Simple run_pymodules_install test, refs #1898

[AndreMiras]

Regression test for the fix introduced in #1898.

[AndreMiras]

The aim is definitely not to deeply test anything, but to at least check that the bug address in #1898 is now covered by a simple test

[inclement]

This reminds me we still need to decide when p4a drops python2 support. Soon would probably be good, it's especially annoying nowadays to do python2 compatibility stuff when it will be necessary for only a short time.

[AndreMiras]

Completely agree. Also I can't wait to use f-strings and annotations :smile

[opacam]

Yes, I'm also waiting for python 3's f-strings, but this will require python >= 3.6...I have no problem with that

As a side note: we could already use python 3 f-strings but now it would require the install of some extra python package like fmt (I've been tempted of doing this in some prs 😆)

[opacam]

Is there any reason to non inherit from unittest.TestCase?

I think that inheriting from unittest.TestCase has several advantages:

• Allows some ides to detect the unittest class and you can run it from the ide with a couple of mouse clicks (pycharm case...probably eclipse too...but has been years since I used it...so I'm not sure of this last one)
• Allows us to use the assert methods of unit.TestCase, which increases the readability

Also I may add that the class unittest.TestCase has been made specially to deal with tests, I don't see the reason of not using it...am I missing something?

[AndreMiras]

Yes the reason is, don't use something if you don't need it, as simple as that 😄
Also known as YAGNI https://en.wikipedia.org/wiki/You_aren%27t_gonna_need_it
I mean we use pytest from tox which advocates for light code with no unnecessary boiler plate. https://docs.pytest.org/en/latest/assert.html

pytest has support for showing the values of the most common subexpressions including calls, attributes, comparisons, and binary and unary operators. (...). This allows you to use the idiomatic python constructs without boilerplate code while not losing introspection information.

I hope nowadays IDE can work with pytest, but I don't know because I've never used any.

As for the readability I actually think assert is far more readable than self.assertSomething.
The human eye just has to scan for assert it's always this keyword with always a space straight after. So no silly camelCase (shame on you Python!), no self. and no variations (assertEqual, assertTrue, assertIsNone, assertIsNotNone...). So I honestly thing the core unittest module makes test less readable than going straight with pytest and plan assert.

Nevertheless, I've added the inheritance boilerplate for you, I know it's gonna take time to convince people, so let's try one step at a time 😄

[opacam]

LGTM!

Thanks @AndreMiras 😄