Am Fri, 5 Jun 2015 22:11:08 +0100 schrieb Jonathan Payne <core-l...@jpayne.net>:
> My problem was that finalization was not being run at all with the G1 > collector. The problem that an object is not detected as unreachable and not enqueued into the finalizer queue is not only a G1 problem. When an application has very short transactions and no objects get promoted then G1 and CMS (and PrallelOld) trigger very seldom an old GC, so this leads to a long time that finalizeable objects do not get cleaned up. Not sure if G1 is especially bad in that regards. I think it might be related to the bug that it never responded to the DGC cleanup unlike other GCs? Something similar to calling System.gc() every hour but with less negative effects would be a helpful tool for those situations. This is also a problem for other memory pools (class loaders, code cache, constant pool, direct buffers) as well. So having a slow background cleanup is a good thing in all cases. > My fix, BTW, was to use a back door (that I added to SharedSecrets) > in all the JDK classes that had a finalize() method, so that when a > resource is properly closed, by calling the close() method for > example, the back door would remove the Finalizer for the specified > object from the linked list of Finalizer objects I think it would be good to offer such an API to early remove an object from the reference queue. Gruss Bernd
