NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part1/9441

Internet Message Format | 1996-08-05 | 2.1 KB

Path: keats.ugrad.cs.ubc.ca!not-for-mail
From: [e-mail] (Kazimir Kylheku)
Newsgroups: comp.lang.c
Subject: Re: malloc question
Date: 10 Mar 1996 10:18:06 -0800
Organization: Computer Science, University of B.C., Vancouver, B.C., Canada
Message-ID: <[e-mail]>
References: <4htonk$[e-mail]> <4hv45e$[e-mail]>
NNTP-Posting-Host: keats.ugrad.cs.ubc.ca

In article <4hv45e$[e-mail]>,
William Kaufman <[e-mail]> wrote:
 >In article <4htonk$[e-mail]> [e-mail] (Alex Chu) writes:
 >]
 >] main()
 >] {
 >]   PITEM head, current;
 >]   head=(PITEM) malloc(sizeof(ITEM));
 >]             ^^^^^^^
 >]   head->val=1;
 >] }
 >]
 >] I want to know why need to use the type casting PITEM in front of the
 >] malloc ?  Please help!
 >
 >    Either:
 >
 >    a) you're using a pre-ANSI compiler which complains that you can't
 >portably cast from (char *) to (PITEM) (in which case, you could go and
 >get an ANSI compiler); or,
 >
 >    b) you aren't including <stdlib.h>, where malloc() is defined as
 >returning (void *), and the compiler thinks it returns (int).

Which is a serious error that summons undefined behavior.  Never mind that the
mapping function from integral types to pointers is implementation defined;
never mind that there is a possible disparity between the size of an int and a
pointer---these issues don't even enter into it. An int return value may not
even be passed back _in the same register_ as a pointer return value! This is
simply a case of calling a function that is declared one way, and defined
another way, cast or no cast.

If I were implementing a C compiler for, say, a 680x0 architecture, I might use
the A0 register for returning pointers, and D0 for returning integers, just
because it is more convenient for the caller to have a pointer ready in an
address register and an integer ready in a data register.

The fact that a function call which has an incompatible declaration (implicit
or otherwise) with respect to the function definition is undefiend would allow
me that implementational latitude.
--