TCK tests for virtual threads

[KyleAure]

These are tests for virtual=true assuming option 2 will win the vote (which is currently in the lead).

When testing on Java 17:

• Deploys an application to that contains annotations and deployment descriptors with virtual=false and virtual=true configurations
• Asserts that when virtual=false a platform thread is returned for ManagedExecutorService, ManagedExecutorScheduledService, ManagedThreadFactory, ForkJoinPool.
• Asserts that when virtual=true a platform thread is returned for ManagedExecutorService, ManagedExecutorScheduledService, ManagedThreadFactory, ForkJoinPool.

When testing on Java 21:

• Deploys an application to that contains annotations and deployment descriptors with virtual=false and virtual=true configurations.
• Asserts that when virtual=false a platform thread is returned for ManagedExecutorService, ManagedExecutorScheduledService, ManagedThreadFactory, ForkJoinPool.
• Asserts that when virtual=true a platform thread is returned for a ForkJoinPool backed by a ManagedThreadFactory as defined by Java.
• Assumes that when virtual=true a virtual thread is returned for ManagedExecutorService, ManagedExecutorScheduledService, ManagedThreadFactory
  • If a platform thread is returned, the test passes and we make no assertions about virtual thread behavior
  • If a virtual thread is returned, the test then makes assertions about virtual thread behavior

[njr-11]

This looks great Kyle! Thanks for writing this and demonstrating this approach to cover virtual=true in the TCK in a way that can be skipped on Java SE 17. I added a few minor comments - mostly on a typo of the word virtual that ended up copied to several places.

[njr-11]

Looks good! I noted one very minor type with two period characters (which might even be my fault if I mistyped my prior suggestion on that one). If you don't want to correct it, don't worry about it.

[Emily-Jiang]

I don't think you should skip the testPlatformThreadFactory for Java 17. It should work same as Java 21.

[njr-11]

I don't think you should skip the testPlatformThreadFactory for Java 17. It should work same as Java 21.

That will depend on the outcome of the vote that has yet to happen over what the behavior of virtual=true on Java SE 17 is. If vendor-specific wins, then we have no choice but to skip it because a vendor could choose to fail deployment. Kyle wrote this pull to investigate whether we would even be able to handle such an outcome, and did a very good job of figuring out a way. This pull will remain in pending state until a vote occurs and then be updated accordingly.

[Emily-Jiang]

I don't think you should fail the deployment when setting virutal=false on Java 17 as it is a valid scenario.
Even setting virutal=true on Java 17 might not fail deployment as the code would only be invoked during run time not something you can detect at the application startup stage.

[KyleAure]

@njr-11 I think this needs a re-review now that it has been updated to comply with the results of the vote.

[njr-11]

Thanks for updating this. I did spot some questionable assertions that are technically not guaranteed by the spec, although in most cases seem unlikely not to be the case.

[KyleAure]

@njr-11 I have handled all of the review comments you left. Please re-review when you get a chance.

[njr-11]

Added suggestions for very minor corrections. Otherwise looks good. Thanks for writing these!

[KyleAure]

This PR has been opened for a while and there have been no other comments/reviews. Now that all the comments have been resolved I'll go ahead and merge this.