JDK-8370887 DelayScheduler Bug Discussion

[Pavel Rappo]

Please use core-libs-dev to discuss this.

On Sat 15 Nov 2025 at 13:01, Kent Dong <ch3cho at qq.com> wrote:

> Hi guys,
>
> During our trial of Java 25, we found this bug JDK-8370887
> <https://bugs.openjdk.org/browse/JDK-8370887> DelayScheduler.replace
> method may break the 4-ary heap in certain scenarios. Since NIO uses
> virtual threads for poller by default in Linux, it may affect the timeout
> of IO operations. If it can be fix, we would be more confident to use Java
> 25 and VT in our production environment.
>
> After a full analysis, we found it is caused by a missing sift-up branch
> in DelayScheduler.replace method, and it's not quite hard to fix. And it
> stays silent in JBS for quite a while. Is there a schedule/plan for it to
> be fixed or maybe we can contribute a fix to it?
>
> Thanks.
>
> Kent Dong
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/jdk-dev/attachments/20251115/1720cb36/attachment.htm>