Re: revised patch for m32r gas relocations

[Doug Evans <dje at transmeta dot com>]

Donald Lindsay writes:
 > > 
 > > Also, methinks I would like to see the rewrite of m32r_elf_lo16_reloc
 > > undone now that there is m32r_elf_generic_reloc.
 > 
 > The logic certainly could be merged, but I'd rather do that as a separate
 > patch, after we agree that this patch works (or agree on how to make it
 > work).

There's no incentive to do the cleanup once the patch goes in,
and since undoing the lo16 rewrite is TRT, it should be part
of this patch.

The cleanup is trivial: go back to the way things were but
instead of calling bfd_elf_generic_reloc, call m32r_elf_generic_reloc
instead.  I'm not talking about spending more than 2 minutes
(plus the time to rerun the tests - but you have to do that anyway to
fix the current problems with m32r_elf_generic_reloc).

m32r_elf_lo16_reloc use to end with this:

  /* Now do the LO16 reloc in the usual way.  */
  return bfd_elf_generic_reloc (abfd, reloc_entry, symbol, data,
				input_section, output_bfd, error_message);

Change that to:

  /* Now do the LO16 reloc in the usual way.
     ??? It would be nice to call bfd_elf_generic_reloc here,
     but we have partial_inplace == TRUE.  bfd_elf_generic_reloc will
     pass the handling back to bfd_install_relocation which will install
     a section relative addend which is wrong.  */
  return m32r_elf_generic_reloc (input_bfd, reloc_entry, symbol, data,
				 input_section, output_bfd, error_message);

[keeping the renaming of abfd to input_bfd]

If the cleanup doesn't go in now, then when doing the cleanup one
might want to run the testsuite AGAIN.  It's _easier_ doing it now.