From owner-freebsd-hackers  Fri Jun 23 08:17:06 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id IAA10190
          for hackers-outgoing; Fri, 23 Jun 1995 08:17:06 -0700
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id IAA10179
          for <hackers@FreeBSD.org>; Fri, 23 Jun 1995 08:16:54 -0700
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id BAA13731; Sat, 24 Jun 1995 01:02:33 +1000
Date: Sat, 24 Jun 1995 01:02:33 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199506231502.BAA13731@godzilla.zeta.org.au>
To: amjudge@dsg.cs.tcd.ie, hackers@FreeBSD.org
Subject: Re: gcc 2.7.0 and FreeBSD 2.0.5
Sender: hackers-owner@FreeBSD.org
Precedence: bulk

>However, when I try to compile a simple C++ program, the
>assembler complains about the use of the .weak pseudo-op.  I thought
>this might just be a limitation in the assembler, so I hacked up a
>freebsd config for binutils-2.5.2 and compiled up the assembler there.
>This got me past the .weak problem and even works for simple programs.

I removed ASM_WEAKEN_LABEL() from gcc-2.7.0/config/i386/freebsd.h.
SUpport for weak symbols seems to be optional (it is only provided
for i386/osfrose, m88k, netbsd and svr4).

>However, when I try with a more complex program (basically any
>multimodule program where a .weak symbol appears more than once, I
>think), I get errors of the form:
>ld: /var/tmp/cc0019921.o: unexpected multiple definitions of symbol `_f__1X', type 0xf
>:1: Definition of symbol `_f__1X' (multiply defined)

>I guess that the FreeBSD ld doesn't fully support weak symbols?

Weak symbols have stab numbers N_WEAKU=0xd to N_WEAKB=0x11 (see
binutils/gas/config/aout_gnu.h).  Old ld's don't support these.

Bruce