The current mechanism in Cactus requires that there are files darwinX.Y for every single OSX release (and also for hpux, freebsd, aix, openbsd, osf, solaris, unicos, unicosmp basically anything that is not GNU/Linux) which is cumbersome as we have to keep track of Apple’s releases and always end up adding a new boilerplate file

#! /bin/sh
# /*@@
#   @file    darwin18.0.0
#   @date    2019-03-17
#   @author  Roland Haas
#   @desc
#            Known architecture stuff for OS X (Darwin) release 18.0.0
#   @enddesc
#   @version $Header$
# @@*/

. $srcdir/known-architectures/darwin

Pull request https://bitbucket.org/cactuscode/cactus/pull-requests/59/cactus-allow-for-globs-in-architecture/diff avoids this by letting the files be named as shell globs and matches those globs against the architecture name.

Kind of ugly but does get the job done. An alternative may be to instead allow a magic comment:

#CCTK ARCHITECTURE=darwin[0-9]*

parse the file and use it if its REGEX matches the architecture.

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2255/reduce-number-architecture-files