Daniel Hahler

btw: I think in general more verbose asserts should be used, e.g. in this case: > `assert message["type"] == "http.response.start", message` or > `assert message["type"] == "http.response.start", message["type"]`

> Do I understand it correctly that the "http.response.template" extension is only used with tests? If so, its type could maybe have a "test." prefix? I came up with a...

It might also make sense to store this in `request.state` instead? But it shows that currently the middleware fails in case of unexpected messages being send - but maybe this...

> We might want to _always_ send the template message as the first message. This would basically mean to move it out of `body_stream` (in https://github.com/encode/starlette/pull/473/files#diff-ea43df86acedfeeae9fcba482f124a2cR49). > In that case...

See also #3844.

@h-vetinari As for handling this during collection time, see https://github.com/pytest-dev/pytest/issues/4377#issuecomment-438173521 for an example, and for docs: https://docs.pytest.org/en/latest/reference.html?highlight=pytest_collection_modifyitems#_pytest.hookspec.pytest_collection_modifyitems. I think it should work to remove the items that "do not make...

Just for reference: this looks related to a crash seen on CI here lately (#7022). /cc @bluetech From the gist: ``` (venv) C:\tmp>python -m pytest test_pywinauto.py ================================================= test session starts...

@aaltat re the gist above: was this really with `-p no:faulthandler`? (since you've said also that it would not happen when using that)

@aaltat no, it's fine - thanks for clarifying. I've edited my above comment.

Maybe related to https://github.com/pytest-dev/pytest/pull/4943 and/or https://github.com/pytest-dev/pytest/pull/4988. > Shouldn't pytest exactly be able to keep track of this, or at least wait until all test operations are completed so that any...