Intermittent test failure: timeout from tasks in test_app.test_invite_scenarios

[njsmith]

Somehow test_invite_scenarios is sometimes failing with a timeout from asks.

This is weird, because that test uses monkeypatching to fake out http requests – it shouldn't be invoking asks at all!

Examples:

https://dev.azure.com/python-trio/snekomatic/_build/results?buildId=966
https://dev.azure.com/python-trio/snekomatic/_build/results?buildId=967

[njsmith]

Oh, we do make one real HTTP request: we use asks to deliver the fake webhook payload to our app. This is supposed to work reliably, because it only involves a localhost connection from our test process to itself. But the traceback shows the webhook URL, so this must be what's failing.

[njsmith]

...ah, and the reason it's failing is that we're using the autojump_clock, because I missed that we're doing real I/O here, so if the localhost connection doesn't succeed instantly, then it immediately times out.

So the simple fix would be to stop using the autojump clock, or to use it but set a non-zero "cushion" value. The reason we're using it is because our webhook handler always sleeps for 1 second, to give github's eventually-consistent databases to settle down. So the cost of not using it is just 1 second per test, and for a cushion to do any good, it would have to be set to less than a second, which is getting dicey with

So it sounds like for now we should just disable autojump_clock here. And we'll be able to re-enable it once we have an easy way to enable virtual networking for tests (python-trio/trio#170).

[njsmith]

Marking this a "good first issue" – the change is literally just to edit the definition of test_invite_scenarios and remove the autojump_clock argument.