conftest: add docs and tests regarding CiTestCase's subp functionality

[OddBloke]

And fix up an inconsistency in tests/helpers.py too by raising a TypeError when subp is called with no arguments as it does under normal operation:

>>> from cloudinit.util import subp
>>> subp()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: subp() missing 1 required positional argument: 'args'

[TheRealFalcon]

Rubber stamp +1 😄

I don't understand why we have a separate subp or why we're disabling it's usage in tests, but the changes here make sense.

[OddBloke]

I don't understand why we have a separate subp

util.subp is a wrapper around subprocess.Popen, you can see it at https://github.com/canonical/cloud-init/blob/master/cloudinit/util.py#L2014. Honestly, as it was first written when we still supported Python 2.6 (and perhaps even earlier), there's probably a lot of it that could be dropped in favour of the stdlib at this point, but that would be a fairly large chunk of work.

or why we're disabling it's usage in tests, but the changes here make sense.

We've run into problems where tests pass in particular environments because the code under test is shelling out and getting the expected result back from the executable it ran, but fail in others. This ensures that we don't have any accidental dependency in our tests on binaries outside of cloud-init. (Also, given the things that cloud-init shells out to do to a system, it's in our interests to be very careful about this sort of thing.)

[smoser]

You should fix the comment 'assertion_error'. Or fix the _fake_subp to raise an assertion_error . Otherwise this is confusing. Your test is named ....raises_assertion_error... and then it insists that the more generic Exception is raised.

[OddBloke]

Yep, this was a copy/paste error from the above tests, will fix.

(The pytest functionality raises AssertionError but I didn't want to change the behaviour of CiTestCase "just" for the sake of consistency.)

[OddBloke]

Thanks for the review, Scott! I believe I've addressed both your comments.