Re: [lmi] Cygwin build failure

[Vadim Zeitlin]

[this is not very important, but keeping up with the one-thing-at-a-time
 principle, let me address this in a separate message]

On Tue, 30 Apr 2019 22:04:22 +0000 Greg Chicares <address@hidden> wrote:

GC> Yet we could do
GC>   make prefix=/opt/lmi/whatever ad_hoc_dir=/tmp/something -f 
install_mingw.make
GC> in which case the parentage is broken.

 I think trying to fool proof makefiles is a lost cause, it's all but
impossible (and really impossible in practice) to write them in a way that
would be immune to all possible redefinitions of the variables on make
command line. So the best we can do is to distinguish between the variables
that could be potentially set on make command line, such as prefix, and all
the other that are not supposed to be set there, which includes ad_hoc_dir.

 So I strongly believe that makefiles should be as simple and short as
possible rather than as smart as possible and IMNSHO lmi makefiles are
already towards the too smart end of the spectrum, so any simplification
would be welcome. In fact, maybe replacing this makefile with another shell
script wouldn't be a bad idea?

VZ