currently, it's kind of fragile. It only triggers one unstable_if trap and it is not the test itself.

<statistics type='deoptimization'>
Deoptimization traps recorded:
1 (100.0%) total
unstable_if/reinterpret/if_icmpeq: 1 (100.0%)
</statistics>

In JDK-8287385, I manage to suppress the superficial unstable_if traps and then I can't pass this test. I would like to make sure the test itself triggers null-check deoptimization and is not subject to external environment.