Re: Hide __[u]{div,mod}di3

[Franz Sirl <Franz dot Sirl-kernel at lauterbach dot com>]

On Monday 27 January 2003 23:08, Andreas Schwab wrote:
> Franz Sirl <Franz.Sirl-kernel@lauterbach.com> writes:
> |> On Monday 27 January 2003 22:38, Andreas Schwab wrote:
> |> > Franz Sirl <Franz.Sirl-kernel@lauterbach.com> writes:
> |> > |> Though note that this _requires_ binutils 2.13.90.0.4+ or 2.13.1+,
> |> > |> otherwise there will be really strange interactions with gcc.
> |> >
> |> > In which way?
> |>
> |> Earlier binutils doesn't correctly link together 2 shared libs
> |> (libgcc.so and libc.so in this case) which have the same version of a
> |> symbol, but aren't both linktime references. The PPC solution was only
> |> possible after hjl fixed a few binutils bugs. The fixes went into
> |> 2.13.90.4 and 2.13.1. IIRC the best case was binutils refusing to link
> |> glibc with some strange error message, but I think it was also possible
> |> to get a corrupted libc.so.
>
> I don't think this is an issue since glibc is using -static-libgcc.

You are right, it was the other way round, gcc libgcc.so (and other apps that 
link against both libc.so and libgcc.so) wouldn't link anymore. Just read up 
the whole lengthy discussion on libc-alpha back in May or June I think. If 
still in doubt just try to build a shared gcc with old binutils :-).

I'm adding

%ifarch ppc
BuildPreReq: binutils >= 2.13.90.0.4
Conflicts: binutils < 2.13.90.0.4
%endif

to Jakubs glibc srpm for a reason ;-).

Franz.