[takluyver]

This is an alternative to #155007, keeping the file descriptor in the object and fixing the __del__ method to close it without trying to close the mmap or memoryview, which was my original idea in #155003.

I could add, in this PR or separately:

• The same posix_fallocate() change as in gh-155003: Don't keep file descriptor in multiprocessing SharedMemory object #155007 - but with this option it's orthogonal.
• A public API to access the fd, probably a .fileno() method, which is a convention for higher level objects which wrap an fd.
• Use mmap(..., trackfd=False) to skip the duplicate fd within mmap (at a cost of breaking potential workarounds using sm._mmap.resize() & .size())

• Issue: multiprocessing SharedMemory.__del__ can fail, leak file descriptor #155003

[takluyver]

Is there anything else I should be doing on this?

[gpshead]

This is likely good. I added a regression test (thank Claude). letting PR CI run and i'll kickoff a buildbot series after that.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @gpshead for commit 04a3ebb 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F155070%2Fmerge

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

[miss-islington-app]

Thanks @takluyver for the PR, and @gpshead for merging it 🌮🎉.. I'm working now to backport this PR to: 3.14, 3.15.
🐍🍒⛏🤖

[miss-islington-app]

Sorry, @takluyver and @gpshead, I could not cleanly backport this to 3.14 due to a conflict.

Please backport manually with cherry_picker, see the devguide for more information.

cherry_picker 4b79d90e935398442d65c59243f8015f062da0f8 3.14

[bedevere-app]

GH-161797 is a backport of this pull request to the 3.15 branch.

[bedevere-app]

GH-161798 is a backport of this pull request to the 3.14 branch.