Finalization and CNI

[Tom Tromey]

>>>>> "Hans" == Boehm, Hans <hans_boehm@hp.com> writes:

Hans> Just to clarify: The garbage collector tries NOT to scan the
Hans> C/C++ new/malloc heap.

Andrew and I talked about some CNI changes the other day.  We'd like
to do 3 things:

* Finish Bryce's binary compatibility work.  I'm not sure how the CNI
  part of this would work.

* Move from using plain pointers to smart pointers, so we can get
  proper null pointer checking on all platforms

* Tighten the rules on where pointers to GCd objects are allowed to
  be kept (to let us reduce the root set)

I don't know when these things will happen.  But the basic idea is
not to promise any sort of binary compatibility until all of them are
done -- each of them represents an ABI change.

Tom