[RFC] Providing init_fini_syms earlier?

[Jakub Jelinek]

On Mon, Jul 11, 2005 at 02:06:55PM -0400, Carlos O'Donell wrote:
> On Mon, Jul 11, 2005 at 10:47:41AM -0700, H. J. Lu wrote:
> > On Mon, Jul 11, 2005 at 10:51:07AM -0400, Carlos O'Donell wrote:
> > > On Mon, Jul 11, 2005 at 07:45:00AM -0700, H. J. Lu wrote:
> > > > That was a glibc bug. I posted a patch for the linker bug.
> > > 
> > > There may have been a linker bug, but it doesn't solve the problem.
> > > The symbol provide by the linker is still too late for the backend to
> > > do anything about the extra reloc space.
> > 
> > Please try the linker patch with the FIXED glibc.
> 
> Could you provide a reference to your libc patch please?
> 
> I know you recently posted:
> 2005-04-13  H.J. Lu  <hongjiu.lu@intel.com>
>         [BZ #974]
> 	...
> 
> Though it doesn't look like that patch would change the definitions of
> the symbols.
> 
> I'm a bit confused because it also looks like the definition in 2.3.2
> were marked hidden.

That's actually a fedora*-branch local change:
http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/csu/elf-init.c.diff?r1=1.3&r2=1.3.2.1&cvsroot=glibc
ld should work correctly with or without that, what really matters
is if the definitions of the symbols are hidden or not (and they are
supposed to be hidden).

	Jakub