Re: Only obj_coff_weak() with BFD_ASSEMBLERs

[Nick Clifton <nickc at redhat dot com>]

Hi Aaron,

> On a related note, I need help with getting relocations to work with
> COFF weak symbols.  When generating relocations, GAS needs to treat weak
> symbols as undefined symbols, not as defined symbols, which is what it
> presently does.  It gets it halfway right, and knows to generate
> relocations for weak symbols where a defined symbol wouldn't need one,
> but the fixup to the frag and the relocation itself don't create correct
> results when linked.  All of the obvious places to add this sort of
> logic are places I am very hesitant to add changes, such as in
> bfd_perform_relocations.

Changing and debugging reloc processing in BFD is notoriously difficult. 
 It is also well known that bfd_perform_relocations() often fails to do 
what is desired so a lot of backends choose to perform their own reloc 
processing.

For simplicity sake this is what I suggest you do for now.  Look at the 
backend specific file for the target you are interested in (coff-i386.c 
??) and try adding to or defining its coff_relocate_section function.

Cheers
  Nick