cc'ing core-libs-dev as they own this test. On 30/06/2015 10:54 PM, Stefan Särne wrote:
Hi all, Please review and sponsor this fix. Bug: https://bugs.openjdk.java.net/browse/JDK-6896810 <https://bugs.openjdk.java.net/browse/JDK-8073718>Webrev: http://cr.openjdk.java.net/~sjohanss/sponsor/ssarne/JDK-6896810/webrev.jdk.00/ <http://cr.openjdk.java.net/%7Esla/stefan/8073718/webrev.00/> The test fails since it run out of memory (again) when it tries to print that the expected OOME is received. Additional allocations are done during printing and the next OOME is thrown out of the test. By freeing the "chain" that hogs the memory on the heap, the test continues and the additional checks of checking that the blocks are empty are performed and pass.
Seems a reasonable approach to free up the memory - or at least allow a GC to reclaim it.
David -----
Tested: Run with all GC combinations. The test is part of tier1 for the jdk and this blocks integration from jdk9/hs to jdk9/dev and I therefore suggest it is integrated directly to jdk9/hs. Thanks, Stefan