Why is "ld -r" not good?

[Sam Ravnborg]

Hi Alan.

Searching for somethign else I stumbled over
following comment from you:

======================================================================
powerpc64-ld: section exceeds stub group size

    * From: Alan Modra <amodra@bigpond.net.au>
    * To: binutils@sourceware.org
    * Date: Tue, 7 Nov 2006 14:04:20 +1030
    * Subject: powerpc64-ld: section exceeds stub group size

There have been rather a lot of complaints about the warnings emitted
by the linker when building a powerpc64 linux kernel.  Up until now
I've resisted doing anything about them, because the warnings proved
useful in tracking down problems with large apps.  Also, I thought
they were avoidable if the kernel people would see the light and not
use "ld -r" as a means of packaging object files.  Well, it seems that
I was wrong.  "ld -r" wasn't the real cause.
======================================================================

Until now I have thought the "ld -r" was a good way to do what
we do in the kernel.

So I wonder what you have in mind here that can improve the
kernel build?

PS. Sorry for quoting you on such an old mail - I'm just curious.

	Sam - kernel build system maintainer