FW: is this a bug ?

[Yangfei (Felix)]

> >
> > But have you actually see it fail? I don't see how, as test grading
> > accepts all results.
> 
> Yes, it fails randomly on our 128-core aarch64 server platform.
> 
> Part of the report was:
> 
> ERROR tests
> Tests break for some reason, other than failing the assert. Correct
> implementations should have none.
> 
> 
> o.o.j.t.atomicity.buffers.DirectByteBufferViewsAtomicityTests.LongViewTest
> 
> Exception in thread "main" java.lang.IllegalStateException: Should have been
> handled within the Runner
> 	at org.openjdk.jcstress.JCStress.runEmbedded(JCStress.java:248)
> 	at org.openjdk.jcstress.ForkedMain.main(ForkedMain.java:55)
> JVM options: [-XX:+UnlockDiagnosticVMOptions, -XX:+WhiteBoxAPI,
> -XX:-RestrictContended, -XX:-TieredCompilation,
> -XX:ReservedCodeCacheSize=128M, -server] Iterations: 5 Time: 1000

Some of our analysis:

 51         private final LongBuffer lb;
 52         private final ShortBuffer sb;
 53
 54         public MyState() {
 55             b = ByteBuffer.allocate(16);
 56             b.order(ByteOrder.nativeOrder());
 57             ib = b.asIntBuffer();
 58             cb = b.asCharBuffer();
 59             db = b.asDoubleBuffer();
 60             fb = b.asFloatBuffer();
 61             lb = b.asLongBuffer();
 62             sb = b.asShortBuffer();
 63         }

94     @JCStressTest
 95     @JCStressMeta(GradeInt.class)
 96     public static class LongViewTest {
 97         @Actor public void actor1(MyState s)                { s.lb.put(0, -1);                                  }
 98         @Actor public void actor2(MyState s, J_Result r) { r.r1 = s.lb.get(0);                               }
 99     }

For the LongBuffer.put & LongBuffer.get operation, the atomicity is different when b is assigned a DirectByteBuffer.  Please check.