RFR(xs): 8199656: Make slow metaspace verifications switchable in debug builds

Thomas Stüfe thomas.stuefe at gmail.com
Thu Mar 15 10:06:06 UTC 2018


Hi David,

On Thu, Mar 15, 2018 at 10:31 AM, David Holmes <david.holmes at oracle.com>
wrote:

> Hi Thomas,
>
> This seems fine to me.
>
> As it is a develop flag no CSR request is needed. (Under rules we recently
> clarified with CSR group, but which have yet to be documented somewhere on
> the wiki.)
>
> Thanks,
> David
>
>
Thanks for the review! Out of curiosity, would a diagnostic switch need a
CSR?

Thomas


>
> On 15/03/2018 5:09 PM, Thomas Stüfe wrote:
>
>> Hi all,
>>
>> may I please have reviews for this small rfe:
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8199656
>> webrev:
>> http://cr.openjdk.java.net/~stuefe/webrevs/8199656-verifymet
>> aspace-switch/webrev.00/webrev/
>>
>> We have a number of useful metaspace verifications which are all switched
>> off via a const bool in code. This change makes these verifications
>> switchable.
>>
>> Note: the switch is develop because all the verifications are ASSERT only,
>> so it makes no sense to make it the switch diagnostic. If we want the
>> tests
>> in product builds too, we need to revise the test code.
>>
>> For the effect this has,
>> try ./hotspot/variant-server/libjvm/gtest/gtestLauncher.exe
>> -XX:-VerifyMetaspace -jdk:./images/jdk/
>> --gtest_filter=MetaspaceAllocationTest*. With verifications, runtime
>> almost
>> triples in fastdebug.
>>
>> Thanks and Kind Regards, Thomas
>>
>>


More information about the hotspot-runtime-dev mailing list