Re: gcj & JMX

[Gary Benson <gbenson at redhat dot com>]

Andrew Haley wrote:
> Andrew John Hughes writes:
> > On Wed, 2007-02-14 at 18:14 +0000, Andrew Haley wrote:
> > > With Classpath, none of this ever happens.  Instead,
> > > gnu.javax.management.registerMBean(Object, ObjectName) does not
> > > call MBeanRegistration.preRegister() unless ObjectName == null.
> > > 
> > > So, the preRegister method is never called.
> > > 
> > > The spec for MBeanServer.registerMBean() says "If the object
> > > name given is null, the MBean must provide its own name by
> > > implementing the MBeanRegistration interface and returning the
> > > name from the preRegister method."
> > > 
> > > I suspect, however, that if the MBean is implements the
> > > MBeanRegistration interface we should call the preRegister
> > > method regardless.
> > > 
> > > I'm going to try it.
> > 
> > Sounds good to me; I hope this fixes the problem.  Maybe I
> > shouldn't stick to the specs and just go with common sense in
> > future... ;)
> 
> Maybe, although the spec doesn't actually say that regstration
> should only be done if there is no name, it just very strongly
> implies it.  :-)

My reading of the various specs is that there's an unspoken assumption
that MBeans _may_ implement the MBeanRegistration interface _if_ they
want to do things before and after registration and/or deregistration
time, and that the spec for MBeanServer.registerMBean() merely changes
the "may" to a "must" for MBeans without a name.

Ok if I commit something like the attached patch?

Cheers,
Gary

Attachment: patch
Description: Text document