This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: egcs-1.1b on sco-3.2v5.0.4 gcc/pic/libgcc.a with mixed COFF and ELF files?
- To: jason at cygnus dot com (Jason Merrill)
- Subject: Re: egcs-1.1b on sco-3.2v5.0.4 gcc/pic/libgcc.a with mixed COFF and ELF files?
- From: hjl at lucon dot org (H.J. Lu)
- Date: Thu, 10 Sep 1998 00:07:00 -0700 (PDT)
- Cc: egcs at cygnus dot com
>
> >>>>> H J Lu <hjl@lucon.org> writes:
>
> >> The Makefiles go to a considerable amount of effort to be sure that
> >> libgcc1 components are _not_ built by the compiler being built. In the
> >> OpenServer case with native cc, t his means libgcc1 is COFF. This sure
> >> looks intentional to me.
>
> > FWIW, guess what native compiler is used to compile libgcc1
> > Linux/x86.
>
> No compiler. From config/t-linux:
>
> # Do not build libgcc1. Let gcc generate those functions. The GNU/Linux
> # C library can handle them.
> LIBGCC1 =
> CROSS_LIBGCC1 =
> LIBGCC1_TEST =
>
Well, that is kind of misleading. The comment doesn't apply to
Linux/x86. In fact, since gcc 2.0, libgcc1 is no longer needed
for x86. The only reason I left it there for gcc 2.0 was at
that time, there were still libc binaries compiled with gcc
1.3x. If I took it out, those libc binaries would have to be
recompiled. But I might not want to compile libc with gcc 2.0.
You can use gcc to compile libgcc1 on any x86 hosts. You will
see no undefined symbols in libgcc1.a, maybe except for
_GLOBAL_OFFSET_TABLE_.
--
H.J. Lu (hjl@gnu.org)