Add test for relative-beyond-top-level false positive

[DanielNoord]

• Add yourself to CONTRIBUTORS if you are a new contributor.
• Write a good description on what the PR does.

Type of Changes

	Type
✓	🐛 Bug fix

Description

This builds upon the fix and later regression fixes introduced in pylint-dev/astroid#1186, pylint-dev/astroid#1200, pylint-dev/astroid#1211

I think running pylint on all the files in the test folder might be a bit overkill but I wanted to show my working solution and get feedback first. As said in the astroid PR I'm not 100% comfortable with namespace packages and importing, but this seems to fix the issue!

Closes #2967
Closes #5131

[DanielNoord]

Do we want to add the extra tests added to astroid for this change to be added to pylint as well?

[coveralls]

Pull Request Test Coverage Report for Build 1350996452

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.01%) to 93.236%

---

Totals
Change from base Build 1350956681:	0.01%
Covered Lines:	13633
Relevant Lines:	14622

---

💛 - Coveralls

[Pierre-Sassoulas]

Do we want to add the extra tests added to astroid for this change to be added to pylint as well?

Well you're right it's notre strictly necessary, but pylint is modifying the path and doing complex thing on top of astroid generally so I would deem keeping them prudent.

[cdce8p]

@DanielNoord Can you add the example from pylint-dev/astroid#1200 here as well?

[Pierre-Sassoulas]

Sorry I forgot that we need an entry in the changelog.

[DanielNoord]

@DanielNoord Can you add the example from PyCQA/astroid#1200 here as well?

Done!

Sorry I forgot that we need an entry in the changelog.

Do we? This was a change in astroid without any code changes in pylint, only tests. I know for a similar case we didn't add anything.

[cdce8p]

@DanielNoord Can you add the example from PyCQA/astroid#1200 here as well?

Done!

Thanks!

Sorry I forgot that we need an entry in the changelog.

Do we? This was a change in astroid without any code changes in pylint, only tests. I know for a similar case we didn't add anything.

It's not strictly required, but probably better. Don't know how many users will actually read the astroid changelog. Just a one liner + 'closes' should be enough though. It might work to reuse the astroid changelog entry for that. (That's what I did before merging your requires astroid update PRs today.)

[DanielNoord]

Hm, tests pass on Mac, but fail on windows...
If anybody has any good clues as to why this might be happening, let me know!

[DanielNoord]

Can we merge this before we get to the bottom of #5131? Or do we want to wait for that issue as well?

[Pierre-Sassoulas]

I think we'll need to add a regression test for the other issue before 2.12 goes out but we don't have to do it in this PR necesssarily. I don't fear conflicts if we don't merge this immediately though.

[DanielNoord]

This should be fixed by pylint-dev/astroid#1211

See explanation of the latest regression there.

[DanielNoord]

At some point we might need to check if the changes to the import resolver actually fixed some of the reported issues with namespace packages.