Re: objcopy test failure on mips-linux

[Andreas Jaeger <aj at suse dot de>]

>>>>> Mark Mitchell writes:

>>>>> "Ralf" == Ralf Baechle <ralf@uni-koblenz.de> writes:
Ralf> On Wed, Mar 15, 2000 at 11:36:15AM +0100, Andreas Jaeger
Ralf> wrote:

>>> This has been introduced by Mark Mitchel in revision 1.12 of
>>> the file: 1999-07-01 Mark Mitchell <mark@codesourcery.com>
>>> 
>>> * elf32-mips.c (_bfd_mips_elf_fake_sections): Use
>>> _bfd_elf_init_reloc_shdr to initialize rel_hdr2.
>>> 
>>> 
>>> Mark, can you comment on this, please?  This part doesn't look
>>> correct to me (read the whole thread for details, I can send it
>>> if you need it).  It breaks objcopy and creates empty rea.text
>>> sections :-(.

Ralf> Afair N32 and 64 ABI files use both kinds of relocations.

Ralf> (Should check the docs but they're about 10000km away ...)

Mark> I'm sorry -- way too much water under that bridge.  I don't remember
Mark> well enough any more.  If you think it's wrong, it may be wrong.

The problem is that with objcopy the text.rel section from the
original file ends in a text.rela section :-(.

Mark> The top-level problem, though, is that on IRIX6 we need both REL and
Mark> RELA relocation sections simultaneously -- as you say both N32 and
Mark> 64-bit objects use both at once.  I'm not surprised it creates empty
Mark> sections; as I recall there didn't seem any way to know up front
Mark> whether or not you needed the sections.

Ok, I understand.  Can't we add a check at write out time to only
write a e.g. REL section if the RELA section is empty?  Currently we
write an empty RELA section.

Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.rhein-neckar.de