On 11/03/17 18:56, Hans Boehm wrote: > > Stepping back, I'm a little surprised at the deprecation. In my opinion, > the problems with finalizers are, in rough order of significance:
I would have thought the most significant problem with finalizers is that finalizers may never run unless the application runs out of memory. In practice, this means that finalizers (and PhantomReferences, Cleaners, and anything else where cleanup is based on reachability) can't be used to manage scarce resources. Given that this is a primary use case for finalizers, I think we're done. > 3) Some less significant issues, like runFinalization() > synchronization, difficulty in properly triggering GC for non-memory > resource exhaustion, slow and deadlocked finalizers, inability to > finalize more than once, potential need for finalizer guardians, > etc. I guess "difficulty in properly triggering GC for non-memory resource exhaustion" is closest to what I just described. > Although I agree we'd be marginally better off without finalizers, > it's unclear to me the improvement is substantial enough to bother > without more of a replacement story. I can't disagree with that. But I welcome this move as an opening gambit. Andrew.
