Adrian Aichner <adrian(a)xemacs.org> wrote:
the packages and xemacsweb CVS modules already require GNU make.
End users don't compile the packages; they just download them.
End users don't build the web pages; they just view them.
End users DO build XEmacs from sources. Requiring GNU make for XEmacs
itself is a much more significant step than requiring it for the
packages and xemacsweb.
I guess this requirement wouldn't be too much to ask for, but
some may
disagree.
I do. I think we will decrease our pool of users if we drop support for
non-GNU makes. It isn't that there are people who cannot build and
install GNU make; it's the nuisance factor. Some sysadmins may just
refuse to play along.
And it isn't as though we are in terrible shape, either. We currently
have a set of Makefiles that seem to work pretty well on the makes
people are currently using. Throwing that all away because of a few
nuisances seems a bit extreme.
The current problem appears to be that there exists a make that fails to
expand variables occurring inside of a pattern matching expression on
the right hand side of an =. Since the same make successfully expands
those variables when := is used instead, we can easily write a configure
test to detect and work around this. I'll do just that in a few
minutes.
Makefiles are bad enough without worrying about different variants,
but then again much to my disgrace I even hack xemacs.mak. Yuck!
If we were starting from scratch, I would worry about the make variants.
But we already have a bunch of infrastructure for dealing with them. I
think it is worth a little effort now and then to keep up with those
variants. Dump the work on me if nobody else wants to do it. I'll
collect all the make variants I can find and successfully build on some
platform I have and try them out from time to time.
--
Jerry James
http://www.ittc.ku.edu/~james/