Path: utzoo!utgpu!attcan!uunet!lll-winken!ames!husc6!cs.utexas.edu!sm.unisys.com!oberon!news
From: news@oberon.USC.EDU (USENET News)
Newsgroups: comp.sys.ibm.pc
Subject: Re: Another msc 5.1 bug
Summary: Compiler contradicts manual
Keywords: msc
Message-ID: <14495@oberon.USC.EDU>
Date: 6 Jan 89 22:34:05 GMT
References: <139@smsdpg.uu.net> <200015@hpmcaa.HP.COM>
Reply-To: gould@physics.usc.edu (Christopher Gould)
Organization: University of Southern California, Los Angeles, CA
Lines: 37

In article <139@smsdpg.uu.net> seg@smsdpg.UUCP (Scott Garfinkle) says:
	>I just ran across a neat new (?) bug in MSC5.1:
	>fun() {
	>	extern int * const a;
	>	int * const b = a;
	>	b[0] = 1;
	>}
	>produces "foo.c(5) : error C2166: lval specifies 'const' object"
	>In other words, you can't use a const pointer to a
	>non-const chunk of memory. 
Scott is right.  This is a bug.

In article <200015@hpmcaa.HP.COM> marco@hpmcaa.HP.COM
(Marco Dalla-Gasperina) says:
	>I'm not sure of this, but...
	>	int * const a      means the value (*a) is constant 
	>	int const * a      means the value (a) is constant
	>Seems backwards but thats what the manual says!
Marco is wrong.  The manual says:
"int const * p" declares a pointer to const,
		cf. sec. 4.4 (p. 62) and sec. 4.4.6 (p. 74), while
"int * const p" declares a constant pointer to int,
		cf. sec. 4.5.5 (p. 81) and sec. 4.7.1 (p. 90).
This is consistent with Harbison & Steele, 2nd ed, sec 11.4.3:
	"int * const const_pointer;
	 const int * pointer_to_const;"
Perhaps intentionally H&S do not use the form "int const *" as
possibly misleading.

Related to this topic, why isn't this example from H&S p.265 illegal?
	const int *p;
	*p = 5;   /* Legal, but ... */

Christopher M. Gould          Internet: gould@physics.usc.edu   (128.125.1.194)
Dept. of Physics              Arpanet:  gould@usc-physics.arpa
Univ. of Southern Calif.      UUCP:     ...{uunet,sdcrdcf}!oberon!physics!gould
Los Angeles, CA 90089-0484    Phone:    (213)743-8521