patch that caused regression PR c++/8913

[Janis Johnson]

The regression reported in PR c++/8913 showed up starting with this
patch, which might merely uncover an existing problem:

gcc:

>2002-12-04  Jason Merrill  <jason@redhat.com>
>
>	PR c++/8461, c++/8625
>	* integrate.c (copy_decl_for_inlining): Handle explicit invisible
>	references.
>	* tree-inline.c (initialize_inlined_parameters): Likewise.
>
>	* tree.c (variably_modified_type_p): Just return an error_mark_node.

gcc/cp:

>2002-12-03  Jason Merrill  <jason@redhat.com>
>
>	PR c++/8461, c++/8625
>	* call.c (convert_for_arg_passing): Don't mess with error_mark_node.
>	(cp_convert_parm_for_inlining): Remove.
>	* cp-lang.c (LANG_HOOKS_TREE_INLINING_CONVERT_PARM_FOR_INLINING): 
>	Remove.
>	* cp-tree.h (ADDR_IS_INVISIREF): Remove.
>	* except.c (stabilize_throw_expr): Remove ADDR_IS_INVISIREF code.

Here's a small test case that causes the ICE, which occurs in the
mainline:

---------------
template <class T> struct Ref { 
  Ref(int *);
  ~Ref();
};

struct User{
  User (Ref<int> obj) {};
};

void test() {
  User a (0);
}
---------------

Output from the compiler:

8913.C: In function `void test()':
8913.C:7: internal compiler error: in simplify_gen_subreg, at simplify-rtx.c:
   2673
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.

This information will also be added to the PR.

I'm working on a script to automate the search for regressions down to
the hour; the rest happens manually.

Janis