[OpenAFS-devel] [PATCH] handle nesting of vars in configure.in properly

[Ken Raeburn]

Jeremy Katz <katzj@redhat.com> writes:
> On Tuesday, October 09 2001, Derrick J Brashear said:
>> Is there some document which defines how autoconf-using packages should
>> behave? 
> Other than the fact that every Makefile generated by default by
> automake/autoconf is this way?  No, not off the top of my head.  

Would you consider the GNU coding standards (as they describe
Makefiles) relevant?  After all, that's sort of what autoconf and
automake are primarily targeted for.


Derek Atkins <warlord@MIT.EDU> writes:
> If you use DESTDIR, as Derrick and Sam have both suggested (and is
> really the FSF defined mechanism to modify the install-root) then you
> don't have to worry about this problem, and you DO get the semantics
> you want.

Are you sure?

$ ./configure --prefix=/afs/mycell.com/foo
$ make all
$ make install prefix=/afs/.mycell.com/foo

What value would you give for DESTDIR?

You *can* get the same effect with DESTDIR ... if you set up some
symlinks, or install to a temporary directory that has a
"afs/mycell.com/foo" subtree.  The "prefix=" version is a lot simpler.

Unfortunately, despite my preference for using prefix, I have to agree
that DESTDIR is the more robust approach.  I could argue that packages
that build things during "make install" right after "make all" are
broken, but they exist, and installing into a different directory is
uncommon enough that many developers would never notice if they get it
wrong.

(And, for whatever it might be worth, GCS does mention optional use of
DESTDIR in install rules.)