2008/6/28 Vladimir G. Ivanovic <vgivanovic(a)comcast.net>:
The build requirements are really not that onerous. I don't have
experience with either system, so I don't speak authoritatively and I'm not
pushing either option, but here's what I've found out:
SCons depends on just Python (already installed in all Linux distributions)
and CMake has no additional requirements. Both systems are available
pre-packaged for Gentoo, Fedora 9, Debian/Ubuntu and apparently cygwin, so
the overhead for a person wishing to build an XEmacs that would use them is
not high. I don't know about pre-packaged versions for Windows or Mac OS X,
but both claim to run and build (natively) for those environments.
The tradeoff seems to be XEmacs build maintainer time (and sanity) vs a
small, possibly zero, user/builder overhead. (It turns out that both were
installed on my Gentoo and Fedora systems.) Since I'm not the XEmacs build
maintainer, officially I don't care.
That being said, I'm partial to the idea of replacing a build environment
based on make and autotools. As a person with a passing interest in
programming languages, I'd prefer SCons because it uses Python, a
full-fledged, modern, widely-used, multi-paradigm, programming environment
that comes with many, many libraries with even more available as add-ons.
But, as I said, I'll defer to those who are directly affected by the choice
of build system: it's not my call (just as Mercurial wasn't).
I have used scons. My experience was that it was very good at
eliminating lots of mess and confusion for the features of autotools +
Makefile that it supports out of the box. The problem I had is that
it didn't support everything I needed out of the box. That left me
with the necessity of writing a bunch of Python code. I eventually
gave up and went back to autotools + Makefile.
On the other hand, the stuff I needed that scons did not support out
of the box is not needed by the XEmacs build system, either. It is
possible that it could simplify our build system. We won't know until
somebody tries. I doubt this proposal will go anywhere unless
somebody steps up to the plate and demonstrates a fully (or at least
mostly) functional replacement build system.
XEmacs-Beta mailing list