libc_hidden_proto breaks things for PowerPC64

[Steve Munroe]

On 08/19/02 09:05 PM, Alan Modra wrote:

> On Mon, Aug 19, 2002 at 09:28:55PM -0400, Roland McGrath wrote:
> > > Roland, what do you mean by "fix the compiler support", especially in
> > > regard to data definitions?
> >
> > With all the intermediate versions we had in this stuff I lost track of
the
> > fact that hidden_def is always done via asm.  I don't know why it's
that
> > way (Jakub?).  I think (with some gcc fixes) it can be done using
> > __attribute__ like the strong_alias macro is done.  What think would be
the
> > best overall solution is to make the __attribute__ approach work
properly
> > in gcc for all platforms, and then use that (and I don't care if we
just
> > punt the hidden_proto optimizations for broken compilers).  ...
>
> OK, barring evidence to the contrary, I think __attribute__ should
> work with the current powerpc64-linux toolchain.  ...

OK Alan can fix the ppc64 gcc. Alan, will you have a patch soon? Will it
get into the gcc3.2 branch?

But what about the other archs? I have not seen an answer to Roland's
question abut why hidden_def is always done via asm.

It seems that the situation for ppc64 and libc-symbols.h does not change
unless the mainline can use __attribute__ instead of asm.

> ... BTW, the hack that
> I added to gcc on 2002-03-22 for exporting .foo as well as foo in
> version scripts is no longer needed.  Same goes for glibc versions.awk
> dotsyms hack (but it may still be useful for AIX support, I'm not
> sure).

I'll save the patch for now but will remove it from my submission. I don't
need any extra controversy ;-)