PATCH: Set forced_local for hidden and internal symbols

[Carlos O'Donell]

On Fri, Jul 08, 2005 at 11:27:02AM -0700, H. J. Lu wrote:
> > 
> > It turns out that we need this patch to handle hidden symbols provided
> > by linker correctly. We should do it since non-weak hidden and internal
> > symbols must be resolved locally.
> > 
> > 
> 
> I think it should be OK to set forced_local for weak hidden symbols
> since it will be resolved to 0 if there is no local defintion.

I don't quite understand the issue, or the case that your current patch
changes. However, it doesn't fix that the linker provides the symbols
too late for the backend to do anything about the allocated dynamic
reloc.

The best patch so far has been to move the provide earlier.

Failing that I am forced to use a relaxation handler to save space by
remove the relocs.

c.