Re: A patch for glibc 2.0/2.1

[Mark Kettenis <kettenis@phys.uva.nl>]

   From: hjl@lucon.org (H.J. Lu)
   Date: Thu, 21 Jan 1999 12:10:44 -0800 (PST)
   Cc: libc-hacker@cygnus.com, drepper@cygnus.com

   > 
   > On Thu, 21 Jan 1999 11:42:19 -0800 (PST), H.J. Lu wrote:
   > >> >
   > >> >Oooops. Here is a patch for glibc 2.0 and 2.1. It will make
   > >> >those functions weak extern in libc.so. Ulrich, could you please
   > >> >check it in for both glibc 2.0 and 2.1? Thanks.
   > >
   > >On my second thought, Ulrich, please ignore this patch. We should
   > >keep them in libc.so so that noone else has to define them. That
   > >means libfoo.so made by
   > >
   > ># gcc -shared -o libfoo.so foo.cc -fPIC
   > >
   > >will work regardless if libstdc++.so is loaded or not.
   > 
   > Then you have to disallow compiling libc with gcc 2.8.  (I personally think
   > this would be a good idea anyway.)
   > 

   Can we add this glibc 2.1? That is we check for gcc 2.8 and disallow it
   for glibc 2.1.

Please do not do this.  egcs may be superior to gcc 2.8.1, but gcc
still is a good C compiler.  I use gcc 2.8.1 exclusively without
problems.  Moreover it still is the latest official release.  We
should find a solution that works for both egcs and gcc 2.8.1.

However I do not see a problem.  My shared libc compiled with gcc 2.8.1
includes `__deregister_frame_info' and `__register_frame_info'.  I
just downloaded the current unstable Debian groff package and the
groff binary (which seems to be compiled with egcs-1.1) runs without
problems on my system.  Are you guys sure that you're not seeing
problems because you still have some binaries/shared libs lying around
that were linked against some old glibc test release?

Mark