From mboxrd@z Thu Jan  1 00:00:00 1970
From: Ian Lance Taylor <ian@cygnus.com>
To: Robert.Wilhelm@Physik.TU-Muenchen.DE
Cc: hjl@innovix.com, gcc2@cygnus.com, raeburn@cygnus.com, gas2@cygnus.com, rwilhelm@Physik.TU-Muenchen.DE
Subject: Re: Bugs in gcc or binutils 2.7?
Date: Thu, 18 Jul 1996 12:46:00 -0000
Message-id: <199607181946.PAA01041@sanguine.cygnus.com>
References: <9607181742.AA01549@nugget.cip.physik.tu-muenchen.de>
X-SW-Source: 1996/msg00067.html

   From: Robert.Wilhelm@Physik.TU-Muenchen.DE
   Date: Thu, 18 Jul 1996 19:42:13 +0200 (MET DST)

   > The assembler has issued this warning for over a year, including the
   > 2.6 release.  I recently (July 5) changed the warning into an error.
   > Do you receive a warning when you use an earlier version of the
   > binutils?

   gas 2.5.2linux   no message
   gas 2.6.0.2      warning
   gas 2.7          error

   BTW: The current gcc snapshot (ss-960707) produces the same asm-output as
   gcc 2.7.2.

This program works correctly when compiled by gcc.  It does not try to
define foo as both weak and global.

extern int foo () __attribute__ ((weak));
foo ()
{
}

I'm inclined to think that if gcc wants to handle #pragma weak, it
should do it the same way as __attribute__ ((weak)).  I think the
current assembler behaviour is reasonable, although I can change it if
necessary.

Ian