bug#8254: race condition in dired.c's scmp function

[Eli Zaretskii <eliz@gnu.org>]

> Date: Tue, 15 Mar 2011 00:31:51 -0700
> From: Paul Eggert <eggert@cs.ucla.edu>
> CC: 8254@debbugs.gnu.org
> 
> On 03/15/2011 12:06 AM, Eli Zaretskii wrote:
> > How about a much simpler fix:
> > 
> >   while (l
> >   	 && (c1 = DOWNCASE ((unsigned char) *s1++),
> > 	     c2 = DOWNCASE ((unsigned char) *s2++),
> > 	     c1 == c2))
> >     l--;
> > 
> > (with suitable declarations of c1 and c2)?  Will that fix the
> > undefined behavior?
> 
> Yes.  But surely it's better to fix the problem so that
> usage of DOWNCASE is less error-prone.  Generally speaking,
> code is easier to read and contains fewer errors
> when function-like macros act like functions.

Maybe, but I wonder if there's a better solution even if we decide to
make these macros functions: I don't like to have the same static
function in every file that includes buffer.h, on platforms that don't
support inline functions.

Anyway, this is Stefan's and Chong's call.  I just voiced my
astonishment that such a simple problem needs such a jumbo change.