Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!rutgers!mit-eddie!think!ames!lll-tis!elxsi!len
From: len@elxsi.UUCP
Newsgroups: comp.os.minix
Subject: Re: Global variables (was Re: More on sh/shar
Message-ID: <408@elxsi.UUCP>
Date: Fri, 5-Jun-87 20:51:57 EDT
Article-I.D.: elxsi.408
Posted: Fri Jun  5 20:51:57 1987
Date-Received: Sat, 6-Jun-87 20:24:12 EDT
References: <1643@encore.UUCP> <581@uokmax.UUCP> <498@sortac.UUCP> <587@uokmax.UUCP>
Reply-To: len@elxsi.UUCP (Len Mills)
Organization: ELXSI Super Computers, San Jose
Lines: 30

In article <587@uokmax.UUCP> rmtodd@uokmax.UUCP (Richard Michael Todd) writes:
>In this example, the 'a's in both files are global variables.  On UNIX
>systems, compiling with "cc foo.c bar.c" will succeed and in the resulting
>program both modules, when referring to 'a' refer to the same memory
>location, thus the program will print out:
>5
>7
>On non-UNIX systems, the linker will complain about multiple definitions
>of the symbol a.
>Note that the functions 'main' and 'bar' above do not have to declare
>within the function body that 'a' is extern, as you did in your example
>code:

Not necessarily true.  There are numerous approaches to the handling
of the consistency among the declaration of the same external name 
or top-level-declaration in different files.  The example in the 
referenced article assumes BSD-style unix systems, as 4.2BSD takes
a mixed strategy approach, and in effect merges the declarations 
at link time, in the manner of a FORTRAN "common" declaration.

ATT-style systems do not take this approach, and will absolutely
cause errors.  For a better explanation than I could manage without
plagarism, see section 4.8 of Harbison & Steele, _A_C_Reference_Manual_,
Prentice-Hall, 1984.

Let's not argue over whether or not BSD and ATT are both UNIX systems.

The above is not intended as a flame, just the facts, folks.

Len