Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83 (MC830713); site vu44.UUCP
Path: utzoo!linus!decvax!mcvax!vu44!keie
From: keie@vu44.UUCP (Ed Keizer)
Newsgroups: net.lang.c
Subject: Rule of thumb for foo++
Message-ID: <297@vu44.UUCP>
Date: Mon, 9-Jan-84 11:12:00 EST
Article-I.D.: vu44.297
Posted: Mon Jan  9 11:12:00 1984
Date-Received: Wed, 11-Jan-84 03:33:49 EST
Organization: VU Informatica, Amsterdam
Lines: 23

Mark Plotnick (mit-eddie!mp) mentioned
        ( (foo+=1) - 1 )
as a rule of the thumb for post-increment
operators like in
        foo++
and said that he didn't know any off-hand exceptions.
Well, I met one last year that shows one of the pitfalls of
overflow.
Assume that foo is declared as an unsigned short and that the
C compiler uses 32-bit int's. Now see what happens if we give
foo the value 0xFFFF.

foo++ delivers the value 0xFFFF, but
( (foo+=1) -1 ) delivers the value 0xFFFFFFFF,
because the result of (foo+=1) is 0 and not 0x10000.
The C reference manual is quite explicit here:
"The value ___of the assignment expression___ is the value
stored in the left operand after the assignment has taken place."

                                        Ed Keizer
                                        Vrije Universiteit
                                        Amsterdam
                                        ...!mcvax!vu44!keie