pebble_check_failed/pebble_check_recovered events always fail consistency check for containers with dashes in their names?

[DnPlas]

When testing pebble_check_failed or pebble_check_recovered events for a container whose name contains dashes (e.g. temporal-worker), the consistency checker unconditionally raises InconsistentScenarioError even when the check_infos are correctly set on the container in the state.

I believe that n _consistency_checker.py, all_checks is built using raw container names all_checks = {(c.name, check.name) for c in state.containers for check in c.check_infos}, but evt_container_name (a bit below that) is derived from event.name, which has been normalised.

To reproduce

In a scenario test case:

check_info = ops.testing.CheckInfo("my-check", failures=3, status=ops.pebble.CheckStatus.DOWN)
container = dataclasses.replace(
      state.get_container("temporal-worker"),
      check_infos={check_info},
)
state = dataclasses.replace(state, containers=[container])
context.run(context.on.pebble_check_failed(container, check_info), state) # <-- will raise InconsistentScenarioError

[tonyandrewmeyer]

Looks a lot like #2511, although we can check to be sure.

[tonyandrewmeyer]

Looks a lot like #2511, although we can check to be sure.

Ah, no, that was runtime, this is Scenario only, although they're both similar normalisation bugs. We handle all the other types of objects, just not notices, oddly.