Re: AIX vtable issues

[David Edelsohn <dje at watson dot ibm dot com>]

>>>>> Mark Mitchell writes:

Mark> For example:

Mark> .toc
Mark> LC..0:
Mark> .tc _ZTT4MostIiE[TC],_ZTT4MostIiE[RW]

Mark> ...

Mark> .csect _test.rw_C[RW],3
Mark> .align 2
Mark> _ZTT4MostIiE:
Mark> .long _ZTV4MostIiE[RW]+12
Mark> .long _ZTC4MostIiE0_2D1IiE[RW]+12
Mark> .long _ZTC4MostIiE4_2D2IiE[RW]+12
Mark> .long _ZTV4MostIiE[RW]+24

Mark> This reults in _ZTT4MostIiE being undefined at link time.

Mark> What's going on here?  (I have the feeling you might have explained
Mark> this before, but I can't quite piece it together.)

	This may have to do with the following code in
rs6000.c:output_toc():

  /* Currently C++ toc references to vtables can be emitted before it
     is decided whether the vtable is public or private.  If this is
     the case, then the linker will eventually complain that there is
     a TOC reference to an unknown section.  Thus, for vtables only,
     we emit the TOC reference to reference the symbol and not the
     section.  */
  if (strncmp ("_vt.", name, 4) == 0)
    {
      RS6000_OUTPUT_BASENAME (file, name);
      if (offset < 0)
        fprintf (file, "%d", offset);
      else if (offset > 0)
        fprintf (file, "+%d", offset);
    }

I assume that vtables no longer have "_vt." in their name, so this code is
not being invoked.  This code needs to check for whatever additional
substring designates a virtual function table.

David