Add support for Python 3.14

[jacobtylerwalls]

Type of Changes

	Type
✓	✨ New feature

[jacobtylerwalls]

Blocked by pylint-dev/astroid#2789

[cdce8p]

#10480 will help resolve a lot of the issues here. On the astroid side I just created pylint-dev/astroid#2793 to add basic support for TemplateStr.

[Pierre-Sassoulas]

Not blocked anymore with astroid 4.0.0b1. Not sure how to best deal with the no-member / no-name-in-module though. I'm not sure that the conditional functional test based on the interpreter are going to work for the expected output too.

[cdce8p]

Not blocked anymore with astroid 4.0.0b1. Not sure how to best deal with the no-member / no-name-in-module though. I'm not sure that the conditional functional test based on the interpreter are going to work for the expected output too.

Will take a proper look at it soon. It at least uncovered a false-positive for used-before-assignment I'll address in a separate PR.

[cdce8p]

CI is looking good now. Interestingly I'm still getting an error locally.

pytest tests/test_functional.py::test_functional[regression_5408]

Crashes with Abort trap: 6 on my machine, MacOS ARM with 3.14.0rc1. If I compile cpython from source it works fine. Maybe it's a fluke?

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.87%. Comparing base (d363fca) to head (8853532).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main   #10467   +/-   ##
=======================================
  Coverage   95.86%   95.87%           
=======================================
  Files         176      176           
  Lines       19168    19168           
=======================================
+ Hits        18376    18377    +1     
+ Misses        792      791    -1     

Files with missing lines	Coverage Δ
pylint/checkers/base/basic_error_checker.py	95.41% <100.00%> (ø)

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Pierre-Sassoulas]

Amazing ! Regarding the abort trap, I don't know, but I think that's outside pylint's purview.

I don't remember if I was the one who removed the new "Forward" functional tests specifically for 3.14 or if it was you (importing from annotationlib instead of typing), do you think it make sense to add it back ?

[cdce8p]

Amazing ! Regarding the abort trap, I don't know, but I think that's outside pylint's purview.

Would you mind running the test on your machine? If the error is more common, it might still be related / caused by pylint or astroid. Though at the moment I suspect it's just some issue on my machine.

I don't remember if I was the one who removed the new "Forward" functional tests specifically for 3.14 or if it was you (importing from annotationlib instead of typing), do you think it make sense to add it back ?

I removed it while pushing 8c4750a. It's not really necessary since with pylint-dev/astroid#2795, typing.ForwardRef just delegates to annotationlib.ForwardRef anyway. Without 8c4750a the test case would fail even with the import from typing.

[github-actions]

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

This comment was generated for commit 8853532

[Pierre-Sassoulas]

Would you mind running the test on your machine?

Sure, but I have ubuntu 24.04.2 LTS (tested with python 3.14 alpha 6 and python 3.14 rc1 installed with uv), no crashes.

If the error is more common, it might still be related / caused by pylint or astroid.

"revealed" right ? As in it's a bug in cpython that we can maybe avoid triggering with some guesswork ?

[cdce8p]

If the error is more common, it might still be related / caused by pylint or astroid.

"revealed" right ? As in it's a bug in cpython that we can maybe avoid triggering with some guesswork ?

Yeah, if it was just on my machine it's probably nothing. However, if it can be reproduced reliably that's a different story.
Looks like it's the latter. Was able to reproduce it in a CI run: https://github.com/cdce8p/pylint/actions/runs/16805911305/job/47598155830

Removing these lines here is enough for the error to go away but that's not really the point.

pylint/tests/functional/r/regression_02/regression_5408.py

Lines 35 to 37 in d363fca

	class Child(_Child):
	def patch(cls):
	MyClass.sub_class.inner_class = cls

Looks like I need to investigate that a bit further.

--
Unless you think otherwise I'd say it doesn't block this PR though. The changes here are itself correct and the issue seems to be with cpython on MacOS exclusively AFAICT at the moment.