[8u-dev] Request for enhancement backport approval for CR 8189170 - Add option to disable stack overflow checking in primordial thread for use with JNI_CreateJavaJVM

[Rob McKenna]

Thanks Buck,

I'll get back to you on this once a decision has been reached.

    -Rob

On 21/01/18 20:26, David Buck wrote:
> Hi!
> 
> Please consider the following enhancement for approval:
> 
> bug report:
> https://bugs.openjdk.java.net/browse/JDK-8189170
> 
> This is an important fix that helps improve our interoperability with the
> language runtime for R and possibly other languages as well. There are
> likely other JNI use cases that also benefit from this as well. One could
> even argue that the current behavior is a bug as it is not clear at all why
> the JVM would be justified in assuming that it completely "owns" the thread
> that calls JNI_CreateJavaVM. Either way, this is an important and safe fix.
> 
> All standard JPRT tests run and passed (hotspot testset). Effectiveness of
> change manually confirmed with dummy JNI invocation test case. The change
> itself is relatively small and very straightforward. It has zero impact on
> JVM processes launched by our own launcher, and will only impact JNI
> programs that use the Invocation API if both the UnlockExperimentalVMOptions
> and DisablePrimordialThreadGuardPages options are passed to the JVM. I
> believe any risk posed by including this in JDK 8 is negligible.
> 
> Cheers,
> -Buck