Xref: utzoo comp.sys.apollo:4832 gnu.gcc:1488
Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!snorkelwacker!mit-eddie!uw-beaver!Teknowledge.COM!unix!hplabs!hp-sdd!apollo!vasta
From: vasta@apollo.HP.COM (John Vasta)
Newsgroups: comp.sys.apollo,gnu.gcc
Subject: Re: Gcc on apollo
Keywords: gcc, apollo
Message-ID: <497390fe.20b6d@apollo.HP.COM>
Date: 27 Mar 90 22:07:00 GMT
References: <18645@boulder.Colorado.EDU> <1990Mar26.155315.29823@cns.umist.ac.uk>
Sender: root@apollo.HP.COM
Reply-To: vasta@apollo.HP.COM (John Vasta)
Followup-To: comp.sys.apollo
Organization: Hewlett-Packard Company, Apollo Division; Chelmsford, MA
Lines: 70

In article <1990Mar26.155315.29823@cns.umist.ac.uk> ran@cns.umist.ac.uk (Bob Nutter) writes:
>Installing gcc on apollos:
>========== === == ========
>
>This is how I went about installing gcc1.37 on our Apollo network
>(several 3000's and 4000's running sr10.1, and cc version 6.6). The
>stages below refer to the README file that come with John Vasta's
>(vasta@apollo.com) changes (see stage 1)
>
>Stage 1 (obtain the software):
>--------
>This is available from labrea.stanford.edu as /pub/gnu/APOLLO.README
>and apollo-gcc-1.37.tar.Z 

I believe that /pub/gnu/APOLLO.README only refers to Leonard Zubkoff's
Emacs port.

>You'll also need the gcc1.37 distribution, the gas1.34 distribution
>and bison.
>
>Stage 3 (building gas):
>-------
>When you try to `make' gas you'll get an error in coff_convert.c, as
>it can't find the include file <apollo/mir.h>. This is because the
>include file is in /usr/include. I put a link in to keep things neat
>from /usr/include/apollo/mir.h to /usr/include/mir.h: ie:
>
>	ln -s /usr/include/apollo/mir.h /usr/include/mir.h
>
>Once this is done, gas will install correctly. Then link `as' to the
>exectutable.

Oops - it looks like mir.h has moved into /usr/include/apollo recently;
I didn't realize it wasn't the same way "out there".

>Stage 5 (building gcc):
>-------
>This goes ok until you get to compiling `hard-params.c'.  It is due
>to finding `volatile' at line 638 (etc). This *should* be ok, as
>we're compiling with `-A nansi' which *should* undef __STDC__ and
>thus define volatile as `static' like this:
>
>	#ifndef __STDC__
>	#define volatile static
>	#endif
>
>However, this doesn't seem to work

hard-params.c is not compiled with -A nansi; the most recent release of
the C compiler (6.7) handles volatile, etc. If you're using an earlier
release, you can edit the gcc Makefile to set HARD_PARAMS_FLAGS = -A nansi

>Stage 6 (testing):
>-------
>The `make bootstrap CFLAGS=""' command (which tests gcc by getting
>it to compile itself) fails due to Domain/OS extensions to stdio.h.
>Basically, gcc doesn't like the `#attribute...' it finds early on.

I don't understand why everyone is having problems with this - my patch
kit comes with a new version of the grammar file, c-parse.y, which handles
this syntax. For some reason it isn't getting rebuilt; the timestamp in
the tar file must be older than the files in the gcc distribution. To
make sure it gets built, do "touch c-parse.y" after installing my changed
files.

Thanks for the assistance.
John Vasta                Hewlett-Packard Apollo Systems Division
vasta@apollo.hp.com       M.S. CHA-01-LT
(508) 256-6600 x6362      300 Apollo Drive, Chelmsford, MA 01824
UUCP: {decwrl!decvax, mit-eddie, attunix}!apollo!vasta