Re: [PATCH] x86-64: support newer relocation types

[Jan Hubicka <jh at suse dot cz>]

> On Sun, Jun 12, 2005 at 02:28:25PM +0200, Jan Hubicka wrote:
> > 
> > > +	  /* Note that sgot is not involved in this
> > > +	     calculation.  We always want the start of .got.plt.  If we
> > > +	     defined _GLOBAL_OFFSET_TABLE_ in a different way, as is
> > > +	     permitted by the ABI, we might have to change this
> > > +	     calculation.  */
> > > +	  relocation -= htab->sgotplt->output_section->vma
> > > +			+ htab->sgotplt->output_offset;
> > > +	  break;
> > > +
> > > +	case R_X86_64_GOTPC32:
> > > +	  /* Use global offset table as symbol value.  */
> > > +	  relocation = htab->sgotplt->output_section->vma
> > > +		       + htab->sgotplt->output_offset;
> > > +	  unresolved_reloc = FALSE;
> > > +	  break;
> > 
> > 
> > There is another difference relative to my implementation:
> > 
> > + 	  /* Note that sgot->output_offset is not involved in this
> > + 	     calculation.  We always want the start of .got.  If we
> > + 	     defined _GLOBAL_OFFSET_TABLE in a different way, as is
> > + 	     permitted by the ABI, we might have to change this
> > + 	     calculation.  */
> > + 	  relocation -= htab->sgot->output_section->vma;
> > + 	  break;
> > + 
> > + 	case R_X86_64_GOTPC32:
> > + 	  /* Use global offset table as symbol value.  */
> > + 	  relocation = htab->sgot->output_section->vma;
> > + 	  unresolved_reloc = FALSE;
> > + 	  break;
> > + 
> > 
> Jan's version is the same as i386. I believe those relocations are
> the same between i386 and x86_64.

Yeah, I just copied it around and it seemed to fit, but I am not quite
sure whether REL/RELA difference does not play important role here...

Honza
> 
> 
> 
> H.J.