handle case when no event loop exists (support pytest-aiohttp)

[smagafurov]

Solves this:

• Latest version causing test failures #52

And this:

• No event loop in the thread 'MainThread' when using test_client fixture aio-libs/aiohttp#1069

Now pytest-asyncio and pytest-aiohttp can act together with this conftest.py:

from aiohttp.test_utils import loop_context


@pytest.yield_fixture
def loop():
    with loop_context() as _loop:
        yield _loop


@pytest.yield_fixture
def event_loop(loop):
    yield loop

[smagafurov]

@asvetlov, hi

see conftest.py in my comment above

How do you think, is it ok to run pytest-asyncio and pytest-aiohttp together in such manner?

[spumer]

Yea! This is very usefull!

[asvetlov]

Looks good at first glance

[argaen]

I really should check open PR first before doing mine 😓 . Looks good to me too, only thing I'm missing there is a test, you can check #65, copy paste for the test should do it.

@Tinche any intentions on merging this? :)

[Tinche]

The patch looks simple enough, I'd be willing to merge it in. But wouldn't a better course of action be to reach out to pytest-aiohttp and see if they would be willing to depend on pytest-asyncio, instead of merging in workarounds?

[argaen]

Now it's pytest-aiohttp the one having this edge case but anyone can do an asyncio.set_event_loop(None) in a pytest fixture and then pytest-asyncio starts crashing. IMO this should be solved regardless of what other libraries do

[smagafurov]

test added (copy pasted from #65)

[bugok]

+1

I'm hitting this problem as well. Would be happy to see this getting merged in.

[Tinche]

Alright, thanks for the contribution. My plea for pytest-asyncio and pytest-aiohttp to work better together still stands :)

[smagafurov]

can we release minor version with this fix?

[Tinche]

I'll get it out over the weekend.