Is JCStress sample test Classic_01_DiningPhilosophers.Arbitrator correct?

[Serguei Katkov]

I'm trying to figure out what guarantees that release forks (setting false
to elements of forks array) is visible for other threads?

Say p1 implements the last iteration in stride, sets true to forks[0] &
forks[1], release waiter monitor (since this moment p2, p3 after taking
waiter monitor must see an update of forks to true) leaves infinite loop
and restore elements to false without any synchronization and goes in
framework for sync with other threads using access to other volatile
variable.

So why must p2 and p3 see false value in forks[0] and forks [1]?
Without that p2 and p3 fall into an infinite loop.

Thanks in advance,
Serguei.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/jcstress-dev/attachments/20220719/74b7946c/attachment-0001.htm>