Document empty structures

[Richard Earnshaw]

> 
> On Monday, April 14, 2003, at 11:20  AM, Mark Mitchell wrote:
> 
> > On Mon, 2003-04-14 at 11:17, Richard Earnshaw wrote:
> >>
> >>>> I'm pretty sure that the C++ language standard does not say that 
> >>>> empty
> >>>> structures have size 1 -- AFAIK it just says that they have 
> >>>> non-zero size
> >>>> (C++98 section 9, paragraph 3).
> >>>
> >>> Correct.  The C++ ABI says they have size 1.
> >>
> >> Even on a machine where STRUCTURE_SIZE_BOUNDARY != 8?  I would have
> >> thought STRUCTURE_SIZE_BOUNDARY / BITS_PER_UNIT was a more sensible
> >> definition.
> >
> > You are even more correct.  I forget what the exact calculation is, but
> > it does indeed take this kind of thing into account.
> 
> Could someone provide a patch?

Not a patch, but it seems that this is the bit of code that handles this 
case:

cp/class.c:

  /* Make sure not to create any structures with zero size.  */
  if (integer_zerop (rli_size_unit_so_far (rli)) && CLASSTYPE_EMPTY_P (t))
    place_field (rli, 
                 build_decl (FIELD_DECL, NULL_TREE, char_type_node));

so we behave as though the class contained a single anonymous char member.

R.