Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!apple!julius.cs.uiuc.edu!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!rutgers!mcnc!xanth!cs.odu.edu!zeil
From: zeil@cs.odu.edu (Steven J. Zeil)
Newsgroups: comp.lang.c++
Subject: Re: Book of the Week Club
Message-ID: <1990Nov29.172517.3695@cs.odu.edu>
Date: 29 Nov 90 17:25:17 GMT
References: <59405@microsoft.UUCP>
Sender: usenet@cs.odu.edu (Usenet News Poster)
Reply-To: zeil@cs.odu.edu (Steven J. Zeil)
Organization: Old Dominion University, Norfolk, VA
Lines: 16
Nntp-Posting-Host: dolphin.cs.odu.edu

In article <59405@microsoft.UUCP> jimad@microsoft.UUCP (Jim ADCOCK) writes:
>Be forwarned however that in his ref counted pointer class he makes the
>common mistake of decrementing the old referenced object before incrementing
>the new referenced object.  This has the effect of possibly prematurely 
>deleting the old referenced object when it is the same object as the newly
>referenced object.  Probably the best fix is just to explicitly test against
>the possibility that the old and new objects are the same.

Actually, that fixes only a special case of the general problem.
Suppose that A points to B, B points to C, and I try to replace the
pointer in A to make it point to C. This passes your proposed test,
but decrementing B's count before incrementing C's could result in
both B and C being deleted, when in fact only B should have been
deleted.

                                                            Steve Z