J9vmTest j9vm.test.jni.CriticalRegionTest testAcquireAndGC_Sleepers8_GCers8: not all threads completed before the timeout

[pshipton]

https://openj9-jenkins.osuosl.org/job/Test_openjdk17_j9_sanity.functional_s390x_linux_Nightly_testList_1/556 - ub20-390-1
J9vmTest_4

+++ j9vm.test.jni.CriticalRegionTest: +++
command: /home/jenkins/workspace/Test_openjdk17_j9_sanity.functional_s390x_linux_Nightly_testList_1/openjdkbinary/j2sdk-image/bin/java  -XX:+UseCompressedOops -Xjit -Xgcpolicy:balanced  -Xdump -Xms64m -Xmx64m    -classpath /home/jenkins/workspace/Test_openjdk17_j9_sanity.functional_s390x_linux_Nightly_testList_1/aqa-tests/TKG/../../jvmtest/functional/VM_Test/VM_Test.jar:/home/jenkins/workspace/Test_openjdk17_j9_sanity.functional_s390x_linux_Nightly_testList_1/aqa-tests/TKG/../TKG/lib/asm-all.jar  j9vm.test.jni.CriticalRegionTest 

Exception in thread "main" java.lang.RuntimeException: testAcquireAndGC_Sleepers8_GCers8: not all threads completed before the timeout (15/16)
	at j9vm.test.jni.CriticalRegionTest.reportError(CriticalRegionTest.java:154)
	at j9vm.test.jni.CriticalRegionTest.testAcquireAndGC(CriticalRegionTest.java:879)
	at j9vm.test.jni.CriticalRegionTest.runNonBlockingTests(CriticalRegionTest.java:745)
	at j9vm.test.jni.CriticalRegionTest.main(CriticalRegionTest.java:198)
no-zero exit value: 1
*** Test FAILED *** (j9vm.test.jni.CriticalRegionTest)

[pshipton]

https://openj9-jenkins.osuosl.org/job/Test_openjdk17_j9_sanity.functional_s390x_linux_Nightly_testList_0/557 - ub20-390-4
J9vmTest_5

+++ j9vm.test.jni.CriticalRegionTest: +++
command: /home/jenkins/workspace/Test_openjdk17_j9_sanity.functional_s390x_linux_Nightly_testList_0/openjdkbinary/j2sdk-image/bin/java  -Xcompressedrefs -Xjit -Xgcpolicy:gencon  -Xdump -Xms64m -Xmx64m    -classpath /home/jenkins/workspace/Test_openjdk17_j9_sanity.functional_s390x_linux_Nightly_testList_0/aqa-tests/TKG/../../jvmtest/functional/VM_Test/VM_Test.jar:/home/jenkins/workspace/Test_openjdk17_j9_sanity.functional_s390x_linux_Nightly_testList_0/aqa-tests/TKG/../TKG/lib/asm-all.jar  j9vm.test.jni.CriticalRegionTest 

Exception in thread "main" java.lang.RuntimeException: testAcquireAndGC_Sleepers1_GCers8: not all threads completed before the timeout (8/9)
	at j9vm.test.jni.CriticalRegionTest.reportError(CriticalRegionTest.java:154)
	at j9vm.test.jni.CriticalRegionTest.testAcquireAndGC(CriticalRegionTest.java:879)
	at j9vm.test.jni.CriticalRegionTest.runNonBlockingTests(CriticalRegionTest.java:744)
	at j9vm.test.jni.CriticalRegionTest.main(CriticalRegionTest.java:198)
no-zero exit value: 1
*** Test FAILED *** (j9vm.test.jni.CriticalRegionTest)
```

[pshipton]

Changes from previous build to first failure. The change wouldn't cause this failure.
75701be...e57170e

[pshipton]

Probably related to #18044
@dmitripivkine @TobiAjila - seems we should reduce the number of GCs, or increase the timeout.

[dmitripivkine]

Probably related to #18044
@dmitripivkine @TobiAjila - seems we should reduce the number of GCs, or increase the timeout.

Test code: https://github.com/eclipse-openj9/openj9/blob/master/test/functional/VM_Test/src/j9vm/test/jni/CriticalRegionTest.java

Looks like the timeout has been increased in the past to 5 seconds. Seems test fails on one group of similar zLinux machines. It might be these machines are so slow that timeout should be increased again (to 10 seconds?).

[pshipton]

We can first try removing one of the two System.gc() calls and see if that resolves it.

https://github.com/eclipse-openj9/openj9/blob/master/test/functional/VM_Test/src/j9vm/test/jni/CriticalRegionTest.java#L143

[dmitripivkine]

We can first try removing one of the two System.gc() calls and see if that resolves it.

https://github.com/eclipse-openj9/openj9/blob/master/test/functional/VM_Test/src/j9vm/test/jni/CriticalRegionTest.java#L143

Hm, It might improve timing but can it create logical problem if SystemGC can not produce results as expected (due behaviour first call can just complete GC Cycle and left floating garbage, unprocessed References etc. behind, so it is the reason we call SystemGC twice)? Can this test be impacted?

[pshipton]

But System.gc() does two calls internally now, since #18044. You think we still need to explicitly call it twice?

[dmitripivkine]

But System.gc() does two calls internally now, since #18044. You think we still need to explicitly call it twice?

Right. Sorry, I forgot, you are correct.

[pshipton]

I looked more closely and removing a call to System.gc() won't help with the timeout since it's in another test that we haven't run yet when the timeout occurs. Still I've removed it and increased the timeout in https://github.com/eclipse-openj9/openj9/compare/master...pshipton:gconce?expand=1, running a grinder to confirm and then will create a PR for review.