gh-119666: fix multiple class-scope comprehensions referencing __class__ - #120295
carljm merged 2 commits into
Conversation
Sorry, something went wrong.
|
Is there any chance this also fixes #119233? |
Sorry, something went wrong.
|
Fine with me. I think the test cases here probably cover both reproducers? |
Sorry, something went wrong.
|
Yeah, the essence of both reproducers is the same: >1 comprehension, in class scope, referencing super or __class__. |
Sorry, something went wrong.
|
Thanks @carljm for the PR 🌮🎉.. I'm working now to backport this PR to: 3.12, 3.13. |
Sorry, something went wrong.
…__class__ (pythonGH-120295) (cherry picked from commit 0ae8579) Co-authored-by: Carl Meyer <carl@oddbird.net>
|
GH-120299 is a backport of this pull request to the 3.13 branch. |
Sorry, something went wrong.
…__class__ (pythonGH-120295) (cherry picked from commit 0ae8579) Co-authored-by: Carl Meyer <carl@oddbird.net>
|
GH-120300 is a backport of this pull request to the 3.12 branch. |
Sorry, something went wrong.
It took a few months, but the fuzzer discovered that the fix in #115139 was not quite sufficient. It was only applied in the case where the __class__ symbol did not exist in the outer class scope, but it actually needs to be applied regardless.
Verified that all listcomp tests (apart from test_frame_locals, which is now intentionally diverging from pre-3.12 comprehension behavior, since the implementation of PEP 667) pass both with comprehension inlining enabled and disabled, so they reflect the same behavior as before Python 3.12.