Re: PATCH: More stdbool tweaking

[Daniel Jacobowitz <drow at mvista dot com>]

On Thu, Feb 07, 2002 at 01:55:34PM -0800, Ian Lance Taylor wrote:
> Andrew Cagney <ac131313@cygnus.com> writes:
> 
> > > I think I need a bucket ... :-)  You'll need to explictly include
> > >> "config.h" as well.  Is it possible to fix bfd.h?  If this is
> > >> hurting GDB it is going to hurt a lot more :-(
> > > It is not possible to fix bfd.h until we purge bool from that header
> > > entirely.  It's a gross namespace issue.  I'll look at that, but only
> > > post-2.12.0 release.
> > 
> > 
> > So I would strongly encourage fixing the problem before the
> > branch/release.  If the problem is messing up GDB now is going to mess
> > up a lot more tools once it is made official in a release.
> > 
> > GDB's managed to survive changes far less mechanical than this one :-)
> 
> The use of bool is not a problem, since bfd.h doesn't use bool.  It
> uses boolean, which nobody appears to be complaining about.
> 
> The only bfd.h issue is the definitions of true and false.  Daniel has
> proposed a simple patch which works provided stdbool.h is included first.  I
> already proposed a simple patch which should work in all cases:
>     http://sources.redhat.com/ml/binutils/2002-02/msg00068.html
> However, I have not tested it.

(Is the assumption that the first thing in an enum has value 0 always
valid?  I think so, but not sure.)

In any case, your patch would not finesse the problem.  False is
defined to 'false', an enum value, in <stdbool.h>.  While the value
false evalutates to 0, that would not fix the problem, since it is not
the same.

I stand by my patch :)

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer