This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: FreeBSD 4.0
- To: espie at quatramaran dot ens dot fr, law at cygnus dot com
- Subject: Re: FreeBSD 4.0
- From: Mike Stump <mrs at wrs dot com>
- Date: Thu, 16 Sep 1999 13:13:22 -0700 (PDT)
- Cc: egcs at egcs dot cygnus dot com
We have seen this same issue a couple of times. HJ wanted to put
--prefix=/usr in the compiler for linux systems, we told him to take a
leap. It is wrong, plain wrong. Only two people in the world should
do that, the end user that wants to clobber his system, or that
distribution manager for /usr.
You are in the exact same case. You want to put things that are only
relevant to your one build, but screw all the other folks in the world.
gcc doesn't cater to people like you. It caters to all the other
folks that want to install a new compiler in a non-conflicting way on
their system.
Once you readjust your philosophy around to match the gcc philosophy,
you will see why what law says is true.
> Date: Thu, 16 Sep 1999 16:41:15 +0200
> From: Marc Espie <espie@quatramaran.ens.fr>
> I concur that, if the end user want to fiddle with a newer gcc snapshot,
> USER_H is in the way.
See, this is wrong. Use the phrase, if the end user wants to fiddle
with a newer gcc snapshot, everything will always just work. If the
distribution packager wants to build up a distribution, they will have
to fiddle with USER_H...
It would be nice if the internal documentation covered this some...
It would be nice if --enable-distributorbuild or somesuch could make
your life easier. Then we serve all folks (good in this case I
think). Then HJ could put the defaulting of prefix to /usr back in.