Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site ulysses.UUCP
Path: utzoo!watmath!clyde!bonnie!akgua!gatech!ulysses!ggs
From: ggs@ulysses.UUCP (Griff Smith)
Newsgroups: net.lang.c
Subject: Re: condition convention 'if (10 == j)...'
Message-ID: <958@ulysses.UUCP>
Date: Sun, 28-Apr-85 21:53:54 EDT
Article-I.D.: ulysses.958
Posted: Sun Apr 28 21:53:54 1985
Date-Received: Mon, 29-Apr-85 04:37:19 EDT
References: <brl-tgr.9509> <137@bocklin.UUCP> <1456@amdahl.UUCP>
Organization: AT&T Bell Laboratories, Murray Hill
Lines: 48

> > To prevent silly mistakes like 
> > 	if (j = 10)
> > I usually write
> > 	if (10 == j)
> > By putting the constant first, I ensure that the compiler will catch the 
> > typo.
> 
> I think this is a good idea.  Any criticisms?  The only problem
> I have with it is that I am not accustomed to reading code written
> this way.
> -- 
> Gordon A. Moffett               ...!{ihnp4,cbosgd,sun}!amdahl!gam

Don't get me wrong, I also think the idea has some charm, but it's not as
innocent a change as it appears to be.  I recently found a bug in a C
compiler: if (j == 0.0) worked, if (0.0 == j) didn't.  It seems the
implementors wanted to avoid creating double precision constant zeros,
so they planned a two-step optimization where

L1:	.long	0x00000000, 0x00000000
	.text
	cmpd2	4(fp),L1

would be reduced to

	cmpd2	4(fp),$0

and then further reduced to

	tstl	4(fp)

Unfortunately, they forgot that "cmpd2 $0,j" doesn't fit the pattern;
the intermediate code pattern escaped to the final output.  Since $0
was not a legal representation of a floating point number (this was not
a VAX), the assembler screamed about an illegal address and sulked.
The compiler had apparently never encountered this case in any of the
UNIX(TM) System source.

The point of all this is: if you want to use unusual conventions, be
prepared to evade bugs in compilers designed for those who are less
imaginative.  Also be prepared to defend against flames from people
who find it hard to read your code.
-- 

Griff Smith	AT&T Bell Laboratories, Murray Hill
Phone:		(201) 582-7736
Internet:	ggs@ulysses.uucp
UUCP:		ulysses!ggs  ( {allegra|ihnp4}!ulysses!ggs )