Patch for R_ARM_THM_PC22 relocs

[Momchil Velikov]

Alan Modra wrote:
> 
> On Fri, 29 Sep 2000, Momchil Velikov wrote:
> 
> > +      upper_insn = bfd_get_16 (abfd, address);
> > +      lower_insn = bfd_get_16 (abfd, address + 2);
> 
> Sorry, I should have picked this up on your previous patch, but this is
> wrong too.  You're assuming big-endian, which isn't always the case for
> arm-elf.  Please use bfd_get_32 and bfd_put_32.

I don't think so. Here is what the ARM manual has to say:
    ...
    In the first instruction the Offset field contains the upper 11 bits
of the target address.
    ...
    In the second instruction the Offset field contains an 11-bit
representation lower half of
    the target address.  
    ...

So, when using bfd_get_32, depending on the host and target endianess,
you can end up with the upper bits of the addendum either in
the upper or the lower bits of the `addend'. 

Regards,
-velco