Re: sparc testsuite failures

[Christian Joensson <christian at j-son dot org>]

Alan's suggested patch makes the testsuite problem for PR 48 go away :)

Cheers,

/ChJ



On Thu, Apr 08, 2004 at 04:21:35PM +0200, Christian Joensson wrote:
> Is something similar needed in elf64-sparc.c at line no 2412?
> 
> Cheers,
> 
> /ChJ
> 
> BTW, testing the below suggested patch right now on sparc-linux.
> 
> 
> On Thu, Apr 08, 2004 at 01:41:42AM +0930, Alan Modra wrote:
> > Hi Jakub,
> >   There are a number of open binutils bug reports for sparc.  This patch
> > should fix PR 48.
> > 
> > 	* elf32-sparc.c (elf32_sparc_relocate_section): Don't abort
> > 	when statically linking PIC code.
> > 
> > Testing h->plt.offset isn't sufficient because when no dynamic objects
> > are involved in the link, h->plt.offset is still a refcount.
> > size_dynamic_sections isn't run.
> > 
> > PR 49 and PR 50 look like they are a result of the sparc backend not
> > being updated to use SYMBOL_CALLS_LOCAL.  Do you plan to fix this
> > before 2.15?
> > 
> > Index: bfd/elf32-sparc.c
> > ===================================================================
> > RCS file: /cvs/src/src/bfd/elf32-sparc.c,v
> > retrieving revision 1.65
> > diff -u -p -r1.65 elf32-sparc.c
> > --- bfd/elf32-sparc.c	27 Mar 2004 10:58:07 -0000	1.65
> > +++ bfd/elf32-sparc.c	7 Apr 2004 15:08:05 -0000
> > @@ -2323,16 +2323,13 @@ elf32_sparc_relocate_section (output_bfd
> >  	  if (h == NULL)
> >  	    break;
> >  
> > -	  if (h->plt.offset == (bfd_vma) -1)
> > +	  if (h->plt.offset == (bfd_vma) -1 || htab->splt == NULL)
> >  	    {
> >  	      /* We didn't make a PLT entry for this symbol.  This
> >  		 happens when statically linking PIC code, or when
> >  		 using -Bsymbolic.  */
> >  	      break;
> >  	    }
> > -
> > -	  if (htab->splt == NULL)
> > -	    abort ();
> >  
> >  	  relocation = (htab->splt->output_section->vma
> >  			+ htab->splt->output_offset
> > 
> > 
> > -- 
> > Alan Modra
> > IBM OzLabs - Linux Technology Centre