Re: Bug generating libc.so.lds

[Daniel Jacobowitz <drow at mvista dot com>]

On Tue, Apr 01, 2003 at 11:10:27AM +0200, Jakub Jelinek wrote:
> On Tue, Apr 01, 2003 at 12:42:33AM -0800, Roland McGrath wrote:
> > > Filtering out the -u __register_frame appears to solve the problem. 
> > > Does that sound right?
> > 
> > Um, that is probably ok.  The only reason I hesitate is that I am not sure
> > why it uses as complex a link command as it does to generate the ld script.
> > AFAICS this script doesn't vary and is in fact the same for all -shared
> > links.  In that case it doesn't matter to include all the target-specific
> > flags and LDLIBS and so forth at all.  In fact it could be generated just
> > once rather than done anew as part of every shared object link we do.
> > But I might be missing something.
> 
> ld -melf32ppclinux -shared --verbose
> vs.
> ld -melf64ppc -shared --verbose
> have different linker scripts, so some options surely have to be passed
> down to the linker besides -shared.

And also -z combreloc changes the linker script, et cetera.  Which is
in LDFLAGS, along with the problematic -u __register_frame, so we can't
just remove LDFLAGS.

However, it should not vary per library given the consistent options
between all libraries.  Doing it once ought to be enough...

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer