Path: utzoo!news-server.csri.toronto.edu!rutgers!dimacs.rutgers.edu!seismo!uunet!pilchuck!dataio!fnx!nazgul!bright
From: bright@nazgul.UUCP (Walter Bright)
Newsgroups: comp.lang.c++
Subject: Re: Petzold's Windows examples with Zortech 2.1
Message-ID: <282@nazgul.UUCP>
Date: 13 Mar 91 17:57:03 GMT
References: <1991Mar6.130447.16332@grebyn.com>
Reply-To: bright@nazgul.UUCP (Walter Bright)
Organization: Zortech, Seattle
Lines: 40

In article <1991Mar6.130447.16332@grebyn.com> ted@grebyn.com (Ted Holden) writes:
/          zwndclass.lpfnWndProc   = WndProc ;
/must be changed to
/          zwndclass.lpfnWndProc   = (void *) WndProc ;
/and then the whole thing works with Zortech 2.1.  Nothing in windows.h would
/seem to suggest such a requirement.  What gives??

The trouble is that ZTC does ANSI type checking of argument and pointer types.
It seems that a large percentage of windows code (including the SDK sample code)
is written assuming that two types are the same *if they have the same number
of bytes of storage* (!). These problems are usually easilly solved by proto-
typing functions using ANSI C rules, and judicious use of casts. For instance,
there are cases where a long is cast to a 4 byte struct as in:
	func((struct X) y);
where y is a long. This can be converted to ANSI C by:
	func(*(struct X*)&y);
In general, this technique can be used when a type 'paint' is required (a paint
is when you change the type of an object without changing any bits).

We are working on a switch to the compiler to allow it to compile code that is
written under the assumption that size==type, but be aware that using this
feature will make it hard to port code from 16 to 32 bit compilers.
It will also conceal hard-to-find bugs in the code.

As a test of your compiler's type checking, it should fail to compile the
following:

	short *s;
	int *i;
	i = s;		/* type mismatch */

	int *p;
	int **pp;
	pp = p;		/* type mismatch */

	char *c;
	signed char *sc;
	sc = c;		/* type mismatch */

etc.