JDK 9 RFR of JDK-8132548: java/lang/ThreadGroup/Stop.java fails with "RuntimeException: Failure"

[Martin Buchholz]

This message is only the reviewers having friendly discussion with each
other!

On Mon, Jul 11, 2016 at 11:44 PM, David Holmes <david.holmes at oracle.com>
wrote:

>     private static final AtomicReference<Throwable> firstThrowable
>>         = new AtomicReference<>();
>>     private static final AtomicReference<Throwable> secondThrowable
>>         = new AtomicReference<>();
>>
>
> Simple volatile Throwable fields would suffice - AtomicReferences are
> overkill.
>
>
But in my later attempt they became locals, which is better style when
possible.


>     private static final Thread second = new Thread(group, () -> {
>>         ready.countDown();
>>         try {
>>
>
> You have to place the countDown() inside the try block else the
> ThreadDeath may escape.
>
>
Oh good point - there's a race window!
(everybody writes racy code)


>             do {} while (true);
>>
>
> Busy-loops are "bad".
>

Occasionally necessary, but yes, should be used only as last resort.  But
here it's not clear which is more wasteful of cpu time...


> I agree this meets the "spec" of checking "did group.stop() cause all
> group members to encounter a ThreadDeath exception". But in relation to
> fixing the original timing problem, well a simple untimed-join() fixed that.
>
>
I like to throw in extra assertions when I've gone to the trouble of
setting up a test scenario.