Re: RFC: Add output_bfd to bfd_link_info

["H. J. Lu" <hjl at lucon dot org>]

On Fri, Apr 23, 2004 at 09:27:20PM +0930, Alan Modra wrote:
> On Thu, Apr 22, 2004 at 07:47:14PM -0700, H. J. Lu wrote:
> > There are a few places where bfd_link_info is modifed to include
> > output_bfd. If output_bfd is in bfd_link_info, it is not needed.
> 
> You're talking about the last param to elf_link_hash_traverse, aren't
> you?  The places I see that pass output_bfd there also need to pass
> other extra data, so adding output_bfd to bfd_link_info won't remove the
> need for a struct containing a bfd_link_info pointer and other data.

That is true. I always find it clearer for bfd_link_info to have both
input and output bfds.

> 
> I'm not against the patch, but where do you plan to use
> info->output_bfd?

In _bfd_generic_link_add_one_symbol. In _bfd_elf_fix_symbol_flags,
there are

     /* Unfortunately, ELF_LINK_NON_ELF is only correct if the symbol
         was first seen in a non-ELF file.  Fortunately, if the symbol
         was first seen in an ELF file, we're probably OK unless the
         symbol was defined in a non-ELF file.  Catch that case here.
         FIXME: We're still in trouble if the symbol was first seen in
         a dynamic object, and then later in a non-ELF regular object.  

We can fix it by checking if the input file has the same format as
the output by

	if (abfd->xvec->flavour != info->output_bfd->xvec->flavour)
	  bfd_mark_foreign_symbol (info->output_bfd, h);

For ELF, it will be

void
_bfd_elf_mark_foreign_symbol (bfd *abfd ATTRIBUTE_UNUSED,
                             struct bfd_link_hash_entry *hash)
{
  struct elf_link_hash_entry *h = (struct elf_link_hash_entry *) hash;
  h->elf_link_hash_flags |= ELF_LINK_NON_ELF;
}

For others, it will be a no op.

H.J.