From mboxrd@z Thu Jan  1 00:00:00 1970
To: 9fans@cse.psu.edu
From: "Douglas A. Gwyn" <DAGwyn@null.net>
Message-ID: <3D9A13FE.E52C05F8@null.net>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
References: <821f6979e86cb87e2f87d83def62d004@vitanuova.com>
Subject: Re: [9fans] function inconsistently declared
Date: Wed,  2 Oct 2002 09:01:36 +0000
Topicbox-Message-UUID: faf164e0-eaca-11e9-9e20-41e7f4b1d025

C H Forsyth wrote:
> in old C,
>         f(a, b)
>         int a;
>         char b;
>         {
>         ...
>         }
> actual parameters for b were always promoted to int
> (since prototypes weren't available, and the compiler didn't maintain a
> file of the types of parameters to internal and external functions, it
> had to assume int for char, double for float etc.)

Actually it didn't *have* to, but it was convenient for the way
PDP-11 function linkage worked.

In the presence of a prototype, it is quite possible for the
compiler to use a linkage that differs from the one used for
unprototyped functions.  In this regard also remember that
use of a variable-argument function requires that the ,...)
prototype be visible; quite a few systems use a linkage for
that situation that is different from the one for unprototyped
functions.  I still occasionally have to fix up some legacy
code like:
	error("foo %d\n",1);	/* no prototype in scope */
	error("punt\n");
	...
	void error(char *fmt, int x) {
		... something involving &x
	}