Re: [PATCH 2/4] Refactor window-system configuration

[Eli Zaretskii <eliz@gnu.org>]

> Date: Fri, 30 Dec 2011 01:07:00 -0800
> From: Daniel Colascione <dancol@dancol.org>
> CC: Dan Nicolaescu <dann@gnu.org>, emacs-devel@gnu.org
> 
> > Daniel, please keep in mind that at least Windows users normally know
> > very little about autoconf, and will have difficulty finding their way
> > around the configure script and its products, and understanding how
> > and where the value of that constant is computed.
> > 
> 
> Who exactly do you imagine being confused by a including macro and under
> what circumstances do you expect this confusion to arise?

The confusion will happen when J.R. Hacker wants to know where and how
the value of that macro is computed.

> A quick glance at config.h will reveal what header is being included

One must know to look in config.h in the first place.

And even then, the issue of _how_ that macro is computed will still
remain.

> The only people who need to care about the configure machinery are
> those who add window systems, and I hope these people would be
> familiar with autoconf.

There's another kind of people who might need to care: those who need
to figure out why their build process does the wrong thing with a
window system.  Suppose your patches, when applied, break the native
w32 build in some rare case, for example, because the wrong window
system is being computed.

> I'd really prefer not to have another big list of window systems in the
> source.

Why "another"?  I understand that you would be moving those other
places to a single header file.