RFR(xs): 8161552: Test issue: VM init failed: GC triggered before VM initialization completed. Try increasing NewSize, current value 768K.

Jon Masamitsu jon.masamitsu at oracle.com
Wed Jul 20 22:40:48 UTC 2016



On 07/20/2016 03:16 PM, Dmitry Fazunenko wrote:
> Jon,
>
> On 20.07.2016 23:56, Jon Masamitsu wrote:
>> Changes look good.
>
> thanks for review.
>
>>
>> You also removed the 3m maximum heap size test as well
>> as the 2m one.  Was that also failing?
>
> No, the test hasn't failed yet with the 3m limit. But 3m doesn't look 
> like a right minimal max heap value.

Ok.  I think moving it to 4m is a reasonable thing.

Jon

> Because there is no formal requirement for the minimal heap size 
> required for Hotspot, I decided to round it to 4m.
>
> I understand the current tests are still fragile, and we need to 
> revise how to test the VM behavior on  small heaps. So, this fix is an 
> attempt to preserve the existing tests via adapting them to the new 
> reality.
>
> Thanks,
> Dima
>
>>
>> Jon
>>
>> On 07/20/2016 10:01 AM, Dmitry Fazunenko wrote:
>>> Hello,
>>>
>>> I'm looking for a couple of reviewers for a simple fix of two tests.
>>> The tests check some basic assertions when VM is started with max 
>>> heap size set to some small values.
>>> The tests work almost everywhere, but there are cases when with 
>>> -Xmx2M GC triggered before VM initialization completed.
>>> 2M is valid but extremely small max heap size. Cases of usages such 
>>> small heap are rather theoretical than practical.
>>>
>>> To fix the problem I made the tests weaker, the cases setting max 
>>> heap to 2M are removed.
>>>
>>> http://cr.openjdk.java.net/~dfazunen/8161552/webrev.00/
>>> https://bugs.openjdk.java.net/browse/JDK-8161552
>>>
>>> Testing: locally + RBT
>>>
>>> Thanks,
>>> Dima
>>
>




More information about the hotspot-gc-dev mailing list