[gclist] Finalization and object orientation.

Daniel Wang danwang@CS.Princeton.EDU
Fri, 28 Mar 1997 15:42:16 -0500

>>>>> "Hans" == Hans Boehm <boehm@hoh.mti.sgi.com> writes:

    Hans> On Mar 28, 10:39am, Eliot & Linda wrote:
    >> So can we agree that in languages which reify execution state and
    >> don't allow pointer forging that reachability is decidable?

    Hans> I don't think so.  Whether or not the context will be accessed is
    Hans> certainly undecidable in general.

Reading the last few messages, I think Eliot and others are confusing the
deciablity of computing the transitive closure of the "points to relation"
with the more general notion of reachability defined as whether or not an
object is every used again by a given program. I think Eliot been talking
about when you can and cannot compute the TC of "points to relation" while
others have been speaking about the more general notion.  

The more general notion is clearly undecidable. Even with a "points to
graph" there's always "semantic garbage" that's never collected, though as
some have pointed out type reconstruction can help you reclaim some of it.

Is this the point of the confusion? Or have I missread everyone?