Log RecursionError out as warning during inference

[jacobtylerwalls]

Partner to pylint-dev/astroid#2385

Type of Changes

	Type
✓	🐛 Bug fix

Description

Closes #9139

[codecov]

Codecov Report

Attention: Patch coverage is 19.58042% with 115 lines in your changes are missing coverage. Please review.

Project coverage is 95.25%. Comparing base (fd6790b) to head (ed4fe10).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #9614      +/-   ##
==========================================
- Coverage   95.83%   95.25%   -0.59%     
==========================================
  Files         174      174              
  Lines       18888    19014     +126     
==========================================
+ Hits        18102    18111       +9     
- Misses        786      903     +117     

Files	Coverage Δ
pylint/checkers/base/comparison_checker.py	100.00% <ø> (ø)
pylint/checkers/refactoring/refactoring_checker.py	98.06% <100.00%> (-0.21%)	⬇️
pylint/checkers/newstyle.py	90.90% <0.00%> (-5.25%)	⬇️
...heckers/refactoring/implicit_booleaness_checker.py	97.60% <0.00%> (-2.41%)	⬇️
pylint/extensions/typing.py	96.02% <62.50%> (-1.66%)	⬇️
pylint/checkers/classes/special_methods_checker.py	92.34% <0.00%> (-3.14%)	⬇️
pylint/checkers/stdlib.py	95.03% <14.28%> (-1.92%)	⬇️
pylint/checkers/strings.py	93.00% <0.00%> (-1.28%)	⬇️
pylint/pyreverse/utils.py	90.34% <14.28%> (-3.86%)	⬇️
pylint/checkers/base/basic_checker.py	96.01% <0.00%> (-2.07%)	⬇️
... and 4 more

[jacobtylerwalls]

Maybe we should do this in astroid after all?

[github-actions]

🤖 According to the primer, this change has no effect on the checked open source code. 🤖🎉

This comment was generated for commit ed4fe10

[Pierre-Sassoulas]

Great ! Clearly a lot of work went into this. I'm not sure New contributor or plugin maintainer will be able to keep up in new code. The exception catching around infer seem almost mandatory and not very DRY now. Maybe we could reraise an inference error with warning from astroid directly ? Or rework safe_infer / infer ?

Other than that I think it's pretty hard to know how much this will be shown in case of a problematic code base. I assume it can happens from more than once, to too much. Maybe we need an antispam limit or some ux refinement (one warning per full lint, with the number of total error ? One warning per second max ?)

[jacobtylerwalls]

Clearly a lot of work went into this.

No, don't worry, this went pretty fast, happy to discuss alternatives :-)

Maybe we could reraise an inference error with warning from astroid directly ?

Yeah, I'm thinking we should do that after all.