Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!rutgers!ucsd!ucbvax!decwrl!hplabs!hp-sdd!ncr-sd!ncrlnk!ncrwat!swdev!neil
From: neil@swdev.Waterloo.NCR.COM (Neil A. Law)
Newsgroups: comp.lang.c
Subject: Re: near and far in MSC
Message-ID: <366@swdev.Waterloo.NCR.COM>
Date: 29 Jul 88 15:15:47 GMT
References: <331@bdrc.UUCP>
Reply-To: neil@swdev.Waterloo.NCR.COM (Neil A. Law)
Organization: E & M Waterloo, NCR Canada Ltd, Waterloo, Ont, Canada
Lines: 46

>
>    result = jess (LOAD_KB, "rules", 0);
>
>The code that handles the variable argument list looks like:
>
>    va_start(ap);
>    while ((args[argno] = va_arg(ap, char *)) != (char *) 0)
                         .
                         .
                         .
>
>I think the problem has to with the difference between near and far
>pointers (eg. the terminating zero in the call will be interpreted as
>a NULL pointer to char by the function jess).  Could someone explain
>to me how one programs under memory models other than the small one?
>
>John C. Lusth
>Becton Dickinson Research Center
>RTP, NC 27709
>
>...!mcnc!bdrc!jcl

John, 
    Your problem is definitely in the difference between near and far pointers
and some inconsistency in your coding. First, short pointers are 2 bytes in
length, far pointers are 4. Second, if you analyze your call to "jess" the 
last parameter your are passing it is an int, not a pointer.

When you call the function "jess" and pass it a "0" as the last parameter,
the compiler pushes an int onto the stack (a 2 byte quantity).  When "jess"
computes the while statement, it is looking for a 4 byte quantity (char *).

Since near pointers are the same length as an int in a small model your code
works fine there.

Solution, cast the 0 in the call to jess with a (char *), this way
the compiler will always push sufficient information on the stack.

Hope this helps.

P.S. We had this problem in droves when we converted a 68000 application to
run on the PC.  On the 68K ints and pointers are the same size so comparisons
of "ptr == 0" worked fine on the 68K but failed on the PC in large model.  It
took our developers a long time to track down all of these inconsistencies and
taught us a valuable lesson, casting is important, especially in dealing with
constants.