Re: [PATCH 2/9] Refactor window-system configuration

[Eli Zaretskii <eliz@gnu.org>]

> Date: Tue, 07 Aug 2012 01:19:27 -0700
> From: Daniel Colascione <dancol@dancol.org>
> 
> This change streamlines the window system selection code in
> configure.in and moves many common function declarations from
> window-specific headers to frame.h.  It introduces a new TERM_HEADER
> macro in config.h: we set this macro to the right header to use for
> the window system for which we're compiling Emacs and have source
> files include it indirectly.  This way, we don't have to teach every
> file about every window system.

A general comment on this refactoring: the current code does not
necessarily assume that a build must have only one type of window
system.  Your refactoring forces that.  Here's an example:

> --- a/src/dispnew.c
> +++ b/src/dispnew.c
> @@ -48,17 +48,9 @@ along with GNU Emacs.  If not, see <http://www.gnu.org/licenses/>.  */
>  
>  #include "syssignal.h"
>  
> -#ifdef HAVE_X_WINDOWS
> -#include "xterm.h"
> -#endif /* HAVE_X_WINDOWS */
> -
> -#ifdef HAVE_NTGUI
> -#include "w32term.h"
> -#endif /* HAVE_NTGUI */
> -
> -#ifdef HAVE_NS
> -#include "nsterm.h"
> -#endif
> +#ifdef HAVE_WINDOW_SYSTEM
> +#include TERM_HEADER
> +#endif /* HAVE_WINDOW_SYSTEM */

The current code could well enough use all 3 of the headers.  Your new
code precludes that.

(Yes, I know that currently including more than one means trouble,
because there will be conflicting declarations.  But still, your
change sets one more obstacle to the long-term goal of allowing
several different frame types in the same session.)

I'd like Stefan's and Chong's opinion on this before delving into the
details of this changeset.