Re: GOLD: RFA: Add support for RX target

[Nick Clifton <nickc at redhat dot com>]

Hi Ian,

  Thanks for the patch review.  Attached is a revised version which 
addresses all of your comments except one:


> > +       // This is a copy of the relocate_section() function in target-reloc.h,
> > +       // except with the additional calls to convert_host to fix up the
> > +       // extracted information.
> > +       // FIXME: There ought to be a better way to do this.
>
> Yes, there ought to be a better way to do this.  If your code is
> correct, then it looks like the problem is that you have a little-endian
> object but the reloc fields are written with big-endian data.  The
> comment at the top of the file suggests that the symbol table is also
> written with big-endian data, in which case I think this code is
> insufficient.
>
> Anyhow, if the reloc fields are consistenly written big-endian, then the
> right approach is not to call convert_host, but instead to change
>
> > +       typedef Reloc_types<elfcpp::SHT_RELA, 32, false>::Reloc Reltype;
>
> to
>
> > +       typedef Reloc_types<elfcpp::SHT_RELA, 32, true>::Reloc Reltype;
>
> That will cause the reloc code to fetch values big-endian rather than
> little-endian.

I found it very hard to get the reloc code work correctly for the RX but 
the current version does that.  I did think about splitting up 
relocate_section in target-reloc.h but I did not want to interfere with 
already working code just for the vagaries of this one target.  (Plus I 
was not sure that I would not introduce new bugs this way.  That and the 
fact that big-endian data/little-endian code is a rare case for the RX).

So, if it is OK with you I would like to leave the code as it currently 
is.  It may be less efficient, but it does work.

Is this revised version OK ?

Cheers
  Nick

Attachment: rx.gold.patch.2
Description: Unix manual page