[lmi] Building lmi on various OSs

[Greg Chicares]

I'm convinced that cross-compiling on GNU/Linux is better for me than
using cygwin in a qemu-kvm virtual machine running 32-bit msw-xp. We
had hoped that Kim could easily try GNU/Linux and decide for herself
whether she prefers that to msw-7, but it is proving difficult to
navigate the bureaucracy. Vadim, you've built lmi on 64-bit msw as well
as GNU/Linux, so let me ask you whether the advantages of the latter
are great enough to justify fighting for GNU/Linux in the office.

We rarely rebuild wx or other libraries. Mostly we make incremental
changes to lmi's own source and the proprietary product files, and run
various (largely automated) tests. Therefore, it doesn't matter much
that GNU/Linux runs 'configure' faster than msw.

We depend on cygwin, and not just to run makefiles--Kim and I both do
a lot of work with *nix utilities in pipelines. I'm concerned that some
sophisticated cygwin users are reporting problems with msw-10, but our
office is not pushing people off msw-7 yet. And if cygwin's issues ever
become intractable, msw-10 promises "wsl", which might meet our needs.

Another possibility is osx, which is more enthusiastically supported by
the corporation than GNU/Linux. AIUI, that's just BSD plus eye candy,
so it ought to work for us...and Kim already uses osx at home.

On the other hand, if we pursue GNU/Linux, it seems that the best we'll
get is an account on a rhel server. I asked for a debian chroot,
thinking that would be best for everyone--we could have more freedom,
and the server would be more secure--but that suggestion seems to have
withered on the vine. If we follow this path, I fear that we'll be
stuck with obsolete tools like make-3.8x .

I should mention one other issue, which has been resolved. Earlier this
year, I considered building lmi only for gtk myself. If system-testing
results matched what Kim builds with msw, then we'd be assured that the
two builds are equivalent. But they don't match. However, results do
match perfectly between native msw and my cross builds in debian, so
that is no longer a problem.

All of that being said, I think we ought to keep using msw-7 in the
office, at least for now, and reevaluate our options when we're forced
to move to msw-10. Am I missing any compelling argument in favor of
pushing for GNU/Linux in the office sooner?