👋 Welcome back shade! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

@pchilano, would be happy to hear what you think here :)

❗ This change is not yet ready to be integrated.
See the Progress checklist in the description for automated requirements.

This backport pull request has now been updated with issue from the original commit.

Webrevs

• 00: Full (a2a0d68c)

Thank you for looking! I'll stress test it a bit more and then request approval for 25u.

Stress tests pass, see PR body. Oh, but we should do JDK 26 first: openjdk/jdk26u#111

/approval request Fixes another Virtual Threads issue. The issue reproduces in 25u, a regression since JDK 24. The patch does not apply cleanly due to minor conflicts, easy to resolve. Reproducer fails without the fix, passes with it. Stress tests pass. Was in mainline for 1.5 months without bugtail. We expect that users would see this issue in real deployments, if not fixed. Risk is medium: touches VT paths, but it fairly mild manner and fixes the VT corner case.

@shipilev
8373120: The approval request has been created successfully.

Hi @shipilev
Are you sure this needs to go to 25.0.4 (July), would it be fine to defer it by one?
As I understand there are no reports from applications running into this.

Hi @shipilev Are you sure this needs to go to 25.0.4 (July), would it be fine to defer it by one? As I understand there are no reports from applications running into this.

Yes, this is more in "taking a calculated risk" territory.

We have plenty of internal customers who want stable VT support, and JDK 25 looks the release that unbreaks some of the stuff for them, e.g. the no longer pinning on synchronized. Our experience also shows that if we have seen VT issues in JDK-internal testing (like this one), they would definitely show up in some production environment at scale. It would be even sadder if VT production code that ran reasonably fine on JDK 21, would upgrade to JDK 25 to cover the significant part of VT corner cases, and then run into new regressions in JDK 25 the fix like this addresses. This is why we are even considering to pick it up to Corretto 25.0.3 ahead of upstream :)

But it is ultimately your call as maintainer, we can live with it either way.