[CLI-2190] Resolve panic that can occur during auto-login without a role or environment

[Steven Gagniere (sgagniere)]

Checklist

1. [CRUCIAL] Is the change for CP or CCloud functionalities that are already live in prod?
   • yes: ok

What

A panic occurs in ccloudAutoLogin if a user account has no environments associated with it, which can happen when that user has no role-bindings. This PR makes two changes:

• Switch from direct access to using Get methods to prevent the panic
• Add an error if the environment is still nil after setting the context.

References

https://confluentinc.atlassian.net/browse/CLI-2190

Test & Review

Ran tests
Manual testing:

• I Invited myself as a new user to my org and checked that running commands which require Authentication does not panic, and instead returns the new error
• Checked that basic functionality still works: login/logout, help, update, version.
• Checked that a user with an environment, but without an active environment, can still run commands without the new error.

[Brian Strauch (brianstrauch)]

Are you assuming that all authenticated commands require an environment? What about confluent environment create?

[Brian Strauch (brianstrauch)]

Never mind, we're assuming the only way a user can have 0 environments is when they have no role bindings.

[Brian Strauch (brianstrauch)]

This seems a bit like a general-purpose bandaid, though... have you tried fixing the panics closer to where they occur?

[Brian Strauch (brianstrauch)]

Also- we should think about writing system tests for this as a follow-up, since this is an issue that's come up twice in the last few months.

[Steven Gagniere (sgagniere)]

Yeah my reasoning was basically as follows: an organization will have at least one environment that can't (normally) be deleted. Every user with a role is able to view the list of environments, so any user account w/ sufficient permissions to create an environment should not be blocked by this.

• My only concern is what happens if a user somehow manages to succeed in deleting all of their environments even though the backend doesn't allow this. I think you mentioned that this has happened before; how was it recovered?

I did fix the panic itself, this block of code is actually downwind of that. The reason I added this was based on some discussion w/ David Hyde (@DABH) about what we should do if this occurs. I think the reasoning was that we want an environment to really do anything meaningful, so we may as well just error out and tell the user about it.

Edit: I stuck it here because it seemed like a reasonably high level place so I wouldn't have to reflect the change across multiple files.

Actually, how's this: instead of an error, I issue this as a warning instead. It will function exactly the same as it does now (minus the panic), but it will alert the user that they're probably missing a role. Edit: And this should sidestep the edge case where someone who should be able to create environments would be blocked.

[David Hyde (DABH)]

Warning seems ok. I kind of like having something like this at a high level because having 0 environments is such an edge case that something is very likely very wrong (e.g. user has no rolebindings) if they hit this scenario.

[Brian Strauch (brianstrauch)]

I think you mentioned that this has happened before; how was it recovered?

I believe CDMUM had to step in and manually create a new default environment