Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!rutgers!mit-eddie!husc6!necntc!cullvax!drw
From: drw@cullvax.UUCP
Newsgroups: comp.lang.c
Subject: Question about the new standard
Message-ID: <812@cullvax.UUCP>
Date: Wed, 18-Feb-87 11:50:25 EST
Article-I.D.: cullvax.812
Posted: Wed Feb 18 11:50:25 1987
Date-Received: Thu, 19-Feb-87 20:27:46 EST
Organization: Cullinet Software, Inc., Westwood, MA
Lines: 39

I've got a question about the new standard.  (Please discuss this only
within the context of the standard, or possible desirable
modifications of it.  I don't want to hear "well, *my* compiler does
it this way...")  Consider the code:

	char *p;
	int i;

	i = *p++ == *p++;

As far as I can tell, it is legitimate to increment p only once.  The
reasoning is: Evaluations of subexpressions are permitted to be
interleaved in any way, except (1) a subexpression whose operator has
a sequence point must not be interleaved with any other evaluation,
and (2) the result of an assignment or prefix inc/decrement must be
stored into its target before the resulting value is used.  Since the
above expression has no sequence points, we can evaluate it in the
following order:

	  1 2 3  4  5 6
	i = *p++ == *p++

A. Calculate p+1 (for 3)
B. Calculate p+1 (for 6)
C. Store p+1 (from A) in p (for 3)
D. Store p+1 (from B) in p (for 6)
E. Indirect (for 2 and 5)
F. Compare (for 4)
G. Assign (for 1)

In fact, the same reasoning works for "*++p == *++p", because the
above sequence of (abstract) execution gets the incremented values
stored before they are used for indirection.

Dale
-- 
Dale Worley		Cullinet Software
UUCP: ...!seismo!harvard!mit-eddie!cullvax!drw
ARPA: cullvax!drw@eddie.mit.edu