Use the "default" directory to access J9's libjvm

[babsingh]

Workaround for eclipse-openj9/openj9#14397.

Signed-off-by: Babneet Singh sbabneet@ca.ibm.com

[babsingh]

I was debating whether to just do minimal changes for exeCallerAccessTest/CallerAccessTest to pass on OSX or use default/libjvm.* with OpenJ9 everywhere. I chose the latter in order to avoid similar issues seen in eclipse-openj9/openj9#14397 with other native tests.

Running personal build to verify these changes: https://hyc-runtimes-jenkins.swg-devops.com/view/OpenJ9%20-%20Personal/job/Pipeline-Build-Test-Personal/11780

• Platforms: x86-64_linux,x86-64_windows,ppc64_aix,x86-64_mac
• Test targets: sanity.functional,sanity.openjdk

Will remove this PR from the draft state if no failures are seen in the personal build.

fyi @pshipton @tajila @keithc-ca

[pshipton]

The problem with using default/libjvm everywhere is that it might hide problems. Users do use, and are expected to use the redirector (from server or j9vm) and not the VM's libjvm directly. We used to have two or more VMs packaged (one for compressed, one non-compressed, one for realtime long ago). We haven't yet decided to do away with the redirector, so we expect it to work.

[babsingh]

re #15 (comment):

in this case, i will update the PR with minimal changes for exeCallerAccessTest/CallerAccessTest to pass on OSX.

[keithc-ca]

I would hope that 'minimal' change is no change here. As Peter said, things should work with the redirector involved (as is the normal use case).

[pshipton]

True, the better approach is to spend time on figuring out what the problem with the redirector or test is, rather than on the workaround.

[babsingh]

Closing. Will exclude the failing test on OSX while the real fix is being developed.