RFR: 8332923: ObjectMonitorUsage.java failed with unexpected waiter_count [v3]

[Alan Bateman]

On Thu, 30 May 2024 06:51:54 GMT, David Holmes <dholmes at openjdk.org> wrote:

>> Hopefully the ports will catch up someday and the alternative implementation can be removed. 
>> 
>> We decided not to rename java.lang.VirtualThread when introducing the alternative implementation as it's just too disruptive. The super class that both implementations extend is BaseVirtualThread so testing for an instance of that is correct for the two implementations.
>> 
>> If it helps the readability then introducing a function to test if a thread is a virtual thread might help. It could use VMContinuations if needed but right now, testing for an instanceof BaseVirtualThread is okay.
>
> Okay. I still think that should be hidden behind the `java_lang_VirtualThread::is_instance` as it is an implementation detail the JVMTI and thread code shouldn't need to know about IMO. Once the alternative implementation is removed I expect these explicit checks for `BaseVirtualThread` will need to be reverted and we could avoid that if we make a change now.

java_lang_VirtualThread::is_instance returning true when the top is not an instance of that class would be a bit strange. java_lang_Thread::is_virtual_thread_instance might be less surprising.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/19405#discussion_r1620105401