Problem with recent GNU ld packages

[Edouard Gomez]

Edouard Gomez (ed.gomez@free.fr) wrote:
> I     tried    to    debug     GNU    ld     to    find     where    the
> (section_type_struct)->_raw_size field was set...  but no luck. Is there
> in this list a GNU ld (coff) guru wanting to help me track the the bug ?

I found the change that is responsible of my bug.

2003-04-15  Brian Ford  <ford@vss.fsi.com>
                                                                                
    * peicode.h (coff_swap_scnhdr_in): If a section holds
    uninitialized data and is from an object file or from an
    executable image that has not initialized the s_size field, or if
    the physical size is padded, use the virtual size (stored in
    s_paddr) instead.

It concerns code in bfd/peicode.h, included in a #ifndef
COFF_NO_HACK_SCNHDR_SIZE preprocessor directive.

In my case, i have:
scnhdr_int->s_name = ".text\0";
scnhdr_int->s_vaddr = 0;
scnhdr_int->s_paddr = 12;
scnhdr_int->s_size = 243;
scnhdr_int->s_scnptr = 112;
scnhdr_int->s_relptr = 355;
scnhdr_int->s_lnnoptr = 0;
scnhdr_int->s_flags = 0x60500020

The "guilty" condition is that one:
  if (scnhdr_int->s_paddr > 0
      && (((scnhdr_int->s_flags & IMAGE_SCN_CNT_UNINITIALIZED_DATA) != 0
          && (! bfd_pe_executable_p (abfd) || scnhdr_int->s_size == 0))
          || scnhdr_int->s_size > scnhdr_int->s_paddr))
    {
      scnhdr_int->s_size = scnhdr_int->s_paddr;
    }

Old code used to be:
  if ((scnhdr_int->s_flags & IMAGE_SCN_CNT_UNINITIALIZED_DATA) != 0
      && (scnhdr_int->s_paddr > 0))
    {
      scnhdr_int->s_size = scnhdr_int->s_paddr;
    }

I hope this enough so a binutils hacker can fix the bug, or if it's not
a binutils bug, just tell me that nasm generates bad pe-i386 object
files.

-- 
Edouard Gomez