RFR: 8168396: Unexpected OOME in GcCauseTest02 and GcTest02
michail.chernov at oracle.com
Mon Oct 31 12:44:46 UTC 2016
Thank you very much for reviewing this!
Could someone of reviewers please take a look at this change. This issue
is integration blocker.
On 27/10/2016 19:05, Dmitry Fazunenko wrote:
> Thanks for addressing my comments. New version looks good to me.
> On 27/10/2016 16:23, Michail Chernov wrote:
>> I'd like to update the webrev after offline discussion with Dmitry
>> Fazunenko. Classes structure was simplified, removed unnecessary
>> Tests GcCauseTest02 and GcTest02 were updated - old version can throw
>> OOME. The reason of OOME is that tests do not consider available free
>> heap and try to fill 70% of all heap. This change makes heap size
>> fixed ( InitialHeapSize == MaxHeapSize ). New version tries to make
>> heap usage more than 70% and invokes System.gc() to update
>> performance counters/pools statistics.
>> On 26/10/2016 16:02, Michail Chernov wrote:
>>> Could you please review this change?
>>> Initial heap size and max heap size were made equal to guarantee
>>> expected memory allocation. Previous version of test can throw OOME
>>> because does not consider existing free heap size.
More information about the hotspot-gc-dev