Re: RFR: 8200195 : serviceability/jvmti/FieldAccessWatch/FieldAccessWatch.java crashes with "assert(thread->thread_state() == _thread_in_native) failed: coming from wrong thread state"

[serguei.spit...@oracle.com]

Hi Alex,

The fix looks good.
Thank you for replacing the javaEnv with the jni_env.

Thanks,
Serguei



On 4/9/18 15:04, Alex Menkov wrote:

updated webrev: http://cr.openjdk.java.net/~amenkov/field_access_graal/webrev.02/

(javaEnv is replaced with jni_env)

On 04/09/2018 14:35, David Holmes wrote:

+1

Though I'd also like to understand why the test only fails sometimes and which other thread is involved?

It always fails with Graal compiler on win, linux and macosx.

Looks like with Graal compiler access/modification events are sent on several threads (sorry, I don't have enough knowledge about Graal).

--alex

Thanks,
David

On 10/04/2018 7:28 AM, Chris Plummer wrote:

Hi Alex,

I'd suggest renaming javaEnv to jni_env to be consistent. Not sure why javaEnv was chosen in the original implementation. Otherwise the changes look good.

thanks,

Chris

On 4/9/18 2:13 PM, Alex Menkov wrote:

Hi all,

Please review a fix for https://bugs.openjdk.java.net/browse/JDK-8200195

webrev: http://cr.openjdk.java.net/~amenkov/field_access_graal/webrev/

The problem with the test is it uses cached JNIEnv value instead using a value passed to the callbacks. JNIEnv is valid only for the current thread.

--alex