Avoid TypeError in call_historic()

[rmfitzpatrick]

The default argument for proc in _HookCaller.call_historic() is None. No attempt should be made to call it.

[goodboy]

Heh, thanks @rmfitzpatrick!

@RonnyPfannschmidt @nicoddemus I think we need some tests to cover this API. Can you guys point me to where this proc is even getting used traditionally - I'm assuming it's pytest-xdist? Maybe we should be decoupling the ad-hoc naming for this (proc: process or processor whaa?) and formalizing an API for how this is intended to be used (say maybe a callback kwarg) and documenting it?

[nicoddemus]

The only place I could find that uses is the call of the pytest_namespace hook:

        def do_setns(dic):
            import pytest
            setns(pytest, dic)

        self.hook.pytest_namespace.call_historic(do_setns, {})
        self.hook.pytest_addoption.call_historic(kwargs=dict(parser=self._parser))

You are right, callback probably would be a better name for it. I'm OK with changing it to a keyword-only argument and documenting it.

[RonnyPfannschmidt]

callback is a bad name - result_callback as at least needed to convey meaning - this processes results

[nicoddemus]

Agree that result_callback is an even better name than callback. 👍

This would warrant a 0.7 release though.

[RonnyPfannschmidt]

the old and the new name can be supported simultaneous, the old name should trigger a deprecation and we need pytest release as well

[goodboy]

Agree that result_callback is an even better name

Fair but what else would the callback be for ;P - either way I like explicit so result_callback works for me.

@rmfitzpatrick Do you mind adding a small test which would have caught this bug without your patch?

[nicoddemus]

Please add a test to avoid future regressions. 🙇

[goodboy]

@nicoddemus I've gone and added the requested test in #119.
Once this is merged that test should pass.

[goodboy]

Closing in favour of #119.