Path: utzoo!attcan!uunet!cs.utexas.edu!sun-barr!apple!bloom-beacon!usc!orion.cf.uci.edu!uci-ics!milne
From: milne@ics.uci.edu (Alastair Milne)
Newsgroups: comp.lang.pascal
Subject: Re: Pointers..
Keywords: pointers traversal dispose
Message-ID: <16404@paris.ics.uci.edu>
Date: 2 Jun 89 01:29:00 GMT
References: <3398@westfort.UUCP> <8835@phoenix.Princeton.EDU> <16969@swrinde.nde.swri.edu> <1150@draken.nada.kth.se>
Sender: news@paris.ics.uci.edu
Reply-To: Alastair Milne <milne@ics.uci.edu>
Organization: Educational Technology Center, Dept. of ICS, UC Irvine
Lines: 33

d87-jse@nada.kth.se (Joakim Sernbrant) writes
>Talking about pointers, is there any reason why I shouldn't do this on
>a PC running DOS without multitasking:
>
>while p <> nil do begin
>  dispose(p);
>  p := p^.next;
>end;

    Problems here start well before multitasking: you are accessing part of a
    variable (p^.next) after having destroyed it (dispose(p);).  Now it may be
    that your implementation of dispose doesn't actually impede your doing
    this (some Pascal's, for instance, have dispose return the pointer passed
    to them set to NIL), but I certainly wouldn't rely on it -- at best, it
    would be very untransportable.

    What I think you actually want is:

       while p <> nil do begin
	  holder := p;
	  p := p^.next;
	  Dispose( holder);
	  end;

     I don't know nearly enough of how OS/2 handles task-switching
     (time-slicing? priority queues?) to know what operations are guarded and
     what aren't; but I think this algorithm should dispose of a list with no
     incoherence problems.  At most, some concurrent application might be
     grabbing the memory this one releases as it is released -- but that
     should cause no problems.


     Alastair Milne