Neil Richards wrote:
Bug 6597112 [1] reports that the implementation of
java.rmi.server.UnicastRemoteObject.exportObject(Remote) has a window
in which the Remote object given to the method is only weakly
referenced by the implementation.
Due to this window, an unfortunately timed GC cycle can free up the
Remote object, causing an exception to be thrown at a later point
(still) within the exportObject() implementation.
I don't know the RMI code so this isn't a formal review, just a
comment/question from skimming the webrev.
You've changed WeakRef so that an instance is initially "strong". Does
this have implications for other uses of WeakRef? I see one in
sun.rmi.transport.ObjectTable for example. I wonder if it would be
better to change the constructors so that the user of WeakRef decides
the reference is initially pinned or not. Alternatively maybe factory
methods should be introduced for all creation of WeakRefs and they be
named so that it's clear where the WeakRefs are initially strong or not.
Same thing with Target where I wonder about other uses as they will need
to know that the impl is initially pinned.
Otherwise it's good to get this bug fixed.
-Alan.
