[13] RFR (T) 8226394: [TESTBUG] vmTestbase/metaspace/flags/maxMetaspaceSize/TestDescription.java fails with java.lang.NoClassDefFoundError

coleen.phillimore at oracle.com coleen.phillimore at oracle.com
Fri Jun 21 14:23:21 UTC 2019



On 6/21/19 10:00 AM, David Holmes wrote:
> On 21/06/2019 5:37 am, coleen.phillimore at oracle.com wrote:
>> Summary: don't use printStackTrace to verify OOM type.
>
> Can you update the bug to show that you decided to workaround the 
> issue in the test rather than forcing pre-initialization of the class 
> library code please.

Yes, I can do that.  I still think it's a potential problem, but only 
this test has noticed it.
>
>> This avoids StackTraceElement needing to load HashedModules type when 
>> out of metaspace OOM.  Tested 100x without failure with config that 
>> averages 30% failure rate.
>>
>> open webrev at 
>> http://cr.openjdk.java.net/~coleenp/2019/8226394.01/webrev
>> bug link https://bugs.openjdk.java.net/browse/JDK-8226394
>
> Fix seems quite reasonable - unsure why it was thought the complete 
> stack was needed in the first place ...

Yes, agree.
Thanks!
Coleen
>
> Thanks,
> David
>
>> Thanks,
>> Coleen



More information about the hotspot-runtime-dev mailing list