Propagate failures in scope level fixture

[mezhaka]

I have a pattern that is used for integration tests in two scope level fixtures now: 1) start a Kubernetes pod and open a tunnel to it; 2) run a service with Docker and expose it's details, such that all test use the service. The code looks like this:

@pytest.fixture(scope="session", autouse=True)
def setup_service():
    # do some `docker run` stuff
        yield
    # raise an error if the container died

Sometimes the service crashes, which results in connection errors in all tests--the snippet above gives me an ERROR at teardown in the very beginning of the pytest output and a list of connection error failures in each test.

The issue is that for the users of the tests it is not clear why exactly the connection error happened--no explicit connection between the the connection errors and a teardown failure, which reports the crashed container. Could you think of a way to make this connection obvious? Ideally, I would rather abort all tests if the docker container went down.

Could someone suggest a way to achieve that?

So far the only idea I have is to create a function scoped fixture which will check the status of the service before the start--for that to work, I need to figure out how to share state between the scope level fixture that creates container and a function level fixture that checks the status of the container,

[mezhaka]

One can share the state between the scope level fixture and function level fixture like this:

class State:
    def is_service_running() -> bool: pass
    
@pytest.fixture(scope="session", autouse=True)
def first():
    yield State()
    
@pytest.fixture(scope="function", autouse=True)
def second(first):
    if not first.is_service_running:
        pytest.fail("duh")
    yield
    if not first.is_service_running:
        pytest.fail("duh")

With this solution, I get the "duh" for each failed test in the test summary and errors when trying to set up a test, which makes it clear what the problem is.