HPPA/SOM Patch
Ian Lance Taylor
ian@airs.com
Thu Jul 1 00:00:00 GMT 1999
Date: Fri, 04 Jun 1999 13:08:43 -0700
From: Mark Klein <mklein@dis.com>
>If so, why is hppa_fix_adjustable being called after set_symtab?
It is part of my earlier fix. There are certain fixups that don't
have a symbol associated with them. In that case, the original
code referenced a dummy symbol. I changed that to the symbol_get_bfdsym()
call below.
*** 2886,2892 ****
case R_ENTRY:
case R_EXIT:
/* There is no symbol associated with these fixups. */
! relocs[i]->sym_ptr_ptr = &dummy_symbol->bsym;
relocs[i]->addend = fixp->fx_offset;
break;
--- 2893,2900 ----
case R_ENTRY:
case R_EXIT:
/* There is no symbol associated with these fixups. */
! relocs[i]->sym_ptr_ptr = (asymbol **) xmalloc (sizeof (asymbol *));
! *relocs[i]->sym_ptr_ptr = symbol_get_bfdsym (dummy_symbol);
relocs[i]->addend = fixp->fx_offset;
break;
That in turn is causing tc_gen_reloc() to be invoked within write_relocs()
and voila!
I don't quite follow that, actually. I don't see why calling
symbol_get_bfdsym is causing tc_gen_reloc to get called. But I'll
take your word for it.
So, did I make a wrong assumption here, or should set_symtab be delayed
until after the bfd_map_over_sections (stdoutput, write_relocs, (char *) 0) ?
No, I believe set_symtab must be called before the relocations are
written out.
One way to avoid it would be to call symbol_get_bfdsym immediately
after creating dummy_symbol, to force it to be converted to a real
symbol before set_symtab is called.
Another way would be not use dummy_symbol for this at all, and instead
call bfd_create_empty_symbol directly to get something you can set the
reloc sym_ptr_ptr field to point to.
Ian
More information about the Binutils
mailing list