NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part2/11211

Internet Message Format | 1996-08-05 | 1.8 KB

Path: lrz-muenchen.de!news
From: [e-mail] (Kurt Watzka)
Newsgroups: comp.lang.c
Subject: Re: Pointers to register
Date: 22 Mar 1996 10:33:31 GMT
Organization: Leibniz-Rechenzentrum, Muenchen (Germany)
Distribution: world
Message-ID: <4itvlr$[e-mail]>
References: <[e-mail]><[e-mail]> <[e-mail]><[e-mail]> <[e-mail]>
NNTP-Posting-Host: sun2.lrz-muenchen.de

[e-mail] (Brian R. Oldham) writes:

>You are quite right.  I checked in dos.h (BYTEREGS) and al, ah etc. are
>all unsigned char. So today I went back to my program and changed the
>declaration of scancode from pointer to int to a pointer to unsigned
>char. Sure enough, my program compiled without complaint, but - guess
>what - when I ran it in a main() the bloody system froze.

>So back to my original question: why does the ostensibly incorrect
>declaration:

>    *ptr = union_member_var;

1) This is not a declaration, and it is not an incorrect statement, because
   *ptr (in your case) is an int, and you _can_ assign an unsigned char
   to an int if an int can hold all possible values of an unsigned char.
   In this case, the result is well defined.

>work, and the correct:

>    ptr = &union_member_var;

If ptr is a pointer to int, and union_member_var is an unsigned char, this
statement is _not_ correct, and if I remember right, someone told you
that already.

Why do you think that it is more correct to assign a pointer to an
unsigned char to a variable that holds a pointer to an int than to
assign an unsigned char to a variable that is an int.

Kurt
--
| Kurt Watzka                             Phone : [phone]
| [e-mail]
| [e-mail]