clarkb | oh wait no it does | 00:00 |
---|---|---|
clarkb | endpoint = self._get_highest_endpoint(service_types, kwargs) | 00:00 |
mordred | jamielennox: sigh. nope. and there it is | 00:00 |
clarkb | where service types is service_types = [self.get_service_type(service_key)] | 00:00 |
clarkb | so just broken cloud config? | 00:00 |
mordred | nothing sets interface=admin | 00:00 |
mordred | yup | 00:00 |
clarkb | and kwargs there has interface set | 00:00 |
mordred | and becaues of how occ was using the previous ksc workaround, it was masking it | 00:00 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Set interface=admin for operator_cloud functional tests https://review.openstack.org/460784 | 00:01 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Set interface=admin for operator_cloud functional tests https://review.openstack.org/460784 | 00:02 |
mordred | jamielennox: ok - a) thank you - I'm not sure how far down the rabbit hole I would have gotten before finding a missing interface :) and b) ^^ I think that should work around it for at least a sec until I can get a patch to devstack to write those clouds.yaml files properly | 00:03 |
clarkb | mordred: you should update your devstack patch too for completeness | 00:03 |
mordred | oh yah - I shall do that indeed | 00:03 |
clarkb | and then we also want to make sure clouds.yaml is using v3 by default | 00:03 |
* clarkb checks a master job | 00:03 | |
jamielennox | mordred: yea, you've hit like the only section of openstack where setting an interface really matters | 00:03 |
clarkb | mordred: http://logs.openstack.org/75/460675/1/check/gate-tempest-dsvm-neutron-full-ubuntu-xenial/76e7cab/logs/etc/openstack/clouds.yaml.txt.gz is master | 00:04 |
clarkb | mordred: it uses v3 but the other issues are all there too | 00:04 |
jamielennox | afaik, that section of keystone v2 is the only thing that is only available on the admin url | 00:04 |
clarkb | so you want to make change against master and backport | 00:04 |
mordred | clarkb: yah | 00:04 |
jamielennox | mordred: i don't know the shade tests well, but wont that set a fairly global interface='admin'? | 00:05 |
jamielennox | like you probably only want it for those keystone v2 admin calls | 00:05 |
mordred | https://review.openstack.org/#/c/460786/ | 00:09 |
clarkb | mordred: reviewed I think bashate will trip on the extra tab there | 00:09 |
clarkb | otherwise lgtm | 00:09 |
mordred | jamielennox: we keep two cloud configs- one that's a normal user and one that's an admin user | 00:09 |
mordred | jamielennox: so that'll set interface=admin for all the admin operations, which should be a reasonable fine thing to do, yeah? | 00:10 |
jamielennox | mordred: yep, i just mean even for like the nova admin calls you probably want to use public | 00:10 |
jamielennox | mordred: oh, you don't know the admin interface saga? | 00:10 |
mordred | jamielennox: I think I've blocked it from my poor brain | 00:11 |
clarkb | why have different interfaces if nevermind I don't want to know its 5pm and I am going for a walk | 00:11 |
mordred | clarkb: it was an early "security" feature, iirc | 00:11 |
jamielennox | so basically the admin interface was only really ever used by keystone, it was decided that things like adding services, etc was only done by a cloud admin and so should be done on a different port and service to everything else | 00:12 |
jamielennox | because we don't trust RBAC? i don't know | 00:12 |
mordred | because FEAR | 00:12 |
jamielennox | so really the whole :35357 port thing should never have actually been exposed to the public | 00:13 |
mordred | jamielennox: so - we should not add interface: admin to the clouds.yaml files for the admin user in devstack - and instead just special case a thing in shade for those tests | 00:13 |
jamielennox | but keystoneclient, a CLI used almost primarily by operators decided it should default to admin interface because it was silly to have people type it in every time | 00:13 |
jamielennox | defaults stick... | 00:13 |
jamielennox | and the world burns | 00:13 |
* mordred sobs | 00:14 | |
jamielennox | mordred: yea, special case like the 5 method calls that need the admin interface, because it's fairly undefined behaviour whether the adminURL in nova catalog can be hit from wherever you are | 00:14 |
mordred | kk. cool | 00:15 |
mordred | and we _really_ only need that for keystonev2 right? | 00:15 |
jamielennox | mordred: yep, one of the things for v3 was to completely remove the special admin cases | 00:16 |
jamielennox | realistically if your deployment set up DNS correctly we could get the whole catalog thing down to 1 url | 00:17 |
jamielennox | ( per service/region) | 00:17 |
mordred | jamielennox: but dns is scary | 00:17 |
jamielennox | right, that would be us as developers telling admins how to set up their deployment | 00:18 |
jamielennox | which is simultaneously heresy and what they complain about when there is no standard method | 00:18 |
mordred | jamielennox: I kinda feel like we should also put in a patch to occ that is "if service_type=identity and api_version=2: interface=admin" - since it literally won't work otherwise | 00:23 |
jamielennox | mordred: it's limited to only those calls though, the OS-KSADM ones | 00:23 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Set interface=admin for keystonev2 keystone tests https://review.openstack.org/460784 | 00:23 |
jamielennox | list users etc will work fine on public | 00:24 |
mordred | jamielennox: but v2 didn't haveany non-admin calls, did it? | 00:24 |
mordred | ah - gotcha | 00:24 |
mordred | so it's just services and endpoints? | 00:24 |
jamielennox | and because admin is only expected for cloud admins it generally an internal ip | 00:24 |
mordred | jamielennox: user password also seems to do OS-KSADM | 00:25 |
jamielennox | mordred: only the admin reset though | 00:26 |
mordred | yah | 00:26 |
mordred | well - hopefully that patch will pass - 'm going to WIP the devstack patches for now | 00:26 |
mordred | jamielennox: so really - the only reason this worked before is because the ksc hack was telling ksc to use the auth_url for the endpoint and not anything from the catalog (which is what the ksc hack was all about) COMBINED withthe fact that devstack puts the admin url into clouds.yaml as the auth_url | 00:28 |
jamielennox | but yea, in for example your create_service() shade call i would expect you to force pass interface='admin' for v2 because otherwise it just doesn't work | 00:29 |
mordred | jamielennox: yah - but we can't pass interface=admin on a call | 00:29 |
jamielennox | yep, and it's wrong that devstack is putting the admin_url into clouds.yaml | 00:29 |
mordred | or, at least, not without a major restructuring | 00:29 |
jamielennox | mordred: you can't via client, you can via adapter | 00:29 |
mordred | ah - good point- we can make that more robust once we're done restifying | 00:30 |
mordred | jamielennox: ok - I pushed up devstack patches for master, stable/ocata and stable/newton making the change to which endpoint goes into clouds.yaml | 00:37 |
mordred | between that and the shade test fixes that _should_ be good - thank you very much for the hepl! | 00:38 |
jamielennox | mordred: i feel like i've fixed that in the past, someone always changes it back | 00:38 |
jamielennox | no worries | 00:38 |
mordred | people LOVE unfixing fixes | 00:38 |
*** slaweq has joined #openstack-shade | 00:39 | |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Add in a bunch of TODOs about interface=admin https://review.openstack.org/460790 | 00:42 |
*** slaweq has quit IRC | 00:44 | |
rods | mordred you around? | 01:07 |
rods | about the idea of altering the message attribute of the exception we were talking before | 01:10 |
rods | something like this seems to work(line 19-20) http://paste.openstack.org/show/608271/ | 01:10 |
*** gouthamr has quit IRC | 01:44 | |
*** gouthamr has joined #openstack-shade | 01:44 | |
*** yolanda has quit IRC | 01:55 | |
*** slaweq has joined #openstack-shade | 02:41 | |
*** slaweq has quit IRC | 02:45 | |
*** gouthamr has quit IRC | 03:50 | |
*** gkadam has joined #openstack-shade | 03:51 | |
*** gouthamr has joined #openstack-shade | 03:51 | |
*** gouthamr has quit IRC | 03:56 | |
*** slaweq has joined #openstack-shade | 04:42 | |
*** slaweq has quit IRC | 04:42 | |
*** slaweq has joined #openstack-shade | 04:44 | |
*** gkadam__ has joined #openstack-shade | 04:48 | |
*** gkadam has quit IRC | 04:49 | |
*** slaweq has quit IRC | 04:49 | |
*** gkadam__ is now known as regain | 04:49 | |
*** regain is now known as gkadam | 04:49 | |
*** gkadam has quit IRC | 04:50 | |
*** gkadam has joined #openstack-shade | 04:51 | |
*** slaweq has joined #openstack-shade | 05:45 | |
*** slaweq has quit IRC | 05:49 | |
*** yfried has joined #openstack-shade | 09:10 | |
*** cdent has joined #openstack-shade | 09:14 | |
*** purplerbot has joined #openstack-shade | 09:16 | |
*** yfried has quit IRC | 09:22 | |
*** cdent has quit IRC | 10:58 | |
*** gkadam has quit IRC | 11:29 | |
*** cdent has joined #openstack-shade | 11:35 | |
*** slaweq has joined #openstack-shade | 11:50 | |
*** slaweq has quit IRC | 11:54 | |
mordred | rods: ++ that looks great | 12:12 |
Shrews | mordred: rods: that paste doesn't make sense to me, actually. what are you trying to do? looks to me like you'd lose the original exception's message when the original intent was to let that info pass through | 12:30 |
mordred | Shrews: so - the problem here is that with restification we have OpenStackCloudExceptions that get thrown by the adapter without nice error messages | 12:31 |
mordred | (they're things like "Not found") | 12:31 |
mordred | so as we restify, we're actually now losing friendly error messages because we're letting adapter throw them on bad http codes | 12:32 |
mordred | this change would let us still wrap those calls in with shade_exceptions and inject a good error message, but keep the traceback in the exception intact | 12:32 |
Shrews | mordred: but with that change, you are now affecting code like this: https://github.com/openstack-infra/shade/blob/master/shade/openstackcloud.py#L795-L796 | 12:35 |
Shrews | instead of raising an exc saying "project not found", you replace the original exception message | 12:35 |
Shrews | and in that particular example, it's less descriptive | 12:35 |
mordred | Shrews: that's an excellent point | 12:36 |
mordred | Shrews, rods: how about we make a new context_manager that only does the message thing - and we add it only right around calls to the adapter? | 12:37 |
mordred | that way we're only doing it for places we know the error message from the adapter is lame and not accidentally for other places in shade thatuse the shade_exceptions context manager | 12:37 |
Shrews | yeah, i think you need the fix closer to the adapter | 12:38 |
Shrews | mordred: now that i look at that paste more closely, i don't think it would actually do anything. 'args' is only processed in __init__(), so i don't think that does what was intended | 12:49 |
mordred | the other idea I had (although I like rods use of the context_manager better) was to add an "error message" option to the adapter, but that feels clunky | 12:50 |
mordred | Shrews: args gets passed to the super ctr | 12:50 |
mordred | and then __str__ does message = Exception.__str__(self) | 12:50 |
Shrews | mordred: yes, but only when first constructed. you aren't constructing a new exception | 12:51 |
mordred | oh right. duh | 12:52 |
Shrews | the 'raise' by itself seems wrong, too. in any case, i feel that's the wrong approach anyway | 12:52 |
Shrews | mordred: why can't you just use the context manager as-is within the adapter? | 12:53 |
mordred | yah. piddle. I think you're right. darn. I liked the outcome of that approach code-wise | 12:53 |
mordred | Shrews: the adapter itself is long-lived, so we'd have to pass the error message we want produced to each call | 12:53 |
mordred | OR - make a context manager that takes the adapter and an error message as an argument and returns us a thing that looks like the adapter ... | 12:54 |
mordred | hang on - that didn't make sense- let me write that in non-working code real quick | 12:56 |
Shrews | mordred: where are the NOT FOUND exceptions being throw from (in shade)? gimme an example | 12:56 |
Shrews | from raw 'post' calls? | 12:56 |
mordred | Shrews: in shade/exc.py there is a function "raise_from_respose" | 12:56 |
mordred | Shrews: in _munch_response on ShadeAdapter in shade/_adapter.py, the first thing done is calling raise_from_response on the respose | 12:57 |
mordred | response | 12:57 |
Shrews | *nod* | 12:57 |
mordred | Shrews: if you look at https://review.openstack.org/#/c/460357/1/shade/openstackcloud.py | 12:58 |
mordred | line 3879 | 12:58 |
Shrews | so... | 12:58 |
mordred | you can see how it kind of goes - the post call will throw an exception there because of raise_from_response, but it will not have any error message like the previous context manager provided | 12:58 |
Shrews | just make raise_from_response throw normal Exception objects, then use shade_exceptions() to wrap the actual calls and give a useful error message? | 12:59 |
mordred | we could just make it post('/volumes', json=dict(volume=kwargs), error_msg="Error in creating volume") and we'd have something to pass to raise_from_response | 12:59 |
mordred | Shrews: well - the thing is - exception wrapping always makes following tracebacks harder, and I was hoping we could get out of the business if we can throw the right exception from the error site now that we'r enot using the python libs | 13:00 |
Shrews | actually, it seems the exceptions raised from raise_from_response are not useful anyway if the messages themselves aren't useful | 13:01 |
mordred | they're useful-ish in that there are a few different exception classes depending on status_code | 13:01 |
mordred | so they give you some ways to handle the errors in code which are nice - but they depend on the servers for the human-readable error message | 13:02 |
mordred | hrm. I also see some dead code in here | 13:04 |
Shrews | well, if you're going to always override the text of the exception (regardless of status code), then they aren't useful | 13:04 |
mordred | yah | 13:04 |
Shrews | but i like your post example above | 13:04 |
mordred | I've got two ideas - let me sketch them up real quick and see what you think | 13:04 |
* Shrews should coffee more first | 13:05 | |
rods | mordred Shrews hello | 13:11 |
*** Matias has quit IRC | 13:13 | |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Add optional error_message to adapter.request https://review.openstack.org/461025 | 13:13 |
mordred | rods, Shrews: ^^ what about that? (completely untested) | 13:13 |
Shrews | mordred: that might work. if you only wanted it for certain status codes, you could instead pass in a mapping from code to message | 13:15 |
Shrews | 404 => "omg what went wrong????" | 13:16 |
Shrews | e.g | 13:16 |
mordred | Shrews: yes - I agree - it would not be difficult to extend it in that way | 13:17 |
mordred | although honestly if we have callsites that want to do that, I think it would be easier for those to just ask for the response un-munched and do the direct response processing themselves :) | 13:18 |
* Shrews must spitballing. you're obviously more familiar with what's needed than i | 13:19 | |
Shrews | s/must/just/ | 13:19 |
rods | mordred I would be more inclined to test your first idea about the context_manager for the message to apply when needed | 13:21 |
rods | but I'm not that familiar with the code base yet and I could be very wrong :) | 13:23 |
Shrews | rods: the context manager idea doesn't work b/c it was designed for existing shade exceptions to pass through unchanged | 13:25 |
Shrews | it's original purpose is to make sure that the code it wraps doesn't let a non-shade exception get through, AND to not alter any shade exceptions that may be thrown | 13:27 |
mordred | Shrews: would you mind helping me figure outwhat's wrong with https://review.openstack.org/#/c/460784 ? | 13:27 |
mordred | Shrews: the errors say AttributeError: 'TestIdentity' object has no attribute 'config' | 13:27 |
Shrews | mordred: easy... the author! | 13:27 |
Shrews | looking | 13:28 |
mordred | but when I look at the chain of setUp methods, I clearly see self.config being set | 13:28 |
mordred | so I'm very confused | 13:28 |
Shrews | mordred: could that be a reaction to the pep8 error there? | 13:29 |
Shrews | http://logs.openstack.org/84/460784/3/check/gate-shade-pep8-ubuntu-xenial/f1fb982/console.html#_2017-04-28_00_26_19_403493 | 13:29 |
mordred | Shrews: for the love of | 13:30 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Set interface=admin for keystonev2 keystone tests https://review.openstack.org/460784 | 13:35 |
mordred | Shrews: maybe that'll be better | 13:35 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Set interface=admin for keystonev2 keystone tests https://review.openstack.org/460784 | 13:36 |
*** cdent has quit IRC | 14:05 | |
*** gouthamr has joined #openstack-shade | 14:44 | |
Shrews | mordred: KeystoneBaseFunctionalTestCase should derive from BaseFunctionalTestCase, not base.TestCase (the cause of the current errors) | 14:46 |
mordred | Shrews: thank you sir | 14:47 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Set interface=admin for keystonev2 keystone tests https://review.openstack.org/460784 | 14:48 |
*** cdent has joined #openstack-shade | 14:52 | |
*** openstackgerrit has quit IRC | 15:17 | |
*** yolanda has joined #openstack-shade | 15:43 | |
*** slaweq has joined #openstack-shade | 16:22 | |
*** cdent has quit IRC | 16:29 | |
*** openstackgerrit has joined #openstack-shade | 16:45 | |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Set interface=admin for keystonev2 keystone tests https://review.openstack.org/460784 | 16:45 |
*** slaweq has quit IRC | 17:04 | |
*** slaweq__ has joined #openstack-shade | 17:06 | |
*** slaweq__ has quit IRC | 17:07 | |
*** slaweq has joined #openstack-shade | 17:19 | |
*** slaweq has quit IRC | 17:24 | |
openstackgerrit | Monty Taylor proposed openstack/os-client-config master: Fix interactions with keystoneauth from newton https://review.openstack.org/461114 | 17:47 |
*** slaweq has joined #openstack-shade | 17:48 | |
mordred | Shrews: oh - I was wrong, we ARE testing git tips of libraries properly in our gate-shade-dsvm-functional-legacy-libs-nv jobs | 17:55 |
mordred | Shrews: so I'm not sure why we didn't notice the occ-caused newton issue before | 17:55 |
*** slaweq has quit IRC | 18:06 | |
*** slaweq has joined #openstack-shade | 18:06 | |
*** yolanda has quit IRC | 18:07 | |
mordred | Shrews: also, https://review.openstack.org/#/c/460784/ fixes the gate | 18:09 |
mordred | clarkb: ^^ that's the fix for yesterday's fun games with newton | 18:09 |
clarkb | mordred: you aren't going to fix in devstack? | 18:10 |
mordred | clarkb: two different fixes it turns out | 18:10 |
mordred | clarkb: the devstack fix won't actually fix the shade issue - it'll just make devstack correct | 18:10 |
clarkb | isn't shade using the devstack clouds.yaml though which would set the interface? | 18:11 |
Shrews | mordred: something is terribly amiss in the nv job: http://logs.openstack.org/84/460784/7/check/gate-shade-dsvm-functional-legacy-libs-nv/f6da27a/console.html#_2017-04-28_17_10_06_185339 | 18:11 |
*** slaweq has quit IRC | 18:11 | |
mordred | shade-side - for those operations in keystone v2 we should be explicitly requesting the admin interface | 18:11 |
mordred | clarkb: nothing else should _ever_ use the admin interface by default it turns out | 18:11 |
mordred | clarkb: so setting interface=admin in clouds.yaml would not be correct | 18:11 |
mordred | Shrews: yah - bug in latest occ release - patch just pushed up to fix that | 18:11 |
clarkb | mordred: right so the issue is that to be an admin you have to know a shibboleth, have access to a special port, and change all your urls | 18:12 |
Shrews | i did not think 'shibboleth' was an actual word | 18:15 |
Shrews | heh | 18:15 |
*** yolanda has joined #openstack-shade | 18:16 | |
clarkb | its like securing your home by giving everyone that should have access keys, but then only letting them in if they know to go around the back and knock with a special knock | 18:17 |
clarkb | mordred: I've +2'd the change but wonder if shade shouldn't also grow a more verbose error message if you are using admin actions that require the admin interface and admin isn't set? | 18:22 |
openstackgerrit | David Shrewsbury proposed openstack-infra/shade master: Add in a bunch of TODOs about interface=admin https://review.openstack.org/460790 | 18:23 |
mordred | clarkb: ah - so - as soon as we migrate these calls to REST - we can enforce interface=admin | 18:23 |
mordred | clarkb: I dropped in a bunch of todo comments here: https://review.openstack.org/#/c/460790/ | 18:23 |
mordred | clarkb: to remind us to do it | 18:23 |
mordred | we can't do it _before_ then, because python-keystoneclient is not flexible enough | 18:23 |
Shrews | just approved that and the other thing | 18:23 |
mordred | Shrews: woot. thakns | 18:24 |
clarkb | mordred: what we can do is if keystone admin and not interface=admin emit verbose error instead of 404ing? | 18:24 |
mordred | clarkb: yes - we can do that now | 18:24 |
mordred | "this call is about to fail, instead of making it, I'm going to go ahead nad stop and tell you so" | 18:24 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Add optional error_message to adapter.request https://review.openstack.org/461025 | 18:26 |
clarkb | mordred: ya because as an end user a 404 in the bowels of the o[penstack api isn't super helpful | 18:27 |
mordred | clarkb: ++ | 18:27 |
*** slaweq has joined #openstack-shade | 18:33 | |
*** slaweq has quit IRC | 18:55 | |
openstackgerrit | Merged openstack-infra/shade master: Set interface=admin for keystonev2 keystone tests https://review.openstack.org/460784 | 19:04 |
mordred | rods, slaweq_: ^^ tests should be working again | 19:05 |
*** slaweq has joined #openstack-shade | 19:20 | |
openstackgerrit | Slawek Kaplonski proposed openstack-infra/shade master: Replace neutronclient with REST API calls in router commands https://review.openstack.org/460518 | 19:34 |
openstackgerrit | Merged openstack-infra/shade master: Add in a bunch of TODOs about interface=admin https://review.openstack.org/460790 | 19:52 |
rods | mordred cool, thx | 20:04 |
rods | anything I can do to help with https://review.openstack.org/#/c/461025/? | 20:04 |
mordred | rods: I'm poking at it right now... let's see what stupid thing I've done now :) | 20:08 |
mordred | wow. I'm good. what a silly typo :) | 20:09 |
rods | :) | 20:13 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Add optional error_message to adapter.request https://review.openstack.org/461025 | 20:26 |
openstackgerrit | Slawek Kaplonski proposed openstack-infra/shade master: Remove neutronclient mocks from ports tests https://review.openstack.org/461149 | 20:41 |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Add some more debugging to the post_test_hook https://review.openstack.org/460779 | 20:50 |
openstackgerrit | Slawek Kaplonski proposed openstack-infra/shade master: Remove neutronclient mocks from quotas tests https://review.openstack.org/461150 | 20:59 |
rods | mordred tests passed on https://review.openstack.org/#/c/461025/3 | 21:22 |
rods | once that commit is merged, I'll push a new ps for https://review.openstack.org/#/c/460357/1 | 21:24 |
openstackgerrit | Monty Taylor proposed openstack/os-client-config master: Fix interactions with keystoneauth from newton https://review.openstack.org/461114 | 21:35 |
mordred | awesome | 21:36 |
mordred | Shrews: if you're still around, I think you liked the approach in https://review.openstack.org/#/c/461025 - but do you want to chime in on it? | 21:37 |
rods | mordred functional tests run against a newton devstack, right? | 21:51 |
mordred | rods: in the legacy tests, yes. in the normal tests it's normal devstack | 21:53 |
rods | ok, thx | 21:53 |
*** cmurphy has quit IRC | 21:59 | |
*** cmurphy has joined #openstack-shade | 22:05 | |
*** gouthamr has quit IRC | 22:14 | |
*** slaweq has quit IRC | 22:22 | |
morgan | mordred: shade is doing something wrong with get domain | 22:55 |
morgan | mordred: trying to chase it down | 22:55 |
morgan | mordred: when you do a get_domain_by name it just does ".get_domain" still vs an actual search | 22:55 |
morgan | so https://identity.example.com/v3/domains/<domain_name> | 22:56 |
morgan | found this out in the conversion away form mock | 22:56 |
mordred | morgan: oh good! | 22:58 |
*** slaweq has joined #openstack-shade | 23:00 | |
morgan | yeaaaaah | 23:00 |
morgan | #funtimes | 23:00 |
morgan | my brain was trying to figure out how the heck we needed to mock domains/domain_name vs domains/domain_id | 23:01 |
morgan | it dawned on me something broken was happening | 23:01 |
morgan | it might be KSC broken fwiw | 23:01 |
openstackgerrit | Morgan Fainberg proposed openstack-infra/shade master: Convert test_role_assignments to requests mock https://review.openstack.org/448313 | 23:03 |
mordred | morgan: blasphemy!!! | 23:04 |
* morgan has 24 more tests to fix there and a follow up to fix the domain behavior | 23:05 | |
morgan | only 2400 LOC so far | 23:05 |
morgan | ... in short... Sorry! | 23:05 |
morgan | it is effectively un-reviewable except "do tests run" | 23:05 |
mordred | yup | 23:06 |
*** slaweq has quit IRC | 23:06 | |
mordred | it's fun to read and be like "wow, that's not efficient" - but it's definitely what the original code does - yay testing frameworks | 23:07 |
*** slaweq has joined #openstack-shade | 23:07 | |
morgan | i wish i could make it better... but really, i can't | 23:07 |
*** slaweq has quit IRC | 23:12 | |
*** rods has quit IRC | 23:13 | |
openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Add some more debugging to the post_test_hook https://review.openstack.org/460779 | 23:15 |
*** rods has joined #openstack-shade | 23:17 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!