Request for review: 6597112: Referential integrity loophole during remote object export

Alan Bateman Alan.Bateman at
Wed Apr 13 11:08:48 UTC 2011

Peter Jones wrote:
> :
> This fix looks good to me.  The synchronization subtlety may be worth a brief comment, to explain that you don't want the Reaper processing to occur in between the null guard and the put/increment actions.
> Cheers,
> -- Peter
Thanks Peter, thanks Neil, looks like we are almost done with this one.

Neil - I did a hg import --no-commit of the last change-set that you 
attached, ran the RMI tests and all seems okay. I noticed a couple of 
minor things:

1. You had "if (null != target.getImpl())" so changed it to "if 
(target.getImpl() != null)" to make is consistent
2. The run method in the test had inconsistent formatting
3. I "moved" the test to 
test/java/rmi/server/UnicastRemoteObject/exportObject so that additional 
tests could be added in the future

The updated webrev is here:

If you are okay with this then I can push this later today.


More information about the core-libs-dev mailing list