…terleaving threadyield() tests fail on both nqp-m and nqp-j, but this may be a case of intuitive semantics not matching true semantics.
…anks psch++ for spotting this.
…ed semaphore tests. Works fine on both nqp-m and nqp-j.
… Works fine on nqp-m and nqp-j, though the lock effectiveness test is a bit slow on nqp-j, and may be too fast on nqp-m.
…s on nqp-m (works fine on nqp-j), so not adding to standard test list yet.
That means that we have openpipe on all platforms/backends except for parrot on windows.
At least, that is the point we can be sure about it.