From: Andrew Crabtree <andrewc AT typhoon DOT rose DOT hp DOT com>
Message-Id: <199802031917.AA077893469@typhoon.rose.hp.com>
Subject: Re: gcc 2.8.0/include files
To: robert DOT hoehne AT gmx DOT net, djgpp-workers AT delorie DOT com
Date: Tue, 03 Feb 1998 11:17:49 PST
Reply-To: andrewc AT rosemail DOT rose DOT hp DOT com
Precedence: bulk


> > /tmp/gnu/egcs-1.0.1/gcc/ginclude/stddef.h:168: warning: `size_t' previously decl ared here
> 
> These are warnings; they should not stop the compilation.  So why do
> you say that Make aborts?
This is just hand compiled, I don't know what exactly all gets included
and what the flags are in the actual make process but it does abort.

> > and thought that moving size_t into stddef.h might be a decent
> > alternative
> Our stddef.h *does* define size_t.
Right - I meant removing it from stdio.h as as well though.  

Anyways - Its a moot point now because I know how to disable 
gcc from using its own header files, which has been the preferred
solution all along.  

In the file $(gccdir)/config/i386/t-go32

add the line

USER_H = 

Also, comment out any other USER_H line if there are any.  Note that
this will disable all header files from being important.  The only
two ansi ones I think are stdarg.h and stddef.h.  All of the other
appear to be machine specific vararg files.  Its possible to specify 
the USER_H macro to just bring in the vararg stuff and nothing else 
I suppose.

I still have fixincludes running in my environment and it doesn't 
appear to harm things.  That can be disabled though by setting
fixincludes=Makefile.in (I think robert said he made a special 
dj-fixincludes script tho)  The only problem still remaining when
compiling is that the built in specs for djgpp don't work, and you
must manually copy a dj specs file over when xgcc is invoked.
Now, as DJ told me, we cannot fix all of the specs (some, like
DJ_MAJOR and MINOR are dependent on DJDEV releases, not on 
compiler releases), but perhaps we should fix the link spec to match
what it should be.

Andy