Godmar, can you please ad a new Jira issue at
http://jira.codehaus.org/browse/CASTOR and attach a valid unit test (e.g. XML schema, classes generated, comments from this thread) to it. If you happen to have a patch ready for us to look at, even better .. ;-). Werner Godmar Back wrote: > The title says it all: if I specify bound properties in my castor > options, Castor will generate bean property change support, despite > the fact that these classes have no changeable fields, and since their > constructor is private, they also cannot (and should not) be > subclassed. > > I believe that these methods (addPropertyChangeListener, > removePropertyChangeListener, and notifyPropertyChangeListener) should > not be added to these classes. (Related, I also believe that the > internal field "_type" should be declared final.) > > Why am I complaining about this? > > I have code that examines Castor-generated code via reflection to see > whether a particular property is itself a Java bean. It does this by > looking for the "addPropertyChangeListener" method. If this method is > found, a property change listener is added. > > Unfortunately, my code mistakes the Castor-generated "types" constants > for changeable Java beans. Since these constants are also referred to > by static member variables, adding a listener to them results in a > humongous memory leak that quickly leads to out-of-memory errors. > > - Godmar > > --------------------------------------------------------------------- > To unsubscribe from this list please visit: > > http://xircles.codehaus.org/manage_email > --------------------------------------------------------------------- To unsubscribe from this list please visit: http://xircles.codehaus.org/manage_email
