Re: bad sh linking error with test case

[Alan Modra <amodra at bigpond dot net dot au>]

On Fri, Apr 05, 2002 at 01:22:30AM +0200, Ralf Corsepius wrote:
> Am Don, 2002-04-04 um 09.27 schrieb Alan Modra:
> > On Tue, Apr 02, 2002 at 08:26:12AM -0600, Joel Sherrill wrote:
> > > 
> > > for 2.12.  sh-rtems-ld is failing to fail when linking an
> > > executable that has undefined symbols.  As I reported earlier,
> > 
> > coff-sh.c:sh_coff_reloc_type_lookup is converting BFD_RELOC_32
> > to R_SH_IMM32CE.  coff-sh.c:sh_relocate_section isn't prepared
> > to deal with this reloc type unless COFF_WITH_PE is defined.
> > 
> > I'd say sh_coff_reloc_type_lookup needs another table for
> > non COFF_WITH_PE.
> Is the patch below what you had in mind?

Yes, something like that, but I see R_SH_IMAGEBASE is a PE only
reloc too.  Don't ask me what the RVA reloc should map to, I
don't know COFF very well at all.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre