RFR: 8042789: org.omg.CORBA.ORBSingletonClass loading no longer uses context class loader
Changes in the ORB loading in 8u5 causes incompatibility changes for some environments where a 3rd party plugin might be deployed over plugin. Some CORBA deployment installations have an incorrect expectation on finding the custom ORBSingleton class using the context classloader. Proper approach is to expect to find such classes on the system classloader. For JDK 9, the 8u5(new) behaviour will remain. For JDK 8u and older update families, I'm proposing we revert back to the old behaviour to aid compatibility in current deployment scenarios. webrev : http://cr.openjdk.java.net/~coffeys/webrev.8042789/webrev/ bug : https://bugs.openjdk.java.net/browse/JDK-8042789 I'm working a testcase for this also. regards, Sean.
Hi Sean I think this makes sense and the change looks fine Best Lance Lance Andersen| Principal Member of Technical Staff | +1.781.442.2037 Oracle Java Engineering 1 Network Drive Burlington, MA 01803 Lance.Andersen@oracle.com Sent from my iPad On Jun 6, 2014, at 1:22 PM, Seán Coffey <sean.coffey@oracle.com> wrote:
Changes in the ORB loading in 8u5 causes incompatibility changes for some environments where a 3rd party plugin might be deployed over plugin. Some CORBA deployment installations have an incorrect expectation on finding the custom ORBSingleton class using the context classloader. Proper approach is to expect to find such classes on the system classloader. For JDK 9, the 8u5(new) behaviour will remain.
For JDK 8u and older update families, I'm proposing we revert back to the old behaviour to aid compatibility in current deployment scenarios. webrev : http://cr.openjdk.java.net/~coffeys/webrev.8042789/webrev/ bug : https://bugs.openjdk.java.net/browse/JDK-8042789
I'm working a testcase for this also.
regards, Sean.
On 06/06/2014 18:22, Seán Coffey wrote:
Changes in the ORB loading in 8u5 causes incompatibility changes for some environments where a 3rd party plugin might be deployed over plugin. Some CORBA deployment installations have an incorrect expectation on finding the custom ORBSingleton class using the context classloader. Proper approach is to expect to find such classes on the system classloader. For JDK 9, the 8u5(new) behaviour will remain.
Right, it doesn't make sense to use the TCCL to locate the system-wide/singleton ORB and then cache it but clearly some applications that were bundle their own ORB with conflicting configuration are running into this. The change in the webrev looks okay to me (for 8u, not 9 where I agree that we keep the correct behavior). -Alan
participants (3)
-
Alan Bateman
-
Lance ＠ Oracle
-
Seán Coffey