo.config can't have an option that's both a listopt and a stringopt | |
| 14:50:36 | efried | yeah, we don't need it to sling a warning. We just need to accept (and in fact prefer) valid_interfaces over interface in oslo conf. | |
| 14:50:43 | mordred | ++ | |
| 14:50:45 | mordred | agree with that | |