[issue44442] Globals (and presumably builtins) are cleared premuturely in FrameObject

STINNER Victor added the comment: I marked bpo-44288 as a duplicate of this issue. -- ___ Python tracker ___ ___ Python-bugs-list

[issue44442] Globals (and presumably builtins) are cleared premuturely in FrameObject

STINNER Victor added the comment: The bpo number is missing from the merged commit, it's: commit ba2f32a983a08c4f64c23c187432e38908639c12 Author: Mark Shannon Date: Thu Jun 17 16:29:15 2021 +0100 Do not clear globals or builtins when calling clear() on a frame object. Reverts

[issue44442] Globals (and presumably builtins) are cleared premuturely in FrameObject

STINNER Victor added the comment: See also bpo-44288: can it be a duplicate? -- ___ Python tracker ___ ___ Python-bugs-list

[issue44442] Globals (and presumably builtins) are cleared premuturely in FrameObject

Mark Shannon added the comment: No problem, I've added a simple test. -- stage: patch review -> ___ Python tracker ___ ___

[issue44442] Globals (and presumably builtins) are cleared premuturely in FrameObject

Change by Mark Shannon : -- resolution: -> fixed stage: -> resolved status: open -> closed ___ Python tracker ___ ___

[issue44442] Globals (and presumably builtins) are cleared premuturely in FrameObject

STINNER Victor added the comment: > Victor, you've mentioned this problem. Did you have a specific example I can > add as a test? Sorry, I didn't have any reproducer. I vaguely recall an error in the unittest module when getting globals from a frame. --

[issue44442] Globals (and presumably builtins) are cleared premuturely in FrameObject

Change by Mark Shannon : -- keywords: +patch pull_requests: +25355 stage: -> patch review pull_request: https://github.com/python/cpython/pull/26768 ___ Python tracker ___

[issue44442] Globals (and presumably builtins) are cleared premuturely in FrameObject

New submission from Mark Shannon : When calling frame.clear(), the globals (and builtins) are cleared. This is not the case in 3.10. We should restore the 3.10 behavior, as there is no reason not to. Victor, you've mentioned this problem. Did you have a specific example I can add as a test?