Oh, the irony of a class called "cycle breaker" being involved in a memory leak!
>From a brief look at this class: http://jira.codehaus.org/secure/attachment/27944/CycleBreaker.java it seems the problem is caused by calls to "startingToCycle" not being matched with calls to "releaseCycleHandle". You may wish to examine the code for this situation. - Godmar On Tue, Jun 23, 2009 at 2:11 PM, Harry Hehl <[email protected]>wrote: > Hello > > With castor 1.2 and 1.3 if equals/hashcode generation is enabled the > generated equals() method calls > org.castor.core.util.CycleBreaker.startingToCycle . > > org.castor.core.util.CycleBreaker contains private static > IdentityHashMap _threadHash = new IdentityHashMap(); > > During testing it was discovered the application was leaking memory. It > was tracked to _threadHash referencing many objects when they are no > longer referenced anywhere else. > > Why would an object not be removed from _threadHash? Not generating > hash/equals methods eliminates the leak, but they are required by the > application. > > Can this cycle breaker code be disabled? i.e. is it really necessary? > > Thanks. > > > > --------------------------------------------------------------------- > To unsubscribe from this list, please visit: > > http://xircles.codehaus.org/manage_email > > >
