On Tue, 27 Oct 2026 06:25:11 GMT, Alan Bateman <[email protected]> wrote:
>> Thanks @viktorklang-ora !
>>
>> I've also added a regression test that managed to timeout close to 100%
>> without the fix. Would appreciate some eyes on it
>
>> @macarte A stress test running all jdk_concurrent tests ~1000 times passed
>> successfully.
>
> @viktorklang-ora Are you going to do any testing where the waiter field is
> changed to volatile? Do we have enough benchmarks to measure that?
@AlanBateman I suspect it to be noise for most-if-not-all benchmarks. Looking
at the usages of `waiter` in LinkedTransferQueue it seems highly plausible that
we could make the field `volatile`. The only question I have is what we do with:
waiter = w; // ensure ordering
VarHandle.fullFence();
@DougLea Any thoughts about making `waiter` `volatile`? 🤔
-------------
PR Comment: https://git.openjdk.org/jdk/pull/33551#issuecomment-6439108506
