Chris Gray wrote: > That much being said, I'm not sure I agree with David's assertion. If the > reference from a ThreadGroup to an unstarted Thread were a > PhantomReference, then IMO it would be enough to associate that reference > with some ReferenceQueue <rq>. <rq> needs to be scanned from time to > time(*), and the undead threads removed from the ThreadGroup
Well I guess we'd have to quantify "overhead". The ReferenceQueue approach doesn't require any array management, but there is a Reference per thread, and extra overhead on Thread creation and start(). As long as there are few unstarted threads that overhead should be fairly low, but it would be highly variable depending on the application. Looking at this from a real-time VM perspective I do not want to add such unpredictable overhead to Thread creation or starting - but then I don't use the Classpath Thread implementation. I'd adopt the current proposal without the ReferenceQueue part. I just don't think this particular leak is worth the cost of the fix. YMMV. Cheers, David Holmes _______________________________________________ Classpath mailing list [EMAIL PROTECTED] http://lists.gnu.org/mailman/listinfo/classpath
