>This source-changes mail reminded me of something I have been meaning to
>suggest for a while...
>
>        Add dependencies on Makefile for  SYSVSHM, SYSVIPC options.
>
>Nothing wrong with that change as such of course - what's wrong is the
>need for it, the way BSD has done kernel options (since day 1) is dumb.

Hi,

I haven't digested read all your long message, but I thought a fair
bit about just doign the quick hack of adding dependencies on
Makefile.  In the end, I put in the quick simple change ony becuase I
also want to pull it up to 1.3.2.

A much better fix than mine is to use NetBSD's defopt machinery, which
is like option, but creates a header file "opt_<OPTION>.h" in the
build directory, which contains "#define OPTION" if the option is set,
and nothing otherwise.  It's like your proposal except it doesn't use
the options/ subdir, and config does need to know about them all.  
So far that hasn't been a showstopper.

I thought about using defopt for __all__ the ``MI'' flag options in
options(4). We could even define the option flags in a file containing
human-readable definitions of what they mean, and use an awk script to
generate either a file of "defopt" definitions or mandoc input.

I agree completely that a better scheme for managing options and
dependencies on options would be very nice.  Ideally, it could even
support `binary' distributions for people with slow machines. Tho' The
momentum there is more towards supporting cross-compilers.