RFR: Samples for race condition

[Vladimir Sitnikov]

On Wed, 30 Jun 2021 21:18:47 GMT, Michael Mirwaldt <github.com+6693355+mmirwaldt at openjdk.org> wrote:

>> jcstress-samples/src/main/java/org/openjdk/jcstress/samples/high/race/condition/RaceCondition_02_CheckThenReactSequence.java line 49:
>> 
>>> 47: @JCStressTest
>>> 48: @Outcome(id = {"true, false", "false, true"}, expect = ACCEPTABLE, desc = "Only one actor got true for the flag in its if-clause")
>>> 49: @Outcome(id = {"true, true"}, expect = FORBIDDEN, desc = "Both actors got true for the flag in their if-clauses")
>> 
>> Could you clarify why do you put `forbidden` here?
>> 
>> The implementation does allow `true, true` outcome since it does not synchronize actors.
>
> Yes, I can. I considered the outcome "true, true" as the undesirable result of the check-then-react-sequence.
> I regard all those race conditions as a kind of "bug patterns" for concurrent programs.
> What do you think and what do you propose?

If you consider JVM, then the outcome is either allowed by the spec or it is forbidden by the spec. I think the outcomes should be marked from the perspective of the specification.

The outcome true-true is allowed by JVM here, so I expect it to be marked as acceptable or acceptable_interesting.

-------------

PR: https://git.openjdk.java.net/jcstress/pull/88