RFR: 8178813: Add test for G1 pre-barrier on dereference of weak JNI handles

[Kim Barrett]

> On May 5, 2017, at 7:58 AM, Mikael Gerdin <mikael.gerdin at oracle.com> wrote:
> 
> Hi Kim,
> 
> On 2017-05-05 07:45, Kim Barrett wrote:
>> Still looking for reviewers.
>> 
>>> On Apr 14, 2017, at 4:41 PM, Kim Barrett <kim.barrett at oracle.com> wrote:
>>> 
>>> Please review this new test of the handling of weak JNI handles during
>>> G1 concurrent mark.
>>> 
>>> CR:
>>> https://bugs.openjdk.java.net/browse/JDK-8178813
>>> 
>>> Webrev:
>>> http://cr.openjdk.java.net/~kbarrett/8178813/hotspot.00/
> 
> Test seems ok to me, I might have made the "static boolean resolve" an instance member and created two different instances of the test class because I dislike that kind of state in static members but I'm not going to block the change because of it.

I think you are right, and I’m going to change it along those lines.  New webrev will be forthcoming.

> Reviewed.
> /Mikael
> 
>>> 
>>> Testing:
>>> JPRT
>>> 
>>> Locally ran new test with G1-specific code in JNIHandles::resolve and
>>> MacroAssembler::resolve_jobject temporarily removed, and verified
>>> expected failures.