NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/lang/c/11711

Internet Message Format | 1992-07-29 | 3.4 KB

Xref: sparky comp.lang.c:11711 comp.sys.amiga.programmer:11783
Path: sparky!uunet!dtix!darwin.sura.net!gallux.gallaudet.edu!gallua!cads_cole
From: [e-mail] (Kevin Cole; Washington DC)
Newsgroups: comp.lang.c,comp.sys.amiga.programmer
Subject: Re: Matt Dillon's DICE (Amiga C) and rand() function
Message-ID: <1992Jul29.120044.1@gallua>
Date: 29 Jul 92 17:00:44 GMT
References: <1992Jul29.081247.1@gallua>
Sender: [e-mail] (News manager)
Organization: Gallaudet Research Institute
Lines: 70

In article <1992Jul29.081247.1@gallua>, I wrote:

> Here's a very short C program that doesn't do what the documentation claims.
> According to Matt Dillon's DICE documents, I can limit the range of values
> returned by rand() if I define the variable RAND_MAX to something other than
> the default.  I've tried various values, but I continue to get these huge
> (32-bit) numbers.  Tell me I'm not missing something obvious.
>
>      #include <stdio.h>
>      #include <stdlib.h>
>      #define RAND_MAX ((1023))
>      main()
>      {
>       int i;
>       for (;;)
>       {
>        i = rand();
>        printf("Random number %04X (%d)\n",i,i);
>       }
>       return(0);
>      }

Thanks all.  The *REAL* answer is "Friends don't let friends program stoned."
I'm recovering from a broken elbow, and the doc's got me on all sorts of brain-
deadening chemicals.

First, I must confess that I carelessly used the word variable, when I meant
constant.  In Matt Dillon's documentation, it says that RAND_MAX is defined in
stdlib.h.  Sure enough there was a

   #define RAND_MAX ((1 << (sizeof(int)*8) - 1)

or something like that.  I foolishly assumed from the doc that changing that
would change the behavior of rand().  If I hadn't been on pain-killers, I might
have realized that the program I wrote never actually USED my re-definition,
and that there was no way for rand() to know about a constant in the main
program that was never actually passed to it.

Some of you explained that what the documentation was actually suggesting was a
re-compilation of the library containing rand(), after redefining RAND_MAX. The
pain-killers have worn off sufficiently enough that I knew I didn't want to do
that.  (But just in case I was still brain-dead, one author pointed out that
this wasn't exactly a bright idea.)

Two of you offered the correct solution:  Use the modulus operator %.  (One was
the guy who explained why I didn't want to recompile rand().)

Anyway, many, many thanks to:

Ben Jackson, Dept Biological Structure-University of Washington
   <[e-mail]>
Anthony Lapadula, University of New Hampshire  -  Durham, NH
   <[e-mail]>
Daniel Barrett, University of Massachusetts, Amherst
   <[e-mail]>
John C. Chu, UB
   <[e-mail]>
Boris
   <[e-mail]>
Chris Torek, Lawrence Berkeley Laboratory, Berkeley
   <[e-mail]>
Nick Ruprecht, Institut fuer Informatik der Universitaet Freiburg
   <[e-mail]>
--

 ========== U.S. Mail =============================== E-Mail =================
 Kevin Cole      <Flatline>           |  Bitnet: [e-mail]
 Gallaudet Research Institute  (GRI)  |  Internet: [e-mail]
 Washington, D.C.  20002-3625         |  UUCP: ...!psuvax!gallua.bitnet!kjcole
 [phone]  x4592                |  The WELL:   kjcole