Latest version causing test failures

[vangheem]

Getting:

E           RuntimeError: Task <Task pending coro=<pytest_fixture_setup.<locals>.wrapper.<locals>.setup() running at /home/travis/build/plone/guillotina/eggs/pytest_asyncio-0.6.0-py3.6.egg/pytest_asyncio/plugin.py:120> cb=[_run_until_complete_cb() at /opt/python/3.6-dev/lib/python3.6/asyncio/base_events.py:176]> got Future <_GatheringFuture pending> attached to a different loop

on many of my tests now.

Downgrading to 0.5.0 works. Is there something I'm doing wrong in my tests?

[Tinche]

Not necessarily, might be a bug in our fixture handling code.

Can you get me a minimal example?

[vangheem]

I'll see if I have time to do that later. The package the test failures started in is open source though.

Here is failure output: https://travis-ci.org/plone/guillotina/builds/237114876

and here is the test server being created: https://github.com/plone/guillotina/blob/master/guillotina/tests/conftest.py#L198

But that's a lot of code to digest and not sure there is anything obvious there.

I'll see what I can do when I have some more time.

[Tinche]

As far as I can see, your tests are actually pytest-aiohttp, not pytest-asyncio tests. You have async def tests without the @pytest.mark.asyncio marker.

I see you have some async def fixtures too, which weren't supported in pytest-asyncio until 0.6. So these were being picked up by pytest-aiohttp too.

So I'm guessing now pytest-asyncio and pytest-aiohttp are in conflict.

[spumer]

I have failure on 0.6.0 too, but due setup event_loop fixture.

fixturedef = <FixtureDef name='event_loop' scope='function' baseid='' >
request = <SubRequest 'event_loop' for <Function 'test_get_gen_products'>>
... # here a lot of code
        if fixturedef.argname == "event_loop" and 'asyncio' in request.keywords:
            loop = outcome.get_result()
            for kw in _markers_2_fixtures.keys():
                if kw not in request.keywords:
                    continue
                policy = asyncio.get_event_loop_policy()
>               old_loop = policy.get_event_loop()


    def get_event_loop(self):
        """Get the event loop.
    
            This may be None or an instance of EventLoop.
            """
        if (self._local._loop is None and
            not self._local._set_called and
            isinstance(threading.current_thread(), threading._MainThread)):
            self.set_event_loop(self.new_event_loop())
        if self._local._loop is None:
            raise RuntimeError('There is no current event loop in thread %r.'
>                              % threading.current_thread().name)
E           RuntimeError: There is no current event loop in thread 'MainThread'.

In previous version asyncio-pytest did not require initialized old loop. But now...

[Tinche]

Do you have pytest-aiohttp installed too?

[spumer]

@Tinche yes, pytest-aiohttp==0.1.3

[Tinche]

Unfortunately pytest-asyncio isn't compatible with pytest-aiohttp any longer - you're seeing the conflicts.

[spumer]

Putting this code into conftest.py solve the problem

import asyncio
policy = asyncio.get_event_loop_policy()
policy.set_event_loop(policy.new_event_loop())

But where is real problem placed? pytest-aiohttp or pytest-asyncio?

[spumer]

The main fail when pytest-aiohttp installed - parameterized async fixtures did not work

[jstephenson]

Shame that they aren't able to coexist presently — I have a mix of code, some of which uses aiohttp and a good bit which doesn't. It would be preferred to use pytest-asyncio for the latter.

Nuking the default loop in aiohttp: https://github.com/aio-libs/aiohttp/blob/4ecb6339ddb60ad732e8a79a32ce75c10ec8546d/aiohttp/pytest_plugin.py#L144

[Zac-HD]

Closed by #52.