Switching to Lab App for OBO testing #163
Conversation
rayluo
left a comment
rayluo
left a comment
There was a problem hiding this comment.
It is good to see the test passes again. Thanks! As we discussed earlier, we would also want to understand the why. I leave a couple questions inline.
| os.getenv("LAB_OBO_CLIENT_SECRET"), "OBO app secret not found") | ||
| def test_acquire_token_obo(self): | ||
| # Some hardcoded, pre-defined settings | ||
| obo_client_id = "23c64cd8-21e4-41dd-9756-ab9e2c23f58c" |
There was a problem hiding this comment.
Per our earlier discussion, have we investigated why our previous test setup suddenly failed in the first place? Most of those previous settings were obtained from MSAL .Net's OBO test case, which seems to be still working?
There was a problem hiding this comment.
MSAL .Nets OBO test case is an app set up in Travis's personal tenant. He had preconfigured the apps to work with the lab account we use. There was a change in the cloud accounts returned by lab api recently. The account we end up using to run these tests was not consented on Travis's app. While we can grant this consent for the first app using interactive auth, we will have to rely on Travis to add consent to this account for the second confidential client app in this test case.
I confirmed with Travis that .NET test cases are failing too because of this change.
So, the best thing to do is move to using the accounts provided by lab so that they take care of this consent internally when they add new accounts that are returned by their API.
Aside, I am talking to Gladwin about how these API changes can be communicated so that we dont break the tests.
tests/test_e2e.py
Outdated
| @unittest.skipUnless( | ||
| os.getenv("OBO_CLIENT_SECRET"), | ||
| "Need OBO_CLIENT_SECRET from https://buildautomation.vault.azure.net/secrets/IdentityDivisionDotNetOBOServiceSecret") | ||
| os.getenv("LAB_OBO_CLIENT_SECRET"), "OBO app secret not found") |
There was a problem hiding this comment.
It is OK to switch to a different set of test setup, but then I would suggest to follow the previous implementation, to leave some hint for future troubleshooting (in case the test case would somehow fail again), and/or for auditing (i.e. to understand where those test settings come from).
rayluo
left a comment
rayluo
left a comment
There was a problem hiding this comment.
Thanks for all the investigation!
No description provided.