Testcases about accessing volatile arrays

[yuanxiang (F)]

Hi All:

         Testcases in
                   org.openjdk.jcstress.tests.accessAtomic.arrays.small.volatiles.*
                   org.openjdk.jcstress.tests.accessAtomic.arrays.large.volatiles.*
         tests the atomicity of accessing to elements in volatile arrays.

         For example, o.o.j.t.accessAtomic.arrays.small.volatiles.TestLong.java

/********Code Begin******************/
@JCStressTest
@Outcome(id = "[0]", expect = Expect.ACCEPTABLE, desc = "Default value for the element. Allowed to see this: data race.")
@Outcome(id = "[-1]", expect = Expect.ACCEPTABLE, desc = "The value set by the actor thread. Observer sees the complete update.")
@Outcome(expect = Expect.FORBIDDEN, desc = "Other values are forbidden: atomicity violation.")
@State
public class LongTest {
    volatile long[] a = new long[1];

    @Actor
    public void actor1() {
        a[0] = -1L;
    }
    @Actor
    public void actor2(LongResult1 r) {
        r.r1 = a[0];
    }
}
/********Code End******************/


         In my opinion, if an array declared as:
                   volatile long[] arr = new long[xxxxx];
         The "volatile" makes sure that accessing to the array reference is atomicity, but not the elements of the array.
         To make sure atomicity of accessing to array elements, we should use class AtomicLongArray.

         Is there something wrong with these testcases?
         Or I make a mistake about these testcases...

         Thanks a lot!

         Xiang Yuan
         2016.6.23