Re: System configuration during install

[Rusty Ballinger]

> David> If possible, I would like to support the widest range of
> David> systems possible with the constraint that I don't as yet
> David> understand how they vary.
>
> I think most people handle this by writing a document for their
> project that says "once `make install' is done, here is how you can
> set things up..." -- with instructions on a system-by-system basis (or
> even relatively generic, assuming the user understands his own
> system).

Yeh, but figuring out where a system puts its init scripts seems like an
excellent job for the configure script.  (And the autoconf macro archive
posted by Bjoern Pedersen might have a macro which handles all that.)

> FWIW I think Hari is right when he says that most people would be
> suprised if `make install' changed the boot-time configuration of
> their machine.

But I think this surprise comes only after experiencing the number of
packages which punt on those portions of the installation process, not
because "make install" means anything less than "install this package."
When you install an RPM package, certainly it's "acceptable" for the
package to print additional instructions which must be performed manually
to complete the installation, but I don't think this is ever the user's
first choice.  The same is true for the install target.

> RPM builders would be especially suprised -- they use `make install'
> while building RPMs.

That's one of the best reasons to not build RPM packages as root.

If your installation procedure has commands which don't involve the
installation of files which will be placed relative to DESTDIR (like
"killall -HUP inetd", "killall -USR1 xinetd", "ldconfig", etc.), you can
ignore unprivileged users' failure to execute those commands, or precede
those commands with 'test -n "$$DESTDIR" || ...' (although I'm not sure
there aren't cases when checking DESTDIR might be wrong).

--Rusty