Relax timings of subshell tests
Looking into improving the flakiness of some of the subshell tests, hoping to fix #1359.
The failures in CI all appear on Windows, either CPython 3.11 or PyPy. On a local Windows machine the tests pass for my using CPython 3.11 but I can reproduce the failures with PyPy 3.9. Increasing the timings of the tests so that the timing variability is proportionately smaller works for me locally, so trying it here in CI.
Would it be possible to use events instead of relying on timings?
Would it be possible to use events instead of relying on timings?
Well yes, the tests could be rewritten in some other way. But that is not my intention here.
Tests were passing fine when I submitted this. Now that it is rebased on top of the zmq-anyio changes there are lots of failures.
Yes, although mostly around subshells. It could be unrelated to subshells, but since you know best how they work, maybe you could help understand the issue?