[Bug c++/98423] The defaulted default constructor defined as deleted when one of variant member has a default member initializer

["jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98423

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |jason at gcc dot gnu.org,
                   |                            |mpolacek at gcc dot gnu.org

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I think it isn't solely about the process_subob_fn
      if (TREE_CODE (arg) == FIELD_DECL
          && TREE_CODE (DECL_CONTEXT (arg)) == UNION_TYPE)
        {
          if (deleted_p)
            *deleted_p = true;
          if (diag)
            error ("union member %q+D with non-trivial %qD", arg, fn);
        }
diagnostics that probably shouldn't be performed if the union type has nsdmi on
any member, isn't it also the:
  if (constexpr_p && !DECL_DECLARED_CONSTEXPR_P (fn))
    {
      *constexpr_p = false;
      if (diag)
        {
          inform (DECL_SOURCE_LOCATION (fn),
                  SFK_DTOR_P (sfk)
                  ? G_("defaulted destructor calls non-%<constexpr%> %qD")
                  : G_("defaulted constructor calls non-%<constexpr%> %qD"),
                  fn);
          explain_invalid_constexpr_fn (fn);
        }
    }
diagnostics (why would a default constructor of a union be non-constexpr just
because some variant member without nsdmi has non-constexpr default
constructor), perhaps the spec_p stuff too.
So for UNION_TYPE we probably want to search for nsdmis in another loop at the
start of walk_field_subobs and punt some tests for some special members if
there are any nsdmis.