Path: utzoo!censor!dybbuk!yunexus!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!cs.utexas.edu!usc!sdd.hp.com!ucsd!ucbvax!agate!darkstar!terra.ucsc.edu!daniel
From: daniel@terra.ucsc.edu (Daniel Edelson)
Newsgroups: comp.lang.c++
Subject: Smart Pointers Aren't Enough
Keywords: garbage collection, smart pointers, frustration
Message-ID: <11707@darkstar.ucsc.edu>
Date: 30 Jan 91 17:37:25 GMT
Article-I.D.: darkstar.11707
Sender: usenet@darkstar.ucsc.edu
Reply-To: daniel@terra.ucsc.edu (Daniel Edelson)
Organization: University of California, Santa Cruz
Lines: 31

Even if we had a ``perfect'' implementation of smart pointers,
I don't think that would be enough to permit robust copying
garbage collection. Unfortunately. 
By perfect I mean that all of the application's pointers of
type (node *) were smart pointers rather than raw pointers.
Perhaps I'm using the word perfect in a misleading way. If so, 
I hope someone will clairify.

Case in point:

	struct node { int i; };

	int * foo()
	{
		smart_pointer_to_node p = new node;
		int * i = &p->i;
		return i;
	}

How do we keep the node from being collected now that 
the smart pointer to it has been destroyed? 
What's the best course of action here? Forbid taking the
address of members of a collected object? I already place
so many restrictions on code that uses my garbage collector
that this seems unattractive. 

I am not a proponent of conservative garbage collection, but the
more I examine these problems the more I lean that way.

Daniel Edelson
daniel@cis.ucsc.edu