Martin Buchholz <martin(a)xemacs.org> writes:
>>>>> "WMP" == William M Perry
<wmperry(a)aventail.com> writes:
WMP> For actual OS things that they would need to check for that XEmacs'
WMP> config.h doesn't take care of... I can't see many other than
AC_CHECK_LIB
WMP> and testing for module-specific includes. But creating the makefiles, etc,
WMP> etc, is definitely best left to the modules themselves.
WMP> I've been meaning for well over a year to make the package system use
WMP> autoconf, but have never found the time. :(
Time....
One idea is that the module needs an XEmacs, and XEmacs already knows
_everything_ about the system.
I argue that this is a false statement. XEmacs _cannot_ know everything
about the system. Just say that we broke LDAP out into a separately
distributed module. Why should the core XEmacs have a configure-time test
for all the oddities of the different LDAP implementations? Or why should
it check for the existence of 'libghttp' just on the off chance that
someone will write a module that uses it?
XEmacs knows everything for the lispy side of things, but a module is not
going to be really worthwhile unless it interfaces with something XEmacs
doesn't know about yet (otherwise, why write it in C unless speed is
killing you?)
And its configuration variables are available from lisp via
config-value-hash-table.
Emacs/W3 used to take advantage of this, and actually uses the same type of
thing in w3-cfg.el, but it still cannot give you everything. There needs
to be a mixture.
-bp