alpha brsgp relocation lossage

[Alan Modra]

On Wed, Nov 06, 2002 at 04:41:51PM -0800, Richard Henderson wrote:
> What I do know is that by moving code from one place to another, we
> change the order in which this frobbing is done, and this new order
> happens to work for this test case.

What you've done is reduce the reloc to one against a section sym.
Section syms, since they have value zero, avoid the need for hacks
like the "Ohhh, this is ugly" one at the end of tc_gen_reloc.

> + 	if (! S_IS_DEFINED (sym) || S_FORCE_RELOC (sym))
> + 	  return 0;

! S_IS_DEFINED is redundant.  See S_FORCE_RELOC.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre