RFR (S) 8231769: Test tools/javac/tree/MakeTypeTest.java fails with -Xcheck:jni
coleen.phillimore at oracle.com
coleen.phillimore at oracle.com
Fri Oct 11 12:01:53 UTC 2019
Trivial change to the patch but still assert for JavaThread before the
cast. Harold pointed out to me that I didn't need to move this
JavaThread assignment.
http://cr.openjdk.java.net/~coleenp/2019/8231769.02/webrev/index.html
Thanks,
Coleen
On 10/11/19 7:53 AM, coleen.phillimore at oracle.com wrote:
>
>
> On 10/10/19 8:33 PM, David Holmes wrote:
>> Hi Coleen,
>>
>> That seems okay to me. I must admit it is not always apparent to me
>> when we need to explicitly call destroy_local.
>
> Yes, I agree. Looking through lots of native code in the JVM
> yesterday, this seems to be convention. And the other
> create_from_platform_dependent_str() did it too.
>>
>> Minor pre-existing nit - can you reverse these two lines please:
>>
>> 386 JavaThread* thread = (JavaThread*)THREAD;
>> 387 assert(thread->is_Java_thread(), "must be java thread");
>>
>
> Done. Thanks for reviewing this!
> Coleen
>
>> Thanks,
>> David
>>
>> On 11/10/2019 9:54 am, coleen.phillimore at oracle.com wrote:
>>> Summary: Delete local jni handles in
>>> create_from_platform_dependent_str() after upcall to Java.
>>>
>>> The JVM_GetProperties function was creating many local refs which
>>> are never cleaned up, hence overflowing the EnsureLocalCapacity call.
>>>
>>> Tested with tier1 -Xcheck:jni locally and tier2-3.
>>>
>>> open webrev at
>>> http://cr.openjdk.java.net/~coleenp/2019/8231769.01/webrev
>>> bug link https://bugs.openjdk.java.net/browse/JDK-8231769
>>>
>>> Thanks,
>>> Coleen
>
More information about the hotspot-runtime-dev
mailing list