Mail Archives: djgpp-workers/2000/06/07/11:04:35
-------- Original Message --------
Subject: Re: GCC and system headers
Date: Wed, 7 Jun 2000 00:11:22 +0200
From: Marc Espie <espie AT quatramaran DOT ens DOT fr>
Organization: Ecole Normale Superieure (quatramaran)
To: lauras AT softhome DOT net
CC: gcc AT gcc DOT gnu DOT org
In article <393D3934 DOT 8AB76C44 AT softhome DOT net> you write:
>Hello,
>
>could somebody explain what are the technical reasons for
>not allowing ports to choose if they want to use standard
>headers provided by GCC? Many times I've seen on this
>mailing list that 'overriding USER_H is a brain-damaged
>feature' without explanation *how* it is brain-damaged.
>
>Thanks in advance,
>Laurynas Biveinis
>
It's standard policy from the gcc development team. Using USER_H means
not using gcc headers, and overriding them with the system headers.
Since the FSF goal is to promote gcc as a standard portable compiler,
USER_H goes against this goal. Also, they would rather support fixincludes,
and not be bothered with system inadequacies that USER_H is likely to
stumble onto.
Of course, other groups (the OpenBSD project for instance) which use gcc
as a system compiler have rather conflicting goals. We would rather *not*
use fixinc, as it tends to produce a mess of modified headers with
#include_next and what not which is not... too clean, compared to what we
would like our system headers to hold.
I would say that it's unlikely either camp is going to budge, since both
are utterly convinced they're doing the right thing.
As a token offer, I'm trying to ensure the FSF way does work correctly on
OpenBSD, though recent snapshots got me stumped on a varargs problems
which I haven't yet figured out...
- Raw text -