fix: require the same object to be in the testing state as in the event #1468
Conversation
…matching identifiers.
|
Why did a bunch of tests get cancelled tho? |
Yeah, something weird is happening there - the cancelled ones are because one failed, which is for some strange "environment inconsistent" thing, which I have not seen before, and doesn't appear related to the changes. I'll look into it (and it obviously blocks merging anyway). |
Basically, it's this: In main: >>> import dataclasses
>>> import ops
>>> from ops import testing
>>> ctx = testing.Context(ops.CharmBase, meta={'name': 'n', 'requires': {'foo': {'interface': 'i'}}})
>>> r = testing.Relation('foo')
>>> r2 = dataclasses.replace(r, interface='x')
>>> r.id, r2.id
(1, 1)
>>> ctx.run(ctx.on.relation_changed(r), testing.State(relations={r2}))
State(config={}, relations=frozenset({Relation(endpoint='foo', interface='x', id=1, local_app_data={}, local_unit_data={'egress-subnets': '192.0.2.0', 'ingress-address': '192.0.2.0', 'private-address': '192.0.2.0'}, remote_app_name='remote', limit=1, remote_app_data={}, remote_units_data={0: {'egress-subnets': '192.0.2.0', 'ingress-address': '192.0.2.0', 'private-address': '192.0.2.0'}})}), networks=frozenset(), containers=frozenset(), storages=frozenset(), opened_ports=frozenset(), leader=False, model=Model(name='qzYc1M0vTmxETBYBRce4', uuid='888c4943-2e56-4913-bf1b-7bb760fbb562', type='kubernetes', cloud_spec=None), secrets=frozenset(), resources=frozenset(), planned_units=1, deferred=[], stored_states=frozenset({StoredState(name='_stored', owner_path=None, content={'event_count': 4}, _data_type_name='StoredStateData')}), app_status=UnknownStatus(), unit_status=UnknownStatus(), workload_version='')
>>> In the branch: >>> import dataclasses
>>> import ops
>>> from ops import testing
>>> ctx = testing.Context(ops.CharmBase, meta={'name': 'n', 'requires': {'foo': {'interface': 'i'}}})
>>> r = testing.Relation('foo')
>>> r2 = dataclasses.replace(r, interface='x')
>>> r.id, r2.id
(1, 1)
>>> ctx.run(ctx.on.relation_changed(r), testing.State(relations={r2}))
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/tameyer/code/operator/testing/src/scenario/context.py", line 665, in run
with self._run(event=event, state=state) as manager:
File "/usr/lib/python3.12/contextlib.py", line 137, in __enter__
return next(self.gen)
^^^^^^^^^^^^^^
File "/home/tameyer/code/operator/testing/src/scenario/context.py", line 687, in _run
with runtime.exec(
File "/usr/lib/python3.12/contextlib.py", line 137, in __enter__
return next(self.gen)
^^^^^^^^^^^^^^
File "/home/tameyer/code/operator/testing/src/scenario/runtime.py", line 435, in exec
check_consistency(state, event, self._charm_spec, self._juju_version)
File "/home/tameyer/code/operator/testing/src/scenario/_consistency_checker.py", line 125, in check_consistency
raise InconsistentScenarioError(
scenario.errors.InconsistentScenarioError: Inconsistent scenario. The following errors were found: cannot emit foo_relation_changed because relation 2 is not in the state (a relation with the same ID is not sufficient - you must pass the object in the state to the event).
>>> In main, |
Not sure what happened here. One test failed with a weird environment error that seems unrelated to the changes, and when I re-ran it, it passed. Then there were several others that were cancelled since that one failed, but they all passed when I re-ran them. I'm assuming this was some GitHub workflows glitch. |
This was incorrectly changed during the import of the ops-scenario code to this repo. As discussed in #1429, having an object in the state that has the same identifier as the object in the event is insufficient and could end up being confusing: it should be the same actual object.
As well as reverting the previous change, adjust the error messages so that they're clearer if this case is triggered.
Fixes #1429