Path: utzoo!dptcdc!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!leah!rds95
From: rds95@leah.Albany.Edu (Robert Seals)
Newsgroups: comp.lang.c
Subject: Address of a goto label?
Keywords: hmmm
Message-ID: <1768@leah.Albany.Edu>
Date: 18 Apr 89 17:39:00 GMT
Organization: The University at Albany, Computer Services Center
Lines: 48

Hi again,

A quote from Ken Arnold's April 1989 "C Advisor" column in Unix Review:

	...you can take the address of
	a goto label and start stuffing
	data into it.

Well, the whole article was tongue-in-chic, and in fact included a reference
that said,

	...write a column which _looked_
	normal, but actually contained
	some harmless misinformation...

So, the idea of taking the address of a goto label seemed thrilling, and
I wrote a program to test it. Thus,

	#include <stdio.h>

	main()
	{
	    int *p;

	HERE:
	    p = &HERE;
	    printf("%d\n", p);
	}

But it don't verk; at least, the FIRST system I tested it on didn't verk.
The error was "undefined symbol" when taking the address of HERE. So, I
checked K&R2, which says:

	The only use of an identifier
	label is as a target of goto.

So, the failed compilation seemed correct. BUT, I tried it on the local
4.3-tahoe, and it compiled, and gave a warning about taking the address
of an array or function, and the & would be ignored, and it ran and the
program printed "76\n" (unoptimized). So I tried it on good old Ultrix
1.2 cc, and got the same as 4.3, except that I got 80 instead of 76.
And then gcc said "undeclared variable."

So I guess that this is an ANSI thingy-doo, and the answer I will get is,
"it's grotesque, but it works on old compilers" to the question (what is
this, Jeopardy?) "what is the deal here, anyway?"

rob