JVMTI NotSuspended test fails with incorrect virtual thread stacktrace

[dipak-bagadiya]

Issue

The stacktrace of virtual thread is incorrect since it doesn't start/match the CONTINUATION_CLASS_NAME
and CONTINUATION_METHOD_NAME.

Test CMD

make test TEST="jtreg:test/hotspot/jtreg/serviceability/jvmti/stress/StackTrace/NotSuspended" JTREG="JAVA_OPTIONS=--enable-preview -Dvm.continuations=true;VERBOSE=all"

Test Output

STDOUT:
Agent_OnLoad started
Agent_OnLoad finished
Synchronization point checkStatus(0) called.
Data 0x7f2f740a4bd8 0x7f2f54002b18
Agent: wait for thread to start
Agent: started
Agent: Stacktrace of virtual thread is incorrect: doesn't start from enter(...):
JVMTI Stack Trace: frame count: 6
 0: java/lang/Object: waitImpl(JI)V
 1: java/lang/Object: wait(JI)V
 2: java/lang/Object: wait(J)V
 3: java/lang/Thread: join(J)V
 4: java/lang/Thread: join()V
 5: com/sun/javatest/regtest/agent/MainWrapper: main([Ljava/lang/String;)V

[babsingh]

@fengxue-IS I believe that this test fails because our implementation of Continuation does not match the RI. Do you agree? If so, do you want to match the RI or permanently exclude the test?

[fengxue-IS]

@fengxue-IS I believe that this test fails because our implementation of Continuation does not match the RI. Do you agree? If so, do you want to match the RI or permanently exclude the test?

This is not due to mismatch of the impl, looking at the test code, seem like we have returned a carrier/platform thread as result of GetVirtualThread call, this should be fixed in jvmtiGetVirtualThread code.

The test code seem to suggest that a JVMTI_ERROR_THREAD_NOT_ALIVE error should be returned when the platform thread passed into GetVirtualThread doesn't have a virtual thread mounted.

Once that have been fixed, we should retest to see if it is still looking at implementation specific details. (I suspect that will be the case as OpenJ9's continuation entry point is Continuation.execute(...) instead of the enter(...) expected by the test case.

[babsingh]

NotSuspended/GetStackTraceNotSuspendedStressTest has been enabled in adoptium/aqa-tests#4554.

The test has been fixed through the following PRs:

• Rename Continuation.execute to match the RI #16786;
• Fix jvmtiGetVirtualThread and make it more robust #16787; and
• Fix jvmtiGetStackTrace #17318.