On Tue, Dec 8, 2020 at 3:54 AM Daniel Fuchs <dfu...@openjdk.java.net> wrote:
> > > RandomFactory is probably overkill here as the test just needs to > exercise untimed and timed get. So just a random boolean rather than a wide > range of random values. > > OK. > There's a conflict between the desires to do more thorough testing and avoid flaky intermittent failures. Especially in j.u.concurrent we embrace test non-determinism, much of it comes from the concurrency we're testing, and retrieving a random seed for reproducibility is not worth the effort. No one does a good job of race prevention through testing.
