Fix finding packages without an __init__.py

[DanielNoord]

Steps

• For new features or bug fixes, add a ChangeLog entry describing what your PR does.
• Write a good description on what the PR does.

Description

Closes #1327

Had no time to write tests yet, so if somebody can come up with something good: please take over this PR, you're more than welcome to.
Just wanted to illustrate that this is what should be added again.

This regression was introduced in 2ee20cc

Type of Changes

	Type
✓	🐛 Bug fix

Related Issue

@Pierre-Sassoulas, this might warrant a 2.9.3 release.

[interifter]

If you have a pointer to how to approach testing asteroid/pylint, in general, I might be able to help. It wouldn’t be too challenging to generate a temp directory that reproduces the problem.

Otherwise, I can write a test method the way I normally would that you can adapt?

[DanielNoord]

For some reason another test broke as well. I have got some spare time now, so I'll take a look myself. Thanks anyway 😄

[DanielNoord]

While I got the test suite to fail, I can't get the added test to fail without the change (thus it doesn't test anything..).

Would need to do some further investigation into what happened before the fauly commit in this function and what is happening now.

[Pierre-Sassoulas]

If you have a pointer to how to approach testing asteroid/pylint, in general, I might be able to help. It wouldn’t be too challenging to generate a temp directory that reproduces the problem.

I think recreating a temp directory that reproduce the issue is the way to go here. Thank you for your help, much appreciated !

Would need to do some further investigation

Could the problem come from the fact we removed the : os.path.expanduser(path)) ? It's the first thing that comes to mind.

[DanielNoord]

Test now fails with previous code and passes with new code.

The mistake was that we removed pathname = _cache_normalize_path(pathname). Whereas _canonicalize_path also normalised the path, all paths that came from path or [] where thus not normalised. This was problematic when path included "." and the normalised version of "." was not included in sys.path.
The tests also represent this scenario: we move into the tempdir and pass "." as path.

[Pierre-Sassoulas]

Let's release 2.9.3 with this.