Fix false-negative of unused-variable on catched exceptions

[ksaketou]

Steps

• Add yourself to CONTRIBUTORS if you are a new contributor.
• Add a ChangeLog entry describing what your PR does.
• If it's a new feature or an important bug fix, add a What's New entry in
  doc/whatsnew/<current release.rst>.
• Write a good description on what the PR does.

Description

This pull request fixes a false-negative of the unused-variable error message caused in case of catched exception inside another except block. It adds a new function for checking the content of the except block for raise and print statements and also other try/except blocks.

Type of Changes

	Type
✓	🐛 Bug fix

Related Issue

Closes #4391

[coveralls]

Coverage increased (+0.008%) to 91.854% when pulling cb7fe74 on ksaketou:unused-var into 6d449f9 on PyCQA:master.

[Pierre-Sassoulas]

Hi @ksaketou, thank you for the merge request. I had a look and the tests are very nice. I feel that the fix although it works, is too specific to the problem and I feel the performance and clarity of the code would be impacted. I don't know what would be the best way to fix this (It's possible to check how flake8 does it because it detect this error), but probably by doing something more generic like checking that the variable was not redefined before being used in all contexts, and not only when there is two imbricated try except. We could merge the test alone and add a TODO where pylint do not work properly yet so we can reuse them later if you want.

[ksaketou]

I agree that the code is more specific than it should be, but I think that this case is pretty specific as well so I don't know how I could implement it in a more general way. I thought that since the bug only existed in the try/except cases, that that's what I had to work on. I will check flake8 and try to fix the bug. So, I shouldn't focus only on the try/except cases? Any suggestions are welcome 😄

Of course we can merge just the tests for now until I figure this out :)

[ksaketou]

I'm working on a new branch for the addition of the tests but when I run python -m pytest, I see tests for other error messages like logging-not-lazy, invalid-length-returned etc failing. Maybe some changes happened on these messages?

[Pierre-Sassoulas]

Can you try to upgrade your dependencies in the virtualenv ? I think you need to have astroid 2.5.7 now.

[Pierre-Sassoulas]

Hey @ksaketou there was a nice comment on this topic here. I seriously underestimated the difficulty of the task at hand, this is clearly not a friendly issue for new contributor, so I apologize for that. Knowing that do you think we should close this merge request ?

[ksaketou]

Hello @Pierre-Sassoulas , it's okay, indeed I had difficulty finding a solution that works and is also elegant. I think we can close this PR.

[Pierre-Sassoulas]

Understood. Sorry again for the bad experience.