gh-137007: Track executor before any possible deallocations#137016
Uh oh!
There was an error while loading. Please reload this page.
Conversation
Fidget-Spinner requested review from brandtbucher and removed request for markshannon
| executor->vm_data.warm= true; | ||
| if (_PyJIT_Compile(executor, executor->trace, length)){ | ||
| Py_DECREF(executor); | ||
| returnNULL; |
There was a problem hiding this comment.
Note to reviewers: the bad dealloc happens right above this line at line 1233/1238 depending on which diff you view.
There's no reproducer because I can't force _PyJIT_Compile to fail except in specific scenarios, or when the GC decides to kick in right when the executor is linked, but not yet tracked. |
devdanzin commented Jul 22, 2025
Unfortunately the patch doesn't fix the issue for me. I double checked that it's applied and still get the abort. Investigating whether the diff necessary to reproduce is the same. |
…/cpython into track-before-deallocating
Fidget-Spinner commented Jul 22, 2025
@devdanzin can you please try again? |
devdanzin commented Jul 22, 2025
Still aborting here. Does it stop reproducing for you? |
picnixz left a comment
picnixz left a comment There was a problem hiding this comment.
In the other PR, when allocating the cold executor, there is an error path before making it immortal. Should we track the cold executor before it is immortal and untrack it before making it immortal?
Fidget-Spinner commented Nov 7, 2025
@devdanzin does it still reproduce now? |
devdanzin commented Nov 9, 2025
Unfortunately I cannot reproduce the crash even without the patch applied. Tried all known MREs, tinkered with them for over two hours to try to make them repro, and fuzzed with the MREs as seeds. No luck reproducing. |
markshannon left a comment
markshannon left a comment There was a problem hiding this comment.
Looks good.
I'm a bit surprised that the first commit didn't fix this. Did you find out why that was?
Fidget-Spinner merged commit 97e1901 into python:mainUh oh!
There was an error while loading. Please reload this page.
Thanks @Fidget-Spinner for the PR 🌮🎉.. I'm working now to backport this PR to: 3.14. |
…thonGH-137016) (cherry picked from commit 97e1901) Co-authored-by: Ken Jin <kenjin@python.org>
GH-142541 is a backport of this pull request to the 3.14 branch. |
4 participants
Uh oh!
There was an error while loading. Please reload this page.