This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Shared library annoyance with gcc-3_0-branch
Richard Henderson <rth@redhat.com> writes:
> On Sun, Feb 18, 2001 at 12:51:29PM -0800, H . J . Lu wrote:
> > If I fold libgcc in gcc 3.0 into glibc tody and change the gcc source
> > to use it instead its own, can I assume it will work with all future
> > versions of gcc from gcc 3.0?
>
> No. We should be free to extend it in backward compatible ways.
I don't think anyone will contest this. The fear seems to be breaking
backward binary compatibility.
> I think this debate is pointless, since neither of us are going
> to change our minds. I think the gcc support routines should be
> in its own library and you think they should be incorporated into
> glibc. I think that we should treat all systems the same for
> uniformity and you think Linux should be special cased.
The question seems more to be the case how do you get libgcc_s.so to
be installed in /lib (or at least where libc lives). You need this
for every unix, otherwise any program you compile with gcc and use at
the wrong moment during bootup will fail.
Moving this stuff into glibc is an interesting idea, but unnecessary.
The scare is that gcc-3.0 which is supposed to finalize the C++ ABI,
will result in the opposite happening. If multiple separate vendors
that rely on gcc as their primary c compiler, take different tactics
the binary situation will not be improved.
A shared libgcc looks necessary to resolve problems of it's being
linked in multiple times by a program. But for a shared libgcc to
work reliably it must be in /lib. So we need gcc's make install to
put it there with an appropriate version number.
Eric