Re: [boost] boost::nil

From: David Abrahams (abrahams_at_[hidden])
Date: 2001-06-08 09:30:22

Is there something wrong with "null"?

There are more-foolproof ways to #undef nil: you use the STLport approach of
making a wrapper header for the MacOS header which defines it. As someone
who's spent a lot of time programming Macs, I don't think this problem is so
bad that it's worth compromising the design of our library for.

-Dave

----- Original Message -----
From: "Chris Little" <cslittle_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Friday, June 08, 2001 10:02 AM
Subject: RE: [boost] boost::nil

It is a design flaw but it has been around a long time.  nil is Pascal's
NULL.  All of the Mac OS's system interfaces were originally defined as
Pascal and nil has been in the system headers since 1984.

Attempting to #undef a system constant is doomed to failure.  It would be
very sensitive to include ordering.  Think of it like trying to redefine
NULL.

Jeremy suggests in another email that boost::nil be renamed
boost::null_pointer or something like that.  I believe this is the route to
take.

Chris

> -----Original Message-----
> From: jk_at_[hidden] [mailto:jk_at_[hidden]]
> Sent: Friday, June 08, 2001 9:44 AM
> To: boost_at_[hidden]
> Subject: Re: [boost] boost::nil
>
>
> 8 Jun 2001 17:23:45 +0400 Chris Little ÎÁÐÉÓÁÌ:
> >On a Mac, nil is a preprocessor macro included in the system
> headers as a
>
> Exactly that way, lowercase macro? Then this is BIG design
> flaw in MacOS.
>
> >legacy of the days when the OS had a Pascal interface.
>
> How about something like
>
> #ifdef nil
> #undef nil
> void * const nil=0; // or what appropriate, in global namespace.
> #endif
>
> somewhere in config.hpp?
>
> --
> jk
>