Also disable to run ParallelPrefix.java in concurrency.
Ok.
Please check the new webrev: http://cr.openjdk.java.net/~mli/8179242/webrev.01/
29 * @run testng/othervm -Xms256m -Xmx1024m ParallelPrefix
Why are you setting the max heap size?
In my test result, 1024M is sufficient, and if I don't set it, I'm not sure
what's the default value for it, so set it explicitly for this memory sensitive
test.
Please let me know if you think it's unnecessary.
Hi Paul,
It just feels a little odd. Given that failed test were reported for machines
with much more than 2g of memory. Implying there is something else going on,
perhaps perturbed by other changes of test infrastructure?
Yes. 2g is just a defensive check in case the test is running on an
system without large memory. Besides of this, there is another
protection: disable ParallelPrefix.java to run in concurrency, so the
test will not compete with other tests for memory. If it still fails in
the future, we can remove large size array test unconditionally or check
getFreePhysicalMemorySize() at runtime.
I think what you have is ok.
Thank you for review, I will push the code.
Thank you
-Hamlin
Paul.
