RFR: 7093328: JVMTI: jvmtiPrimitiveFieldCallback always report 0's for static primitives

Daniel D. Daugherty daniel.daugherty at oracle.com
Wed Aug 29 07:16:31 PDT 2012

Logistics issue: This review request went out to OpenJDK alias, but
I think this webrev is not accessible outside of Oracle. The webrev
should be publicly accessible when the review is public. However,
in this case, this is a one line fix so IMHO you could get away with
a context diff in the suggested fix of the bug report.

On 8/29/12 1:24 AM, Rickard Bäckman wrote:
> Hi all,
> can I get a couple of reviews for the following bug fix:
> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7093328
> webrev: http://rbackman.se.oracle.com/~rbackman/7093328/
> Thanks
> /R

Thumbs up.

     No content comments.

     This is a perfect example of how casting can bite you. :-)

     The buggy line:

     1165     address addr = (address)k + offset;

     dates back to the original implementation of this function:

     D 1.47 05/08/22 20:30:26 ab23780 77 76  01476/01330/01852
     6195957: further clean-up and caching of field maps

     In the buggy code, a value is copied from the instanceKlass
     at the offset instead of from the Java mirror. If the value
     copied from the instanceKlass happened to match the expected
     value, then that was pure luck.

     The bug report claims that this last "worked" in 6u26. I suspect
     that "working" is defined as returning a non-zero value.


More information about the hotspot-runtime-dev mailing list