@mknyszek requested issue #72991 to be considered for backport to the next 1.24 minor release.
@gopherbot Please open a backport issue for Go 1.24.
This is a regression in 1.24 only. Unfortunately the fix uncovered some subtly wrong and scary things about asynchronous preemption. The changes include a fix, but given how subtle it is, we have decided to let it soak for a little bit before actually committing to the backport. Apologies to those waiting. There's a risk of introducing random crashes into programs not affected by this issue.
Comment From: mknyszek
This context is lost in the comment copied over, but we still do want to backport a fix for this, sometime before the Go 1.25 release. The bug is preventing users from upgrading to Go 1.24 from Go 1.23. The fix itself is small and safe, in particular when not considering asynchronous preemption. Assuming we don't see any show-stoppers in the next month or so, I think we can safely backport the fix.
Comment From: kostix
@mknyszek, looks like we've hit this issue here at my $dayjob (1.24.4 vs 1.23.9). Is there any commit we could build Go with and see whether it fixes our issue?
Comment From: mknyszek
Yup, try patching Go 1.24 with https://go.dev/cl/682356.
Comment From: gopherbot
Change https://go.dev/cl/682356 mentions this issue: [release-branch.go1.24] runtime: prevent unnecessary zeroing of large objects with pointers
Comment From: kostix
Yup, try patching Go 1.24 with https://go.dev/cl/682356.
Tried by cherry-picking 8f82bf4 on top of the go1.24.4 tag.
Unfortunately, it did not fix the problem for us, which is not good: basically this means, that our code and/or workload appears to trigger some other regression in 1.24.
Comment From: cagedmantis
Approved as this is a regression.
Comment From: gopherbot
Closed by merging CL 682356 (commit 390ffce7d60abf7d5317c0a78beef64affd0059a) to release-branch.go1.24.
Comment From: mknyszek
@mknyszek, looks like we've hit this issue here at my $dayjob (1.24.4 vs 1.23.9). Is there any commit we could build Go with and see whether it fixes our issue?
Apologies for the late reply. I'd encourage you to file a new issue with as many details as you can share, if you haven't already.