[Bug ld/4844] auto-import and const data import problem

[ralf dot habacker at freenet dot de]

------- Additional Comments From ralf dot habacker at freenet dot de  
2007-09-29 10:14 -------
Subject: Re:  auto-import and const data import problem

nickc at redhat dot com schrieb:
> ------- Additional Comments From nickc at redhat dot com  2007-09-28 11:57 
> -------
> Hi Ralf,
>
>   OK, please could you try out the uploaded patch.  There are several points
> that need to be looked at:
>
>   * The fix is only enabled for x86 PE ports.  Does it need to be enabled for
> other PE ports ?  (ARM, MIPS, SH, etc).
>   
I don't know. I can only speek for cygwin/mingw on win32.
>   * The new script is only used when --enable-auto-import is present on the
> linker command line, not when the linker magically enables auto import for
> itself.  (There appears to be code in pe{p}.em:...find_data_imports() that
> magically enables auto importing.  I was not sure what should be done in this
> situation).
>   
Really ? I found that data imports are ignored is auto-import is disabled.

static void
pe_find_data_imports (void)
{
  struct bfd_link_hash_entry *undef, *sym;

<<<<<<
  if (link_info.pei386_auto_import == 0)
 >>>>>>
    return;

...

}
>   * There are still some input sections that are mapped to the .rdata output
> section, even with the new script active:
>  .rdata BLOCK(__section_alignment__) :
>   {
>      *(.eh_frame)
>     ___RUNTIME_PSEUDO_RELOC_LIST__ = .;
>     __RUNTIME_PSEUDO_RELOC_LIST__ = .;
>     *(.rdata_runtime_pseudo_reloc)
>     ___RUNTIME_PSEUDO_RELOC_LIST_END__ = .;
>     __RUNTIME_PSEUDO_RELOC_LIST_END__ = .;
>   }
>
>   Should these input sections (and symbols) be moved to the .data section as 
> well ?
>   
yes, the related linker script moved those definitions to .data too.

The rdata section is completly empty

  .rdata BLOCK(__section_alignment__) :
  {
  }


> Most importantly, I have not checked the patched linker when used to link real
> applications.  I am not really set up here to do that, so I hope that you will
> be able to test it for me.
>   
I will try to test it. Unfortunally I cannot find a gcc4 binary for 
win32 on the web for mingw/cygwin. Both are still using 3.4.x. Do you 
know any additional resource ?

Cheers

Ralf



-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=4844

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.