Message-ID:	<3F25F8B9.32167C0@worldnet.att.net>
From:	Paul Cousoulis <paulcsouls AT worldnet DOT att DOT net>
X-Mailer:	Mozilla 4.78 [en] (Win95; U)
X-Accept-Language:	en
MIME-Version:	1.0
Newsgroups:	comp.os.msdos.djgpp
Subject:	Re: integer overflow
References:	<3F246120 DOT 63C3753C AT worldnet DOT att DOT net> <bg4dkn$8et$1 AT newsreader1 DOT netway DOT at>
Lines:	64
Date:	Tue, 29 Jul 2003 04:32:30 GMT
NNTP-Posting-Host:	12.81.72.7
X-Complaints-To:	abuse AT worldnet DOT att DOT net
X-Trace:	bgtnsc04-news.ops.worldnet.att.net 1059453150 12.81.72.7 (Tue, 29 Jul 2003 04:32:30 GMT)
NNTP-Posting-Date:	Tue, 29 Jul 2003 04:32:30 GMT
Organization:	AT&T Worldnet
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp
Reply-To:	djgpp AT delorie DOT com

Thanks,

I was trying to get this tgt-edif code which was written for Linux to
compile with DJGPP and this code was in it. It Looked okay to me. My
reference book for C is 'A Book on C' by Ira Pohl and Al Kelley,
Copyright 1998. I guess it's getting a little dated.

Thanks
Paul

News Reader wrote:
> 
> Your compiled code will not result in an overflow.
> You are just getting an overflow warning from the
> compiler, because the sign bit is affected twice during
> the evaluation of your code. It was mentioned in this
> thread that DJGPP defaults to signed 32-bit constants.
> 
> In order to prevent compiler warnings you can
> change your code as follows:
> 
>     unsigned res;
> 
>     res=1; res<<=31; res--;    // was: (1<<31)-1;
> 
> Other examples which will work are:
> 
>     res = (1u<<31)-1;
>     res = ((unsigned)1<<31)-1;
>     res = (unsigned)(1<<31)-1;
> 
> Please note that using 'unsigned res' instead of
> 'unsigned long res' will not change anything because
> DJGPP will produce 32-bit integers in both cases.
> 
> If your intention is to double the size of the variable
> you have to use 'long long' in your declaration.
> This will get you to 64 bits and your maximum
> shift count will therefore be 63 (unsigned).
> 
> Unfortunately this does not help much because
> your code will still be evaluated based on 32-bit
> signed integer constants - unless you use one of
> the following methods:
> 
> long long res;
> 
>      res = (1LL<<31)-1;
> or
>      res = ((long long)1<<31)-1;
> or
>      res=1; res<<=31; res--;
> 
> "Paul Cousoulis" <paulcsouls AT worldnet DOT att DOT net> wrote in message
> news:3F246120 DOT 63C3753C AT worldnet DOT att DOT net...
> >
> > Why does this code result in an integer overflow?
> >
> > unsigned long res;
> >
> > res = (1<<31)-1;
> >
> > Thanks
> > Paul

- Raw text -