RFR (XS) 8153277: [TESTBUG] gc/arguments/TestMaxMinHeapFreeRatioFlags is too sensitive for stray allocations in verifyRatio

Dmitry Fazunenko dmitry.fazunenko at oracle.com
Fri Apr 1 16:07:10 UTC 2016


Still looks good :)
-- Dima

On 01.04.2016 18:58, Aleksey Shipilev wrote:
> Thanks Dmitry!
>
> I realized that my hs-rt workspace was older than I thought. Rebased the
> patch, and verified it still fixes the trouble:
>    http://cr.openjdk.java.net/~shade/8153277/webrev.01/
>
> Thanks,
> -Aleksey
>
> On 04/01/2016 05:56 PM, Dmitry Fazunenko wrote:
>> Hi Aleksey,
>>
>> The change looks good.
>> Pls, fix the copyright before push.
>>
>> Thanks,
>> Dima
>>
>> On 01.04.2016 16:59, Aleksey Shipilev wrote:
>>> Hi,
>>>
>>> Please review a preemptive fix in TestMaxMinHeapFreeRatioFlags test.
>>> This failure only reproduces when you switch the JEP 280 concat strategy
>>> to something more advanced than it is now.
>>>
>>> The test would fail if we silently bring more objects in the heap in the
>>> middle of the test. Notably, this will happen inside of verifyRatio
>>> helper method which does a String concat -- on first invocation it will
>>> lazily initialize j.l.invoke infra. We need to prime the utility methods
>>> before using them within the test.
>>>
>>> Bug:
>>>     https://bugs.openjdk.java.net/browse/JDK-8153277
>>>
>>> Webrev:
>>>     http://cr.openjdk.java.net/~shade/8153277/webrev.00/
>>>
>>> Testing: offending test; going to run JPRT -testset hotspot during a
>>> push.
>>>
>>> Thanks,
>>> -Aleksey
>>>
>



More information about the hotspot-gc-dev mailing list