OpenJ9DiagnosticsMXBean and other mxbean testing is not running

[pshipton]

#718 added OpenJ9DiagnosticsMXBean and a corresponding test. The test is not running and this should be fixed. This isn't the only mxbean test which isn't running, there are others as well: testOSMXBeanLocal, testOSMXBeanRemote, testJCMMXBeanLocal, testJCMMXBeanRemote. There are more but I think they have special machine requirements.

https://ci.eclipse.org/openj9/job/Test-Extended-JDK8-linux_ppc-64_cmprssptrs_le/134/consoleFull

02:38:29 ===============================================
02:38:29 Running test testOpenJ9DiagnosticsMXBean_0 ...
02:38:29 ===============================================
02:38:29 testOpenJ9DiagnosticsMXBean_0 Start Time: Thu Apr 19 06:38:16 2018 Epoch Time (ms): 1524119896667
02:38:29 Skipped due to jvm options ( -Xcompressedrefs -Xint -Xgcpolicy:optthruput -Xnocompressedrefs ) => testOpenJ9DiagnosticsMXBean_0_SKIPPED
02:38:29 testOpenJ9DiagnosticsMXBean_0 Finish Time: Thu Apr 19 06:38:16 2018 Epoch Time (ms): 1524119896679
02:38:29
02:38:29 ===============================================
02:38:29 Running test testOpenJ9DiagnosticsMXBean_1 ...
02:38:29 ===============================================
02:38:29 testOpenJ9DiagnosticsMXBean_1 Start Time: Thu Apr 19 06:38:16 2018 Epoch Time (ms): 1524119896700
02:38:29 Skipped due to jvm options ( -Xcompressedrefs -Xjit -Xgcpolicy:optthruput -Xnocompressedrefs ) => testOpenJ9DiagnosticsMXBean_1_SKIPPED
02:38:29 testOpenJ9DiagnosticsMXBean_1 Finish Time: Thu Apr 19 06:38:16 2018 Epoch Time (ms): 1524119896713

[smlambert]

The test was added to run in Mode100, Mode101, which do not apply to certain buildspecs: https://github.com/eclipse/openj9/blob/master/test/TestConfig/resources/ottawa.csv#L17 and L18 show this.

Not sure why those modes were selected (or any with -Xnocompressedrefs), but would suggest replacing the variation contents in the playlist <variation>-Xjit -Xgcpolicy:optthruput</variation> and <variation>-Xint -Xgcpolicy:optthruput</variation> in place of the Mode101, Mode100. I should have caught this in review. We should also search all existing variations for those 140 or so modes that are -Xnocompressedrefs, as they will be skipped since OpenJ9 is only compressedrefs spec (so far... #1669)

Rethink why we are choosing those modes to run in, rethink to use commandline options in the variation tag instead of short-form Modexxx options, as Modexxxx is legacy from old world test infrastructure that we hope to move away from.

[pshipton]

Using NoOptions (i.e. default options) would be fine, for OpenJ9DiagnosticsMXBean and the others (testOSMXBeanLocal, testOSMXBeanRemote, testJCMMXBeanLocal, testJCMMXBeanRemote) if they have the same issue. The same I think for most any test unless its a GC specific test. I think using these particular modes is just historical, i.e. it made sense long long ago.

If/when we fix the modes for OpenJ9DiagnosticsMXBean it probably needs to be excluded because I don't think it works consistently.

[pshipton]

#1745 should help with the intermittent issues that occur running the OpenJ9DiagnosticsMXBean test.

[smlambert]

Looking for an update as to whether this issue been addressed satisfactorily now?

[pshipton]

Looking for an update as to whether this issue been addressed satisfactorily now?

I'm going to say no. OpenJ9DiagnosticsMXBean is still configured for Mode100, and 101. This apparently doesn't stop it from running on 32-bit JVMs, as we recently needed to exclude it there because the test hangs #2213.

The other MXBean testing mores aren't fixed yet either. I opened #1866 to do so but found an intermittent failure which I expect needs to be resolved before enabling the tests.