bug#8881: config.h double inclusion

[Alfred M. Szmidt]

   > I'm thinking that maybe <config.h> should be generated with a double
   > inclusion guard

   But the general rule is that <config.h> must always be included
   first, no?  So there shouldn't ever be a possibility of including
   it twice.

Right, which is the case in inetutils in all places where we use
<config.h>.  The problem pops up when you have another header file
that includes <config.h> (as the first thing) to use macros from it
(for libinetutils.h it is PACKAGE_BUGREPORT and HAVE_FORK or
something).

   It might be better to have config.h do something like this:

   #ifdef CONFIG_H
   # error "config.h included twice"
   #endif
   #define CONFIG_H

Not against it, but I think that:

#include <config.h>
#include <argp.h>       /* or some other gnulib header... */
#include <config.h>     /* via some internal #include; where it is the first 
line  */

should be valid to do.

   > A bit later, we do:
   > 
   >   #include <libinetutils.h>
   > 
   > which will cause warnings like:
   > 
   >   ../config.h:561:1: warning "HAVE_DECL_PROGRAM_INVOCATION_NAME" redefined

   Sp libinetutils.h includes config.h?  Then that is a problem.  One
   possible fix is to arrange for libinetutils.h to be built from
   libinetutils.in.h, the same way that (for example) stdint.h is
   built from stdint.in.h in gnulib.

Right, sorry if I wasn't clear.  I would prefer not to, the fix that I
might go for if we can't fix this in autoconf would be: to simply do:

#ifndef HAVE_CONFIG_H
#error "config.h hasn't been included; please include it"
#endif

But I think that autoconf should have the guard for double inclusion.