c++: Implement __is_{nothrow_,}convertible [PR106784]

  To improve compile times, the C++ library could use compiler built-ins
rather than implementing std::is_convertible (and _nothrow) as class
templates.  This patch adds the built-ins.  We already have
__is_constructible and __is_assignable, and the nothrow forms of those.

Microsoft (and clang, for compatibility) also provide an alias called
__is_convertible_to.  I did not add it, but it would be trivial to do
so.

I noticed that our __is_assignable doesn't implement the "Access checks
are performed as if from a context unrelated to either type" requirement,
therefore std::is_assignable / __is_assignable give two different results
here:

  class S {
    operator int();
    friend void g(); // #1
  };

  void
  g ()
  {
    // #1 doesn't matter
    static_assert(std::is_assignable<int&, S>::value, "");
    static_assert(__is_assignable(int&, S), "");
  }

This is not a problem if __is_assignable is not meant to be used by
the users.

This patch doesn't make libstdc++ use the new built-ins, but I had to
rename a class otherwise its name would clash with the new built-in.

Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?

	PR c++/106784

gcc/c-family/ChangeLog:

	* c-common.cc (c_common_reswords): Add __is_convertible and
	__is_nothrow_convertible.
	* c-common.h (enum rid): Add RID_IS_CONVERTIBLE and
	RID_IS_NOTHROW_CONVERTIBLE.

gcc/cp/ChangeLog:

	* constraint.cc (diagnose_trait_expr): Handle CPTK_IS_CONVERTIBLE
	and CPTK_IS_NOTHROW_CONVERTIBLE.
	* cp-objcp-common.cc (names_builtin_p): Handle RID_IS_CONVERTIBLE
	RID_IS_NOTHROW_CONVERTIBLE.
	* cp-tree.h (enum cp_trait_kind): Add CPTK_IS_CONVERTIBLE and
	CPTK_IS_NOTHROW_CONVERTIBLE.
	(is_convertible): Declare.
	(is_nothrow_convertible): Likewise.
	* cxx-pretty-print.cc (pp_cxx_trait_expression): Handle
	CPTK_IS_CONVERTIBLE and CPTK_IS_NOTHROW_CONVERTIBLE.
	* method.cc (is_convertible): New.
	(is_nothrow_convertible): Likewise.
	* parser.cc (cp_parser_primary_expression): Handle RID_IS_CONVERTIBLE
	and RID_IS_NOTHROW_CONVERTIBLE.
	(cp_parser_trait_expr): Likewise.
	* semantics.cc (trait_expr_value): Handle CPTK_IS_CONVERTIBLE and
	CPTK_IS_NOTHROW_CONVERTIBLE.
	(finish_trait_expr): Likewise.

libstdc++-v3/ChangeLog:

	* include/std/type_traits: Rename __is_nothrow_convertible to
	__is_nothrow_convertible_lib.
	* testsuite/20_util/is_nothrow_convertible/value_ext.cc: Likewise.

gcc/testsuite/ChangeLog:

	* g++.dg/ext/has-builtin-1.C: Enhance to test __is_convertible and
	__is_nothrow_convertible.
	* g++.dg/ext/is_convertible1.C: New test.
	* g++.dg/ext/is_convertible2.C: New test.
	* g++.dg/ext/is_nothrow_convertible1.C: New test.
	* g++.dg/ext/is_nothrow_convertible2.C: New test.
---
 gcc/c-family/c-common.cc                      |   2 +
 gcc/c-family/c-common.h                       |   1 +
 gcc/cp/constraint.cc                          |   6 +
 gcc/cp/cp-objcp-common.cc                     |   2 +
 gcc/cp/cp-tree.h                              |   4 +
 gcc/cp/cxx-pretty-print.cc                    |   6 +
 gcc/cp/method.cc                              |  31 ++
 gcc/cp/parser.cc                              |  10 +
 gcc/cp/semantics.cc                           |   8 +
 gcc/testsuite/g++.dg/ext/has-builtin-1.C      |   6 +
 gcc/testsuite/g++.dg/ext/is_convertible1.C    | 269 +++++++++++++++++
 gcc/testsuite/g++.dg/ext/is_convertible2.C    |  46 +++
 .../g++.dg/ext/is_nothrow_convertible1.C      | 270 ++++++++++++++++++
 .../g++.dg/ext/is_nothrow_convertible2.C      |  19 ++
 libstdc++-v3/include/std/type_traits          |   4 +-
 .../is_nothrow_convertible/value_ext.cc       |   4 +-
 16 files changed, 684 insertions(+), 4 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/is_convertible1.C
 create mode 100644 gcc/testsuite/g++.dg/ext/is_convertible2.C
 create mode 100644 gcc/testsuite/g++.dg/ext/is_nothrow_convertible1.C
 create mode 100644 gcc/testsuite/g++.dg/ext/is_nothrow_convertible2.C


base-commit: 32d8123cd6ce87acb557aec230e8359051316f9f
  

On 9/22/22 09:39, Marek Polacek wrote:
> To improve compile times, the C++ library could use compiler built-ins
> rather than implementing std::is_convertible (and _nothrow) as class
> templates.  This patch adds the built-ins.  We already have
> __is_constructible and __is_assignable, and the nothrow forms of those.
> 
> Microsoft (and clang, for compatibility) also provide an alias called
> __is_convertible_to.  I did not add it, but it would be trivial to do
> so.
> 
> I noticed that our __is_assignable doesn't implement the "Access checks
> are performed as if from a context unrelated to either type" requirement,
> therefore std::is_assignable / __is_assignable give two different results
> here:
> 
>    class S {
>      operator int();
>      friend void g(); // #1
>    };
> 
>    void
>    g ()
>    {
>      // #1 doesn't matter
>      static_assert(std::is_assignable<int&, S>::value, "");
>      static_assert(__is_assignable(int&, S), "");
>    }
> 
> This is not a problem if __is_assignable is not meant to be used by
> the users.

That's fine, it's not.

> This patch doesn't make libstdc++ use the new built-ins, but I had to
> rename a class otherwise its name would clash with the new built-in.

Sigh, that's going to be a hassle when comparing compiler versions on 
preprocessed code.

> Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
> 
> 	PR c++/106784
> 
> gcc/c-family/ChangeLog:
> 
> 	* c-common.cc (c_common_reswords): Add __is_convertible and
> 	__is_nothrow_convertible.
> 	* c-common.h (enum rid): Add RID_IS_CONVERTIBLE and
> 	RID_IS_NOTHROW_CONVERTIBLE.
> 
> gcc/cp/ChangeLog:
> 
> 	* constraint.cc (diagnose_trait_expr): Handle CPTK_IS_CONVERTIBLE
> 	and CPTK_IS_NOTHROW_CONVERTIBLE.
> 	* cp-objcp-common.cc (names_builtin_p): Handle RID_IS_CONVERTIBLE
> 	RID_IS_NOTHROW_CONVERTIBLE.
> 	* cp-tree.h (enum cp_trait_kind): Add CPTK_IS_CONVERTIBLE and
> 	CPTK_IS_NOTHROW_CONVERTIBLE.
> 	(is_convertible): Declare.
> 	(is_nothrow_convertible): Likewise.
> 	* cxx-pretty-print.cc (pp_cxx_trait_expression): Handle
> 	CPTK_IS_CONVERTIBLE and CPTK_IS_NOTHROW_CONVERTIBLE.
> 	* method.cc (is_convertible): New.
> 	(is_nothrow_convertible): Likewise.
> 	* parser.cc (cp_parser_primary_expression): Handle RID_IS_CONVERTIBLE
> 	and RID_IS_NOTHROW_CONVERTIBLE.
> 	(cp_parser_trait_expr): Likewise.
> 	* semantics.cc (trait_expr_value): Handle CPTK_IS_CONVERTIBLE and
> 	CPTK_IS_NOTHROW_CONVERTIBLE.
> 	(finish_trait_expr): Likewise.
> 
> libstdc++-v3/ChangeLog:
> 
> 	* include/std/type_traits: Rename __is_nothrow_convertible to
> 	__is_nothrow_convertible_lib.
> 	* testsuite/20_util/is_nothrow_convertible/value_ext.cc: Likewise.
> 
> gcc/testsuite/ChangeLog:
> 
> 	* g++.dg/ext/has-builtin-1.C: Enhance to test __is_convertible and
> 	__is_nothrow_convertible.
> 	* g++.dg/ext/is_convertible1.C: New test.
> 	* g++.dg/ext/is_convertible2.C: New test.
> 	* g++.dg/ext/is_nothrow_convertible1.C: New test.
> 	* g++.dg/ext/is_nothrow_convertible2.C: New test.
> ---
>   gcc/c-family/c-common.cc                      |   2 +
>   gcc/c-family/c-common.h                       |   1 +
>   gcc/cp/constraint.cc                          |   6 +
>   gcc/cp/cp-objcp-common.cc                     |   2 +
>   gcc/cp/cp-tree.h                              |   4 +
>   gcc/cp/cxx-pretty-print.cc                    |   6 +
>   gcc/cp/method.cc                              |  31 ++
>   gcc/cp/parser.cc                              |  10 +
>   gcc/cp/semantics.cc                           |   8 +
>   gcc/testsuite/g++.dg/ext/has-builtin-1.C      |   6 +
>   gcc/testsuite/g++.dg/ext/is_convertible1.C    | 269 +++++++++++++++++
>   gcc/testsuite/g++.dg/ext/is_convertible2.C    |  46 +++
>   .../g++.dg/ext/is_nothrow_convertible1.C      | 270 ++++++++++++++++++
>   .../g++.dg/ext/is_nothrow_convertible2.C      |  19 ++
>   libstdc++-v3/include/std/type_traits          |   4 +-
>   .../is_nothrow_convertible/value_ext.cc       |   4 +-
>   16 files changed, 684 insertions(+), 4 deletions(-)
>   create mode 100644 gcc/testsuite/g++.dg/ext/is_convertible1.C
>   create mode 100644 gcc/testsuite/g++.dg/ext/is_convertible2.C
>   create mode 100644 gcc/testsuite/g++.dg/ext/is_nothrow_convertible1.C
>   create mode 100644 gcc/testsuite/g++.dg/ext/is_nothrow_convertible2.C
> 
> diff --git a/gcc/c-family/c-common.cc b/gcc/c-family/c-common.cc
> index c0f15f4cab1..dce3045c9f2 100644
> --- a/gcc/c-family/c-common.cc
> +++ b/gcc/c-family/c-common.cc
> @@ -541,6 +541,8 @@ const struct c_common_resword c_common_reswords[] =
>     { "__is_constructible", RID_IS_CONSTRUCTIBLE, D_CXXONLY },
>     { "__is_nothrow_assignable", RID_IS_NOTHROW_ASSIGNABLE, D_CXXONLY },
>     { "__is_nothrow_constructible", RID_IS_NOTHROW_CONSTRUCTIBLE, D_CXXONLY },
> +  { "__is_convertible", RID_IS_CONVERTIBLE, D_CXXONLY },
> +  { "__is_nothrow_convertible", RID_IS_NOTHROW_CONVERTIBLE, D_CXXONLY },
>     { "__reference_constructs_from_temporary", RID_REF_CONSTRUCTS_FROM_TEMPORARY,
>   					D_CXXONLY },
>     { "__reference_converts_from_temporary", RID_REF_CONVERTS_FROM_TEMPORARY,
> diff --git a/gcc/c-family/c-common.h b/gcc/c-family/c-common.h
> index 2f592f5cd58..31397d80029 100644
> --- a/gcc/c-family/c-common.h
> +++ b/gcc/c-family/c-common.h
> @@ -184,6 +184,7 @@ enum rid
>     RID_IS_UNION,                RID_UNDERLYING_TYPE,
>     RID_IS_ASSIGNABLE,           RID_IS_CONSTRUCTIBLE,
>     RID_IS_NOTHROW_ASSIGNABLE,   RID_IS_NOTHROW_CONSTRUCTIBLE,
> +  RID_IS_CONVERTIBLE,		RID_IS_NOTHROW_CONVERTIBLE,
>     RID_REF_CONSTRUCTS_FROM_TEMPORARY,
>     RID_REF_CONVERTS_FROM_TEMPORARY,
>   
> diff --git a/gcc/cp/constraint.cc b/gcc/cp/constraint.cc
> index 568318f0ba1..5839bfb4b52 100644
> --- a/gcc/cp/constraint.cc
> +++ b/gcc/cp/constraint.cc
> @@ -3697,6 +3697,12 @@ diagnose_trait_expr (tree expr, tree args)
>       case CPTK_HAS_UNIQUE_OBJ_REPRESENTATIONS:
>         inform (loc, "  %qT does not have unique object representations", t1);
>         break;
> +    case CPTK_IS_CONVERTIBLE:
> +      inform (loc, "  %qT is not convertible from %qE", t2, t1);
> +      break;
> +    case CPTK_IS_NOTHROW_CONVERTIBLE:
> +	inform (loc, "  %qT is not %<nothrow%> convertible from %qE", t2, t1);

It's odd that the existing diagnostics quote "nothrow", which is not a 
keyword.  I wonder why these library traits didn't use "noexcept"?

> +      break;
>       case CPTK_REF_CONSTRUCTS_FROM_TEMPORARY:
>         inform (loc, "  %qT is not a reference that binds to a temporary "
>   	      "object of type %qT (direct-initialization)", t1, t2);
> diff --git a/gcc/cp/cp-objcp-common.cc b/gcc/cp/cp-objcp-common.cc
> index 1ffac08c32f..64975699351 100644
> --- a/gcc/cp/cp-objcp-common.cc
> +++ b/gcc/cp/cp-objcp-common.cc
> @@ -463,6 +463,8 @@ names_builtin_p (const char *name)
>       case RID_IS_NOTHROW_ASSIGNABLE:
>       case RID_IS_NOTHROW_CONSTRUCTIBLE:
>       case RID_UNDERLYING_TYPE:
> +    case RID_IS_CONVERTIBLE:
> +    case RID_IS_NOTHROW_CONVERTIBLE:
>       case RID_REF_CONSTRUCTS_FROM_TEMPORARY:
>       case RID_REF_CONVERTS_FROM_TEMPORARY:
>         return true;
> diff --git a/gcc/cp/cp-tree.h b/gcc/cp/cp-tree.h
> index f19ecafc266..e4d89207e2a 100644
> --- a/gcc/cp/cp-tree.h
> +++ b/gcc/cp/cp-tree.h
> @@ -1407,6 +1407,8 @@ enum cp_trait_kind
>     CPTK_IS_CONSTRUCTIBLE,
>     CPTK_IS_NOTHROW_ASSIGNABLE,
>     CPTK_IS_NOTHROW_CONSTRUCTIBLE,
> +  CPTK_IS_CONVERTIBLE,
> +  CPTK_IS_NOTHROW_CONVERTIBLE,
>     CPTK_REF_CONSTRUCTS_FROM_TEMPORARY,
>     CPTK_REF_CONVERTS_FROM_TEMPORARY
>   };
> @@ -7116,6 +7118,8 @@ extern tree forward_parm			(tree);
>   extern bool is_trivially_xible			(enum tree_code, tree, tree);
>   extern bool is_nothrow_xible			(enum tree_code, tree, tree);
>   extern bool is_xible				(enum tree_code, tree, tree);
> +extern bool is_convertible			(tree, tree);
> +extern bool is_nothrow_convertible		(tree, tree);
>   extern bool ref_xes_from_temporary		(tree, tree, bool);
>   extern tree get_defaulted_eh_spec		(tree, tsubst_flags_t = tf_warning_or_error);
>   extern bool maybe_explain_implicit_delete	(tree);
> diff --git a/gcc/cp/cxx-pretty-print.cc b/gcc/cp/cxx-pretty-print.cc
> index 44590830a61..e18143e39a9 100644
> --- a/gcc/cp/cxx-pretty-print.cc
> +++ b/gcc/cp/cxx-pretty-print.cc
> @@ -2696,6 +2696,12 @@ pp_cxx_trait_expression (cxx_pretty_printer *pp, tree t)
>       case CPTK_IS_NOTHROW_CONSTRUCTIBLE:
>         pp_cxx_ws_string (pp, "__is_nothrow_constructible");
>         break;
> +    case CPTK_IS_CONVERTIBLE:
> +      pp_cxx_ws_string (pp, "__is_convertible");
> +      break;
> +    case CPTK_IS_NOTHROW_CONVERTIBLE:
> +      pp_cxx_ws_string (pp, "__is_nothrow_convertible");
> +      break;
>       case CPTK_REF_CONSTRUCTS_FROM_TEMPORARY:
>         pp_cxx_ws_string (pp, "__reference_constructs_from_temporary");
>         break;
> diff --git a/gcc/cp/method.cc b/gcc/cp/method.cc
> index 573ef016f82..c35a59fe56c 100644
> --- a/gcc/cp/method.cc
> +++ b/gcc/cp/method.cc
> @@ -2236,6 +2236,37 @@ ref_xes_from_temporary (tree to, tree from, bool direct_init_p)
>     return ref_conv_binds_directly (to, val, direct_init_p).is_false ();
>   }
>   
> +/* Return true if FROM can be converted to TO using implicit conversions,
> +   or both FROM and TO are possibly cv-qualified void.  NB: This doesn't
> +   implement the "Access checks are performed as if from a context unrelated
> +   to either type" restriction.  */
> +
> +bool
> +is_convertible (tree from, tree to)

You didn't want to add conversion to is*_xible?

> +{
> +  if (VOID_TYPE_P (from) && VOID_TYPE_P (to))
> +    return true;
> +  tree expr = build_stub_object (from);
> +  expr = perform_implicit_conversion (to, expr, tf_none);
> +  if (expr == error_mark_node)
> +    return false;
> +  return !!expr;
> +}
> +
> +/* Like is_convertible, but the conversion is also noexcept.  */
> +
> +bool
> +is_nothrow_convertible (tree from, tree to)
> +{
> +  if (VOID_TYPE_P (from) && VOID_TYPE_P (to))
> +    return true;
> +  tree expr = build_stub_object (from);
> +  expr = perform_implicit_conversion (to, expr, tf_none);
> +  if (expr == NULL_TREE || expr == error_mark_node)
> +    return false;
> +  return expr_noexcept_p (expr, tf_none);
> +}
> +
>   /* Categorize various special_function_kinds.  */
>   #define SFK_CTOR_P(sfk) \
>     ((sfk) >= sfk_constructor && (sfk) <= sfk_move_constructor)
> diff --git a/gcc/cp/parser.cc b/gcc/cp/parser.cc
> index 3cbe0d69de1..bb83d1c78f6 100644
> --- a/gcc/cp/parser.cc
> +++ b/gcc/cp/parser.cc
> @@ -5922,6 +5922,8 @@ cp_parser_primary_expression (cp_parser *parser,
>   	case RID_IS_CONSTRUCTIBLE:
>   	case RID_IS_NOTHROW_ASSIGNABLE:
>   	case RID_IS_NOTHROW_CONSTRUCTIBLE:
> +	case RID_IS_CONVERTIBLE:
> +	case RID_IS_NOTHROW_CONVERTIBLE:
>   	case RID_REF_CONSTRUCTS_FROM_TEMPORARY:
>   	case RID_REF_CONVERTS_FROM_TEMPORARY:
>   	  return cp_parser_trait_expr (parser, token->keyword);
> @@ -11008,6 +11010,14 @@ cp_parser_trait_expr (cp_parser* parser, enum rid keyword)
>         kind = CPTK_IS_NOTHROW_CONSTRUCTIBLE;
>         variadic = true;
>         break;
> +    case RID_IS_CONVERTIBLE:
> +      kind = CPTK_IS_CONVERTIBLE;
> +      binary = true;
> +      break;
> +    case RID_IS_NOTHROW_CONVERTIBLE:
> +      kind = CPTK_IS_NOTHROW_CONVERTIBLE;
> +      binary = true;
> +      break;
>       case RID_REF_CONSTRUCTS_FROM_TEMPORARY:
>         kind = CPTK_REF_CONSTRUCTS_FROM_TEMPORARY;
>         binary = true;
> diff --git a/gcc/cp/semantics.cc b/gcc/cp/semantics.cc
> index 86562071612..92fc795df40 100644
> --- a/gcc/cp/semantics.cc
> +++ b/gcc/cp/semantics.cc
> @@ -12044,6 +12044,12 @@ trait_expr_value (cp_trait_kind kind, tree type1, tree type2)
>       case CPTK_IS_NOTHROW_CONSTRUCTIBLE:
>         return is_nothrow_xible (INIT_EXPR, type1, type2);
>   
> +    case CPTK_IS_CONVERTIBLE:
> +      return is_convertible (type1, type2);
> +
> +    case CPTK_IS_NOTHROW_CONVERTIBLE:
> +      return is_nothrow_convertible (type1, type2);
> +
>       case CPTK_REF_CONSTRUCTS_FROM_TEMPORARY:
>         return ref_xes_from_temporary (type1, type2, /*direct_init=*/true);
>   
> @@ -12165,6 +12171,8 @@ finish_trait_expr (location_t loc, cp_trait_kind kind, tree type1, tree type2)
>       case CPTK_IS_TRIVIALLY_CONSTRUCTIBLE:
>       case CPTK_IS_NOTHROW_ASSIGNABLE:
>       case CPTK_IS_NOTHROW_CONSTRUCTIBLE:
> +    case CPTK_IS_CONVERTIBLE:
> +    case CPTK_IS_NOTHROW_CONVERTIBLE:
>       case CPTK_REF_CONSTRUCTS_FROM_TEMPORARY:
>       case CPTK_REF_CONVERTS_FROM_TEMPORARY:
>         if (!check_trait_type (type1)
> diff --git a/gcc/testsuite/g++.dg/ext/has-builtin-1.C b/gcc/testsuite/g++.dg/ext/has-builtin-1.C
> index fe25cb2f669..17dabf648cf 100644
> --- a/gcc/testsuite/g++.dg/ext/has-builtin-1.C
> +++ b/gcc/testsuite/g++.dg/ext/has-builtin-1.C
> @@ -131,3 +131,9 @@
>   #if !__has_builtin (__builtin_is_pointer_interconvertible_with_class)
>   # error "__has_builtin (__builtin_is_pointer_interconvertible_with_class) failed"
>   #endif
> +#if !__has_builtin (__is_convertible)
> +# error "__has_builtin (__is_convertible) failed"
> +#endif
> +#if !__has_builtin (__is_nothrow_convertible)
> +# error "__has_builtin (__is_nothrow_convertible) failed"
> +#endif
> diff --git a/gcc/testsuite/g++.dg/ext/is_convertible1.C b/gcc/testsuite/g++.dg/ext/is_convertible1.C
> new file mode 100644
> index 00000000000..2e72945bceb
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ext/is_convertible1.C
> @@ -0,0 +1,269 @@
> +// PR c++/106784
> +// { dg-do compile { target c++11 } }
> +
> +#define SA(X) static_assert((X),#X)
> +
> +template<typename From, typename To>
> +struct is_convertible {
> +  static const bool value = __is_convertible(From, To);
> +};
> +
> +struct from_int {
> +  from_int(int);
> +};
> +
> +struct from_charp {
> +  from_charp(const char *);
> +};
> +
> +struct to_int {
> +  operator int();
> +};
> +
> +typedef int Fn(int);
> +typedef char Arr[10];
> +enum E { XYZZY };
> +
> +SA(!__is_convertible(int, void));
> +SA(__is_convertible(int, int));
> +SA(__is_convertible(int, from_int));
> +SA(__is_convertible(long, from_int));
> +SA(__is_convertible(double, from_int));
> +SA(__is_convertible(const int, from_int));
> +SA(__is_convertible(const int&, from_int));
> +SA(__is_convertible(to_int, int));
> +SA(__is_convertible(to_int, const int&));
> +SA(__is_convertible(to_int, long));
> +SA(!__is_convertible(to_int, int&));
> +SA(!__is_convertible(to_int, from_int));
> +SA(!__is_convertible(int, Fn));
> +SA(!__is_convertible(int, Fn*));
> +SA(!__is_convertible(int, Fn&));
> +SA(!__is_convertible(int, Arr));
> +SA(!__is_convertible(int, Arr&));
> +SA(!__is_convertible(int, int&));
> +SA(__is_convertible(int, const int&));
> +SA(!__is_convertible(const int, int&));
> +SA(__is_convertible(const int, const int&));
> +SA(!__is_convertible(int, int*));
> +
> +SA(!__is_convertible(int, E));
> +SA(__is_convertible(E, int));
> +
> +SA(__is_convertible(int&, int));
> +SA(__is_convertible(int&, int&));
> +SA(__is_convertible(int&, const int&));
> +SA(!__is_convertible(const int&, int&));
> +SA(__is_convertible(const int&, const int&));
> +SA(!__is_convertible(int&, int*));
> +SA(!__is_convertible(int&, void));
> +SA(!__is_convertible(int&, Fn));
> +SA(!__is_convertible(int&, Fn*));
> +SA(!__is_convertible(int&, Fn&));
> +SA(!__is_convertible(int&, Arr));
> +SA(!__is_convertible(int&, Arr&));
> +
> +SA(!__is_convertible(int*, int));
> +SA(!__is_convertible(int*, int&));
> +SA(!__is_convertible(int*, void));
> +SA(__is_convertible(int*, int*));
> +SA(__is_convertible(int*, const int*));
> +SA(!__is_convertible(const int*, int*));
> +SA(__is_convertible(const int*, const int*));
> +SA(!__is_convertible(int*, Fn));
> +SA(!__is_convertible(int*, Fn*));
> +SA(!__is_convertible(int*, Fn&));
> +SA(!__is_convertible(int*, Arr));
> +SA(!__is_convertible(int*, Arr&));
> +SA(!__is_convertible(int*, float*));
> +
> +SA(__is_convertible(void, void));
> +SA(!__is_convertible(void, char));
> +SA(!__is_convertible(void, char&));
> +SA(!__is_convertible(void, char*));
> +SA(!__is_convertible(char, void));
> +SA(__is_convertible(const void, void));
> +SA(__is_convertible(void, const void));
> +SA(__is_convertible(const void, const void));
> +SA(!__is_convertible(void, Fn));
> +SA(!__is_convertible(void, Fn&));
> +SA(!__is_convertible(void, Fn*));
> +SA(!__is_convertible(void, Arr));
> +SA(!__is_convertible(void, Arr&));
> +
> +SA(!__is_convertible(Fn, void));
> +SA(!__is_convertible(Fn, Fn));
> +SA(__is_convertible(Fn, Fn*));
> +SA(__is_convertible(Fn, Fn&));
> +SA(!__is_convertible(int(int), int(int)));
> +SA(__is_convertible(int(int), int(&)(int)));
> +SA(__is_convertible(int(int), int(&&)(int)));
> +SA(__is_convertible(int(int), int(*)(int)));
> +SA(__is_convertible(int(int), int(*const)(int)));
> +SA(!__is_convertible(int(int), char));
> +SA(!__is_convertible(int(int), char*));
> +SA(!__is_convertible(int(int), char&));
> +
> +SA(!__is_convertible(Fn&, void));
> +SA(!__is_convertible(Fn&, Fn));
> +SA(__is_convertible(Fn&, Fn&));
> +SA(__is_convertible(Fn&, Fn*));
> +SA(!__is_convertible(Fn&, Arr));
> +SA(!__is_convertible(Fn&, Arr&));
> +SA(!__is_convertible(Fn&, char));
> +SA(!__is_convertible(Fn&, char&));
> +SA(!__is_convertible(Fn&, char*));
> +
> +SA(!__is_convertible(Fn*, void));
> +SA(!__is_convertible(Fn*, Fn));
> +SA(!__is_convertible(Fn*, Fn&));
> +SA(__is_convertible(Fn*, Fn*));
> +SA(!__is_convertible(Fn*, Arr));
> +SA(!__is_convertible(Fn*, Arr&));
> +SA(!__is_convertible(Fn*, char));
> +SA(!__is_convertible(Fn*, char&));
> +SA(!__is_convertible(Fn*, char*));
> +
> +SA(!__is_convertible(Arr, void));
> +SA(!__is_convertible(Arr, Fn));
> +SA(!__is_convertible(Arr, Fn*));
> +SA(!__is_convertible(Arr, Fn&));
> +SA(!__is_convertible(Arr, Arr));
> +SA(!__is_convertible(Arr, Arr&));
> +SA(__is_convertible(Arr, const Arr&));
> +SA(!__is_convertible(Arr, volatile Arr&));
> +SA(!__is_convertible(Arr, const volatile Arr&));
> +SA(!__is_convertible(const Arr, Arr&));
> +SA(__is_convertible(const Arr, const Arr&));
> +SA(__is_convertible(Arr, Arr&&));
> +SA(__is_convertible(Arr, const Arr&&));
> +SA(__is_convertible(Arr, volatile Arr&&));
> +SA(__is_convertible(Arr, const volatile Arr&&));
> +SA(__is_convertible(const Arr, const Arr&&));
> +SA(!__is_convertible(Arr&, Arr&&));
> +SA(!__is_convertible(Arr&&, Arr&));
> +SA(!__is_convertible(Arr, char));
> +SA(__is_convertible(Arr, char*));
> +SA(__is_convertible(Arr, const char*));
> +SA(!__is_convertible(Arr, char&));
> +SA(!__is_convertible(const Arr, char*));
> +SA(__is_convertible(const Arr, const char*));
> +SA(!__is_convertible(int, int[1]));
> +SA(!__is_convertible(int[1], int[1]));
> +SA(!__is_convertible(int[1], int(&)[1]));
> +SA(__is_convertible(int(&)[1], int(&)[1]));
> +SA(__is_convertible(int(&)[1], const int(&)[1]));
> +SA(!__is_convertible(const int(&)[1], int(&)[1]));
> +SA(!__is_convertible(int[1][1], int*));
> +SA(!__is_convertible(int[][1], int*));
> +
> +SA(!__is_convertible(Arr&, void));
> +SA(!__is_convertible(Arr&, Fn));
> +SA(!__is_convertible(Arr&, Fn*));
> +SA(!__is_convertible(Arr&, Fn&));
> +SA(!__is_convertible(Arr&, Arr));
> +SA(__is_convertible(Arr&, Arr&));
> +SA(__is_convertible(Arr&, const Arr&));
> +SA(!__is_convertible(const Arr&, Arr&));
> +SA(__is_convertible(const Arr&, const Arr&));
> +SA(!__is_convertible(Arr&, char));
> +SA(__is_convertible(Arr&, char*));
> +SA(__is_convertible(Arr&, const char*));
> +SA(!__is_convertible(Arr&, char&));
> +SA(!__is_convertible(const Arr&, char*));
> +SA(__is_convertible(const Arr&, const char*));
> +SA(__is_convertible(Arr, from_charp));
> +SA(__is_convertible(Arr&, from_charp));
> +
> +struct B { };
> +struct D : B { };
> +
> +SA(__is_convertible(D, B));
> +SA(__is_convertible(D*, B*));
> +SA(__is_convertible(D&, B&));
> +SA(!__is_convertible(B, D));
> +SA(!__is_convertible(B*, D*));
> +SA(!__is_convertible(B&, D&));
> +
> +/* These are taken from LLVM's test/SemaCXX/type-traits.cpp.  */
> +
> +struct I {
> +  int i;
> +  I(int _i) : i(_i) { }
> +  operator int() const {
> +    return i;
> +  }
> +};
> +
> +struct F
> +{
> +  float f;
> +  F(float _f) : f(_f) {}
> +  F(const I& obj)
> +    : f(static_cast<float>(obj.i)) {}
> +  operator float() const {
> +    return f;
> +  }
> +  operator I() const {
> +    return I(static_cast<int>(f));
> +  }
> +};
> +
> +SA(__is_convertible(I, I));
> +SA(__is_convertible(I, const I));
> +SA(__is_convertible(I, int));
> +SA(__is_convertible(int, I));
> +SA(__is_convertible(I, F));
> +SA(__is_convertible(F, I));
> +SA(__is_convertible(F, float));
> +SA(__is_convertible(float, F));
> +
> +template<typename>
> +struct X {
> +  template<typename U> X(const X<U>&);
> +};
> +
> +SA(__is_convertible(X<int>, X<float>));
> +SA(__is_convertible(X<float>, X<int>));
> +
> +struct Abstract {
> +  virtual void f() = 0;
> +};
> +
> +SA(!__is_convertible(Abstract, Abstract));
> +
> +class hidden {
> +  hidden(const hidden&);
> +  friend void test ();
> +};
> +
> +SA(__is_convertible(hidden&, hidden&));
> +SA(__is_convertible(hidden&, const hidden&));
> +SA(__is_convertible(hidden&, volatile hidden&));
> +SA(__is_convertible(hidden&, const volatile hidden&));
> +SA(__is_convertible(const hidden&, const hidden&));
> +SA(__is_convertible(const hidden&, const volatile hidden&));
> +SA(__is_convertible(volatile hidden&, const volatile hidden&));
> +SA(__is_convertible(const volatile hidden&, const volatile hidden&));
> +SA(!__is_convertible(const hidden&, hidden&));
> +
> +void
> +test ()
> +{
> +  /* __is_convertible(hidden, hidden) should be false despite the
> +     friend declaration above, because "Access checks are performed
> +     as if from a context unrelated to either type", but we don't
> +     implement that for the built-in (std::is_convertible works as
> +     expected).  This is the case for __is_assignable as well.  */
> +  //SA(!__is_convertible(hidden, hidden));
> +}
> +
> +void
> +test2 ()
> +{
> +  struct X { };
> +  struct Y {
> +    explicit Y(X); // not viable for implicit conversions
> +  };
> +  SA(!__is_convertible(X, Y));
> +}
> diff --git a/gcc/testsuite/g++.dg/ext/is_convertible2.C b/gcc/testsuite/g++.dg/ext/is_convertible2.C
> new file mode 100644
> index 00000000000..9b46e264379
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ext/is_convertible2.C
> @@ -0,0 +1,46 @@
> +// PR c++/106784
> +// { dg-do compile { target c++20 } }
> +// Adapted from <https://en.cppreference.com/w/cpp/types/is_convertible>.
> +
> +#include <string>
> +#include <string_view>
> +
> +#define SA(X) static_assert((X),#X)
> +
> +class E { public: template<class T> E(T&&) { } };
> +
> +int main()
> +{
> +    class A {};
> +    class B : public A {};
> +    class C {};
> +    class D { public: operator C() { return c; }  C c; };
> +
> +    SA(__is_convertible(B*, A*));
> +    SA(!__is_convertible(A*, B*));
> +    SA(__is_convertible(D, C));
> +    SA(!__is_convertible(B*, C*));
> +    SA(__is_convertible(A, E));
> +
> +    using std::operator "" s, std::operator "" sv;
> +
> +    auto stringify = []<typename T>(T x) {
> +        if constexpr (std::is_convertible_v<T, std::string> or
> +                      std::is_convertible_v<T, std::string_view>) {
> +            return x;
> +        } else {
> +            return std::to_string(x);
> +        }
> +    };
> +
> +    const char* three = "three";
> +
> +    SA(!__is_convertible(std::string_view, std::string));
> +    SA(__is_convertible(std::string, std::string_view));
> +
> +    auto s1 = stringify("one"s);
> +    auto s2 = stringify("two"sv);
> +    auto s3 = stringify(three);
> +    auto s4 = stringify(42);
> +    auto s5 = stringify(42.);
> +}
> diff --git a/gcc/testsuite/g++.dg/ext/is_nothrow_convertible1.C b/gcc/testsuite/g++.dg/ext/is_nothrow_convertible1.C
> new file mode 100644
> index 00000000000..bb7243e4611
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ext/is_nothrow_convertible1.C
> @@ -0,0 +1,270 @@
> +// PR c++/106784
> +// { dg-do compile { target c++11 } }
> +// Like is_convertible1.C, but conversion functions are made noexcept.
> +
> +#define SA(X) static_assert((X),#X)
> +
> +template<typename From, typename To>
> +struct is_nothrow_convertible {
> +  static const bool value = __is_nothrow_convertible(From, To);
> +};
> +
> +struct from_int {
> +  from_int(int) noexcept;
> +};
> +
> +struct from_charp {
> +  from_charp(const char *) noexcept;
> +};
> +
> +struct to_int {
> +  operator int() noexcept;
> +};
> +
> +typedef int Fn(int);
> +typedef char Arr[10];
> +enum E { XYZZY };
> +
> +SA(!__is_nothrow_convertible(int, void));
> +SA(__is_nothrow_convertible(int, int));
> +SA(__is_nothrow_convertible(int, from_int));
> +SA(__is_nothrow_convertible(long, from_int));
> +SA(__is_nothrow_convertible(double, from_int));
> +SA(__is_nothrow_convertible(const int, from_int));
> +SA(__is_nothrow_convertible(const int&, from_int));
> +SA(__is_nothrow_convertible(to_int, int));
> +SA(__is_nothrow_convertible(to_int, const int&));
> +SA(__is_nothrow_convertible(to_int, long));
> +SA(!__is_nothrow_convertible(to_int, int&));
> +SA(!__is_nothrow_convertible(to_int, from_int));
> +SA(!__is_nothrow_convertible(int, Fn));
> +SA(!__is_nothrow_convertible(int, Fn*));
> +SA(!__is_nothrow_convertible(int, Fn&));
> +SA(!__is_nothrow_convertible(int, Arr));
> +SA(!__is_nothrow_convertible(int, Arr&));
> +SA(!__is_nothrow_convertible(int, int&));
> +SA(__is_nothrow_convertible(int, const int&));
> +SA(!__is_nothrow_convertible(const int, int&));
> +SA(__is_nothrow_convertible(const int, const int&));
> +SA(!__is_nothrow_convertible(int, int*));
> +
> +SA(!__is_nothrow_convertible(int, E));
> +SA(__is_nothrow_convertible(E, int));
> +
> +SA(__is_nothrow_convertible(int&, int));
> +SA(__is_nothrow_convertible(int&, int&));
> +SA(__is_nothrow_convertible(int&, const int&));
> +SA(!__is_nothrow_convertible(const int&, int&));
> +SA(__is_nothrow_convertible(const int&, const int&));
> +SA(!__is_nothrow_convertible(int&, int*));
> +SA(!__is_nothrow_convertible(int&, void));
> +SA(!__is_nothrow_convertible(int&, Fn));
> +SA(!__is_nothrow_convertible(int&, Fn*));
> +SA(!__is_nothrow_convertible(int&, Fn&));
> +SA(!__is_nothrow_convertible(int&, Arr));
> +SA(!__is_nothrow_convertible(int&, Arr&));
> +
> +SA(!__is_nothrow_convertible(int*, int));
> +SA(!__is_nothrow_convertible(int*, int&));
> +SA(!__is_nothrow_convertible(int*, void));
> +SA(__is_nothrow_convertible(int*, int*));
> +SA(__is_nothrow_convertible(int*, const int*));
> +SA(!__is_nothrow_convertible(const int*, int*));
> +SA(__is_nothrow_convertible(const int*, const int*));
> +SA(!__is_nothrow_convertible(int*, Fn));
> +SA(!__is_nothrow_convertible(int*, Fn*));
> +SA(!__is_nothrow_convertible(int*, Fn&));
> +SA(!__is_nothrow_convertible(int*, Arr));
> +SA(!__is_nothrow_convertible(int*, Arr&));
> +SA(!__is_nothrow_convertible(int*, float*));
> +
> +SA(__is_nothrow_convertible(void, void));
> +SA(!__is_nothrow_convertible(void, char));
> +SA(!__is_nothrow_convertible(void, char&));
> +SA(!__is_nothrow_convertible(void, char*));
> +SA(!__is_nothrow_convertible(char, void));
> +SA(__is_nothrow_convertible(const void, void));
> +SA(__is_nothrow_convertible(void, const void));
> +SA(__is_nothrow_convertible(const void, const void));
> +SA(!__is_nothrow_convertible(void, Fn));
> +SA(!__is_nothrow_convertible(void, Fn&));
> +SA(!__is_nothrow_convertible(void, Fn*));
> +SA(!__is_nothrow_convertible(void, Arr));
> +SA(!__is_nothrow_convertible(void, Arr&));
> +
> +SA(!__is_nothrow_convertible(Fn, void));
> +SA(!__is_nothrow_convertible(Fn, Fn));
> +SA(__is_nothrow_convertible(Fn, Fn*));
> +SA(__is_nothrow_convertible(Fn, Fn&));
> +SA(!__is_nothrow_convertible(int(int), int(int)));
> +SA(__is_nothrow_convertible(int(int), int(&)(int)));
> +SA(__is_nothrow_convertible(int(int), int(&&)(int)));
> +SA(__is_nothrow_convertible(int(int), int(*)(int)));
> +SA(__is_nothrow_convertible(int(int), int(*const)(int)));
> +SA(!__is_nothrow_convertible(int(int), char));
> +SA(!__is_nothrow_convertible(int(int), char*));
> +SA(!__is_nothrow_convertible(int(int), char&));
> +
> +SA(!__is_nothrow_convertible(Fn&, void));
> +SA(!__is_nothrow_convertible(Fn&, Fn));
> +SA(__is_nothrow_convertible(Fn&, Fn&));
> +SA(__is_nothrow_convertible(Fn&, Fn*));
> +SA(!__is_nothrow_convertible(Fn&, Arr));
> +SA(!__is_nothrow_convertible(Fn&, Arr&));
> +SA(!__is_nothrow_convertible(Fn&, char));
> +SA(!__is_nothrow_convertible(Fn&, char&));
> +SA(!__is_nothrow_convertible(Fn&, char*));
> +
> +SA(!__is_nothrow_convertible(Fn*, void));
> +SA(!__is_nothrow_convertible(Fn*, Fn));
> +SA(!__is_nothrow_convertible(Fn*, Fn&));
> +SA(__is_nothrow_convertible(Fn*, Fn*));
> +SA(!__is_nothrow_convertible(Fn*, Arr));
> +SA(!__is_nothrow_convertible(Fn*, Arr&));
> +SA(!__is_nothrow_convertible(Fn*, char));
> +SA(!__is_nothrow_convertible(Fn*, char&));
> +SA(!__is_nothrow_convertible(Fn*, char*));
> +
> +SA(!__is_nothrow_convertible(Arr, void));
> +SA(!__is_nothrow_convertible(Arr, Fn));
> +SA(!__is_nothrow_convertible(Arr, Fn*));
> +SA(!__is_nothrow_convertible(Arr, Fn&));
> +SA(!__is_nothrow_convertible(Arr, Arr));
> +SA(!__is_nothrow_convertible(Arr, Arr&));
> +SA(__is_nothrow_convertible(Arr, const Arr&));
> +SA(!__is_nothrow_convertible(Arr, volatile Arr&));
> +SA(!__is_nothrow_convertible(Arr, const volatile Arr&));
> +SA(!__is_nothrow_convertible(const Arr, Arr&));
> +SA(__is_nothrow_convertible(const Arr, const Arr&));
> +SA(__is_nothrow_convertible(Arr, Arr&&));
> +SA(__is_nothrow_convertible(Arr, const Arr&&));
> +SA(__is_nothrow_convertible(Arr, volatile Arr&&));
> +SA(__is_nothrow_convertible(Arr, const volatile Arr&&));
> +SA(__is_nothrow_convertible(const Arr, const Arr&&));
> +SA(!__is_nothrow_convertible(Arr&, Arr&&));
> +SA(!__is_nothrow_convertible(Arr&&, Arr&));
> +SA(!__is_nothrow_convertible(Arr, char));
> +SA(__is_nothrow_convertible(Arr, char*));
> +SA(__is_nothrow_convertible(Arr, const char*));
> +SA(!__is_nothrow_convertible(Arr, char&));
> +SA(!__is_nothrow_convertible(const Arr, char*));
> +SA(__is_nothrow_convertible(const Arr, const char*));
> +SA(!__is_nothrow_convertible(int, int[1]));
> +SA(!__is_nothrow_convertible(int[1], int[1]));
> +SA(!__is_nothrow_convertible(int[1], int(&)[1]));
> +SA(__is_nothrow_convertible(int(&)[1], int(&)[1]));
> +SA(__is_nothrow_convertible(int(&)[1], const int(&)[1]));
> +SA(!__is_nothrow_convertible(const int(&)[1], int(&)[1]));
> +SA(!__is_nothrow_convertible(int[1][1], int*));
> +SA(!__is_nothrow_convertible(int[][1], int*));
> +
> +SA(!__is_nothrow_convertible(Arr&, void));
> +SA(!__is_nothrow_convertible(Arr&, Fn));
> +SA(!__is_nothrow_convertible(Arr&, Fn*));
> +SA(!__is_nothrow_convertible(Arr&, Fn&));
> +SA(!__is_nothrow_convertible(Arr&, Arr));
> +SA(__is_nothrow_convertible(Arr&, Arr&));
> +SA(__is_nothrow_convertible(Arr&, const Arr&));
> +SA(!__is_nothrow_convertible(const Arr&, Arr&));
> +SA(__is_nothrow_convertible(const Arr&, const Arr&));
> +SA(!__is_nothrow_convertible(Arr&, char));
> +SA(__is_nothrow_convertible(Arr&, char*));
> +SA(__is_nothrow_convertible(Arr&, const char*));
> +SA(!__is_nothrow_convertible(Arr&, char&));
> +SA(!__is_nothrow_convertible(const Arr&, char*));
> +SA(__is_nothrow_convertible(const Arr&, const char*));
> +SA(__is_nothrow_convertible(Arr, from_charp));
> +SA(__is_nothrow_convertible(Arr&, from_charp));
> +
> +struct B { };
> +struct D : B { };
> +
> +SA(__is_nothrow_convertible(D, B));
> +SA(__is_nothrow_convertible(D*, B*));
> +SA(__is_nothrow_convertible(D&, B&));
> +SA(!__is_nothrow_convertible(B, D));
> +SA(!__is_nothrow_convertible(B*, D*));
> +SA(!__is_nothrow_convertible(B&, D&));
> +
> +/* These are taken from LLVM's test/SemaCXX/type-traits.cpp.  */
> +
> +struct I {
> +  int i;
> +  I(int _i) noexcept : i(_i)  { }
> +  operator int() const noexcept {
> +    return i;
> +  }
> +};
> +
> +struct F
> +{
> +  float f;
> +  F(float _f) noexcept : f(_f) {}
> +  F(const I& obj) noexcept
> +    : f(static_cast<float>(obj.i)) {}
> +  operator float() const noexcept {
> +    return f;
> +  }
> +  operator I() const noexcept {
> +    return I(static_cast<int>(f));
> +  }
> +};
> +
> +SA(__is_nothrow_convertible(I, I));
> +SA(__is_nothrow_convertible(I, const I));
> +SA(__is_nothrow_convertible(I, int));
> +SA(__is_nothrow_convertible(int, I));
> +SA(__is_nothrow_convertible(I, F));
> +SA(__is_nothrow_convertible(F, I));
> +SA(__is_nothrow_convertible(F, float));
> +SA(__is_nothrow_convertible(float, F));
> +
> +template<typename>
> +struct X {
> +  template<typename U> X(const X<U>&) noexcept;
> +};
> +
> +SA(__is_nothrow_convertible(X<int>, X<float>));
> +SA(__is_nothrow_convertible(X<float>, X<int>));
> +
> +struct Abstract {
> +  virtual void f() = 0;
> +};
> +
> +SA(!__is_nothrow_convertible(Abstract, Abstract));
> +
> +class hidden {
> +  hidden(const hidden&);
> +  friend void test ();
> +};
> +
> +SA(__is_nothrow_convertible(hidden&, hidden&));
> +SA(__is_nothrow_convertible(hidden&, const hidden&));
> +SA(__is_nothrow_convertible(hidden&, volatile hidden&));
> +SA(__is_nothrow_convertible(hidden&, const volatile hidden&));
> +SA(__is_nothrow_convertible(const hidden&, const hidden&));
> +SA(__is_nothrow_convertible(const hidden&, const volatile hidden&));
> +SA(__is_nothrow_convertible(volatile hidden&, const volatile hidden&));
> +SA(__is_nothrow_convertible(const volatile hidden&, const volatile hidden&));
> +SA(!__is_nothrow_convertible(const hidden&, hidden&));
> +
> +void
> +test ()
> +{
> +  /* __is_nothrow_convertible(hidden, hidden) should be false despite the
> +     friend declaration above, because "Access checks are performed
> +     as if from a context unrelated to either type", but we don't
> +     implement that for the built-in (std::is_convertible works as
> +     expected).  This is the case for __is_assignable as well.  */
> +  //SA(!__is_nothrow_convertible(hidden, hidden));
> +}
> +
> +void
> +test2 ()
> +{
> +  struct X { };
> +  struct Y {
> +    explicit Y(X); // not viable for implicit conversions
> +  };
> +  SA(!__is_nothrow_convertible(X, Y));
> +}
> diff --git a/gcc/testsuite/g++.dg/ext/is_nothrow_convertible2.C b/gcc/testsuite/g++.dg/ext/is_nothrow_convertible2.C
> new file mode 100644
> index 00000000000..aa089173b75
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ext/is_nothrow_convertible2.C
> @@ -0,0 +1,19 @@
> +// PR c++/106784
> +// { dg-do compile { target c++11 } }
> +
> +#define SA(X) static_assert((X),#X)
> +
> +struct A { };
> +struct B { };
> +
> +struct M {
> +  operator A();
> +  operator B() noexcept;
> +  M(const A&);
> +  M(const B&) noexcept;
> +};
> +
> +SA(!__is_nothrow_convertible(A, M));
> +SA(!__is_nothrow_convertible(M, A));
> +SA(__is_nothrow_convertible(B, M));
> +SA(__is_nothrow_convertible(M, B));
> diff --git a/libstdc++-v3/include/std/type_traits b/libstdc++-v3/include/std/type_traits
> index 94e73eafd2f..1797b9e97f7 100644
> --- a/libstdc++-v3/include/std/type_traits
> +++ b/libstdc++-v3/include/std/type_traits
> @@ -1453,7 +1453,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
>   
>     // is_nothrow_convertible for C++11
>     template<typename _From, typename _To>
> -    struct __is_nothrow_convertible
> +    struct __is_nothrow_convertible_lib
>       : public __is_nt_convertible_helper<_From, _To>::type
>       { };
>   
> @@ -2999,7 +2999,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
>       struct __is_nt_invocable_impl<_Result, _Ret,
>   				  __void_t<typename _Result::type>>
>       : __or_<is_void<_Ret>,
> -	    __is_nothrow_convertible<typename _Result::type, _Ret>>::type
> +	    __is_nothrow_convertible_lib<typename _Result::type, _Ret>>::type
>       { };
>     /// @endcond
>   
> diff --git a/libstdc++-v3/testsuite/20_util/is_nothrow_convertible/value_ext.cc b/libstdc++-v3/testsuite/20_util/is_nothrow_convertible/value_ext.cc
> index 0f896428537..d736d2ca260 100644
> --- a/libstdc++-v3/testsuite/20_util/is_nothrow_convertible/value_ext.cc
> +++ b/libstdc++-v3/testsuite/20_util/is_nothrow_convertible/value_ext.cc
> @@ -19,10 +19,10 @@
>   
>   #include <type_traits>
>   
> -// Test the non-standard __is_nothrow_convertible trait
> +// Test the non-standard __is_nothrow_convertible_lib trait
>   
>   template<typename From, typename To>
> -  using is_nothrow_convertible = std::__is_nothrow_convertible<From, To>;
> +  using is_nothrow_convertible = std::__is_nothrow_convertible_lib<From, To>;
>   
>   #define IS_NT_CONVERTIBLE_DEFINED
>   #include "value.cc"
> 
> base-commit: 32d8123cd6ce87acb557aec230e8359051316f9f
  

On Thu, Sep 22, 2022 at 06:14:44PM -0400, Jason Merrill wrote:
> On 9/22/22 09:39, Marek Polacek wrote:
> > To improve compile times, the C++ library could use compiler built-ins
> > rather than implementing std::is_convertible (and _nothrow) as class
> > templates.  This patch adds the built-ins.  We already have
> > __is_constructible and __is_assignable, and the nothrow forms of those.
> > 
> > Microsoft (and clang, for compatibility) also provide an alias called
> > __is_convertible_to.  I did not add it, but it would be trivial to do
> > so.
> > 
> > I noticed that our __is_assignable doesn't implement the "Access checks
> > are performed as if from a context unrelated to either type" requirement,
> > therefore std::is_assignable / __is_assignable give two different results
> > here:
> > 
> >    class S {
> >      operator int();
> >      friend void g(); // #1
> >    };
> > 
> >    void
> >    g ()
> >    {
> >      // #1 doesn't matter
> >      static_assert(std::is_assignable<int&, S>::value, "");
> >      static_assert(__is_assignable(int&, S), "");
> >    }
> > 
> > This is not a problem if __is_assignable is not meant to be used by
> > the users.
> 
> That's fine, it's not.
 
Okay then.  libstdc++ needs to make sure then that it's handled right.

> > This patch doesn't make libstdc++ use the new built-ins, but I had to
> > rename a class otherwise its name would clash with the new built-in.
> 
> Sigh, that's going to be a hassle when comparing compiler versions on
> preprocessed code.

Yeah, I guess :/.  Kind of like __integer_pack / __make_integer_seq.

> > --- a/gcc/cp/constraint.cc
> > +++ b/gcc/cp/constraint.cc
> > @@ -3697,6 +3697,12 @@ diagnose_trait_expr (tree expr, tree args)
> >       case CPTK_HAS_UNIQUE_OBJ_REPRESENTATIONS:
> >         inform (loc, "  %qT does not have unique object representations", t1);
> >         break;
> > +    case CPTK_IS_CONVERTIBLE:
> > +      inform (loc, "  %qT is not convertible from %qE", t2, t1);
> > +      break;
> > +    case CPTK_IS_NOTHROW_CONVERTIBLE:
> > +	inform (loc, "  %qT is not %<nothrow%> convertible from %qE", t2, t1);
> 
> It's odd that the existing diagnostics quote "nothrow", which is not a
> keyword.  I wonder why these library traits didn't use "noexcept"?

Eh, yeah, only "throw" is.  The quotes were deliberately added in
<https://gcc.gnu.org/pipermail/gcc-patches/2019-May/522333.html>.  Should
I prepare a separate patch to use "%<noexcept%>" rather than "%<nothrow%>"?
OTOH, the traits have "nothrow" in their names, so maybe just go back to
"nothrow"?
 
> > --- a/gcc/cp/method.cc
> > +++ b/gcc/cp/method.cc
> > @@ -2236,6 +2236,37 @@ ref_xes_from_temporary (tree to, tree from, bool direct_init_p)
> >     return ref_conv_binds_directly (to, val, direct_init_p).is_false ();
> >   }
> > +/* Return true if FROM can be converted to TO using implicit conversions,
> > +   or both FROM and TO are possibly cv-qualified void.  NB: This doesn't
> > +   implement the "Access checks are performed as if from a context unrelated
> > +   to either type" restriction.  */
> > +
> > +bool
> > +is_convertible (tree from, tree to)
> 
> You didn't want to add conversion to is*_xible?

No, it didn't look like a good fit.  It does things we don't need, and
also has if VOID_TYPE_P -> return error_mark_node; which would be wrong
for __is_convertible.

I realized I'm not testing passing an incomplete type to the built-in,
but since that is UB, I reckon we don't need to test it (we issue
"error: invalid use of incomplete type").

Marek
  

On Thu, 22 Sept 2022 at 23:14, Jason Merrill wrote:
> On 9/22/22 09:39, Marek Polacek wrote:
> > This patch doesn't make libstdc++ use the new built-ins, but I had to
> > rename a class otherwise its name would clash with the new built-in.
>
> Sigh, that's going to be a hassle when comparing compiler versions on
> preprocessed code.

Good point. Clang has some gross hacks that we could consider. When it
sees a declaration that uses a built-in name, it disables the built-in
for the remainder of the translation unit. It does this precisely to
allow a new Clang to compile old std::lib headers where a built-in
like __is_assignable was used as a normal class template, not the
built-in (because no such built-in existed at the time the library
code was written). For us, this is only really a problem when
bisecting bugs and using a newer compiler to compile .ii files from
odler headers, but for Clang combining a new Clang with older
libstdc++ headers is a hard requirement (recall that when Clang was
first deployed to macOS it had to consume the system's libstdc++ 4.2
headers).

It's a big kluge, but it would mean that a new GCC could happily
consume preprocessed code from older libstdc++ headers.
  

On Fri, 23 Sept 2022 at 15:34, Marek Polacek wrote:
>
> On Thu, Sep 22, 2022 at 06:14:44PM -0400, Jason Merrill wrote:
> > On 9/22/22 09:39, Marek Polacek wrote:
> > > To improve compile times, the C++ library could use compiler built-ins
> > > rather than implementing std::is_convertible (and _nothrow) as class
> > > templates.  This patch adds the built-ins.  We already have
> > > __is_constructible and __is_assignable, and the nothrow forms of those.
> > >
> > > Microsoft (and clang, for compatibility) also provide an alias called
> > > __is_convertible_to.  I did not add it, but it would be trivial to do
> > > so.
> > >
> > > I noticed that our __is_assignable doesn't implement the "Access checks
> > > are performed as if from a context unrelated to either type" requirement,
> > > therefore std::is_assignable / __is_assignable give two different results
> > > here:
> > >
> > >    class S {
> > >      operator int();
> > >      friend void g(); // #1
> > >    };
> > >
> > >    void
> > >    g ()
> > >    {
> > >      // #1 doesn't matter
> > >      static_assert(std::is_assignable<int&, S>::value, "");
> > >      static_assert(__is_assignable(int&, S), "");
> > >    }
> > >
> > > This is not a problem if __is_assignable is not meant to be used by
> > > the users.
> >
> > That's fine, it's not.
>
> Okay then.  libstdc++ needs to make sure then that it's handled right.

That's fine, the type traits in libstdc++ are always "a context
unrelated to either type", unless users do something idiotic like
declare std::is_assignable as a friend.

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1339r1.pdf
wants to explicitly say that's idiotic.
  

On Fri, Sep 23, 2022 at 03:40:23PM +0100, Jonathan Wakely wrote:
> On Thu, 22 Sept 2022 at 23:14, Jason Merrill wrote:
> > On 9/22/22 09:39, Marek Polacek wrote:
> > > This patch doesn't make libstdc++ use the new built-ins, but I had to
> > > rename a class otherwise its name would clash with the new built-in.
> >
> > Sigh, that's going to be a hassle when comparing compiler versions on
> > preprocessed code.
> 
> Good point. Clang has some gross hacks that we could consider. When it
> sees a declaration that uses a built-in name, it disables the built-in
> for the remainder of the translation unit. It does this precisely to
> allow a new Clang to compile old std::lib headers where a built-in
> like __is_assignable was used as a normal class template, not the
> built-in (because no such built-in existed at the time the library
> code was written). For us, this is only really a problem when
> bisecting bugs and using a newer compiler to compile .ii files from
> odler headers, but for Clang combining a new Clang with older
> libstdc++ headers is a hard requirement (recall that when Clang was
> first deployed to macOS it had to consume the system's libstdc++ 4.2
> headers).
> 
> It's a big kluge, but it would mean that a new GCC could happily
> consume preprocessed code from older libstdc++ headers.

Ah, you're right, it must be this lib/Parse/ParseDeclCXX.cpp code:

    // GNU libstdc++ 4.2 and libc++ use certain intrinsic names as the
    // name of struct templates, but some are keywords in GCC >= 4.3
    // and Clang. Therefore, when we see the token sequence "struct
    // X", make X into a normal identifier rather than a keyword, to
    // allow libstdc++ 4.2 and libc++ to work properly.
    TryKeywordIdentFallback(true);
 
Whether we want to do this, I'm not sure.

Marek
  

On 9/23/22 10:34, Marek Polacek wrote:
> On Thu, Sep 22, 2022 at 06:14:44PM -0400, Jason Merrill wrote:
>> On 9/22/22 09:39, Marek Polacek wrote:
>>> To improve compile times, the C++ library could use compiler built-ins
>>> rather than implementing std::is_convertible (and _nothrow) as class
>>> templates.  This patch adds the built-ins.  We already have
>>> __is_constructible and __is_assignable, and the nothrow forms of those.
>>>
>>> Microsoft (and clang, for compatibility) also provide an alias called
>>> __is_convertible_to.  I did not add it, but it would be trivial to do
>>> so.
>>>
>>> I noticed that our __is_assignable doesn't implement the "Access checks
>>> are performed as if from a context unrelated to either type" requirement,
>>> therefore std::is_assignable / __is_assignable give two different results
>>> here:
>>>
>>>     class S {
>>>       operator int();
>>>       friend void g(); // #1
>>>     };
>>>
>>>     void
>>>     g ()
>>>     {
>>>       // #1 doesn't matter
>>>       static_assert(std::is_assignable<int&, S>::value, "");
>>>       static_assert(__is_assignable(int&, S), "");
>>>     }
>>>
>>> This is not a problem if __is_assignable is not meant to be used by
>>> the users.
>>
>> That's fine, it's not.
>   
> Okay then.  libstdc++ needs to make sure then that it's handled right.
> 
>>> This patch doesn't make libstdc++ use the new built-ins, but I had to
>>> rename a class otherwise its name would clash with the new built-in.
>>
>> Sigh, that's going to be a hassle when comparing compiler versions on
>> preprocessed code.
> 
> Yeah, I guess :/.  Kind of like __integer_pack / __make_integer_seq.
> 
>>> --- a/gcc/cp/constraint.cc
>>> +++ b/gcc/cp/constraint.cc
>>> @@ -3697,6 +3697,12 @@ diagnose_trait_expr (tree expr, tree args)
>>>        case CPTK_HAS_UNIQUE_OBJ_REPRESENTATIONS:
>>>          inform (loc, "  %qT does not have unique object representations", t1);
>>>          break;
>>> +    case CPTK_IS_CONVERTIBLE:
>>> +      inform (loc, "  %qT is not convertible from %qE", t2, t1);
>>> +      break;
>>> +    case CPTK_IS_NOTHROW_CONVERTIBLE:
>>> +	inform (loc, "  %qT is not %<nothrow%> convertible from %qE", t2, t1);
>>
>> It's odd that the existing diagnostics quote "nothrow", which is not a
>> keyword.  I wonder why these library traits didn't use "noexcept"?
> 
> Eh, yeah, only "throw" is.  The quotes were deliberately added in
> <https://gcc.gnu.org/pipermail/gcc-patches/2019-May/522333.html>.  Should
> I prepare a separate patch to use "%<noexcept%>" rather than "%<nothrow%>"?
> OTOH, the traits have "nothrow" in their names, so maybe just go back to
> "nothrow"?

The latter, I think.  Or possibly "no-throw".  I guess -Wformat-diag 
wants "nothrow" quoted because of the attribute of that name.

>>> --- a/gcc/cp/method.cc
>>> +++ b/gcc/cp/method.cc
>>> @@ -2236,6 +2236,37 @@ ref_xes_from_temporary (tree to, tree from, bool direct_init_p)
>>>      return ref_conv_binds_directly (to, val, direct_init_p).is_false ();
>>>    }
>>> +/* Return true if FROM can be converted to TO using implicit conversions,
>>> +   or both FROM and TO are possibly cv-qualified void.  NB: This doesn't
>>> +   implement the "Access checks are performed as if from a context unrelated
>>> +   to either type" restriction.  */
>>> +
>>> +bool
>>> +is_convertible (tree from, tree to)
>>
>> You didn't want to add conversion to is*_xible?
> 
> No, it didn't look like a good fit.  It does things we don't need, and
> also has if VOID_TYPE_P -> return error_mark_node; which would be wrong
> for __is_convertible.
> 
> I realized I'm not testing passing an incomplete type to the built-in,
> but since that is UB, I reckon we don't need to test it (we issue
> "error: invalid use of incomplete type").

But your patch does test that, in the existing call to check_trait_type 
from finish_trait_expr?

The patch is OK.

Jason
  

On Fri, Sep 23, 2022 at 11:54:53AM -0400, Jason Merrill wrote:
> On 9/23/22 10:34, Marek Polacek wrote:
> > On Thu, Sep 22, 2022 at 06:14:44PM -0400, Jason Merrill wrote:
> > > On 9/22/22 09:39, Marek Polacek wrote:
> > > > To improve compile times, the C++ library could use compiler built-ins
> > > > rather than implementing std::is_convertible (and _nothrow) as class
> > > > templates.  This patch adds the built-ins.  We already have
> > > > __is_constructible and __is_assignable, and the nothrow forms of those.
> > > > 
> > > > Microsoft (and clang, for compatibility) also provide an alias called
> > > > __is_convertible_to.  I did not add it, but it would be trivial to do
> > > > so.
> > > > 
> > > > I noticed that our __is_assignable doesn't implement the "Access checks
> > > > are performed as if from a context unrelated to either type" requirement,
> > > > therefore std::is_assignable / __is_assignable give two different results
> > > > here:
> > > > 
> > > >     class S {
> > > >       operator int();
> > > >       friend void g(); // #1
> > > >     };
> > > > 
> > > >     void
> > > >     g ()
> > > >     {
> > > >       // #1 doesn't matter
> > > >       static_assert(std::is_assignable<int&, S>::value, "");
> > > >       static_assert(__is_assignable(int&, S), "");
> > > >     }
> > > > 
> > > > This is not a problem if __is_assignable is not meant to be used by
> > > > the users.
> > > 
> > > That's fine, it's not.
> > Okay then.  libstdc++ needs to make sure then that it's handled right.
> > 
> > > > This patch doesn't make libstdc++ use the new built-ins, but I had to
> > > > rename a class otherwise its name would clash with the new built-in.
> > > 
> > > Sigh, that's going to be a hassle when comparing compiler versions on
> > > preprocessed code.
> > 
> > Yeah, I guess :/.  Kind of like __integer_pack / __make_integer_seq.
> > 
> > > > --- a/gcc/cp/constraint.cc
> > > > +++ b/gcc/cp/constraint.cc
> > > > @@ -3697,6 +3697,12 @@ diagnose_trait_expr (tree expr, tree args)
> > > >        case CPTK_HAS_UNIQUE_OBJ_REPRESENTATIONS:
> > > >          inform (loc, "  %qT does not have unique object representations", t1);
> > > >          break;
> > > > +    case CPTK_IS_CONVERTIBLE:
> > > > +      inform (loc, "  %qT is not convertible from %qE", t2, t1);
> > > > +      break;
> > > > +    case CPTK_IS_NOTHROW_CONVERTIBLE:
> > > > +	inform (loc, "  %qT is not %<nothrow%> convertible from %qE", t2, t1);
> > > 
> > > It's odd that the existing diagnostics quote "nothrow", which is not a
> > > keyword.  I wonder why these library traits didn't use "noexcept"?
> > 
> > Eh, yeah, only "throw" is.  The quotes were deliberately added in
> > <https://gcc.gnu.org/pipermail/gcc-patches/2019-May/522333.html>.  Should
> > I prepare a separate patch to use "%<noexcept%>" rather than "%<nothrow%>"?
> > OTOH, the traits have "nothrow" in their names, so maybe just go back to
> > "nothrow"?
> 
> The latter, I think.  Or possibly "no-throw".  I guess -Wformat-diag wants
> "nothrow" quoted because of the attribute of that name.

OK, let me see.
 
> > > > --- a/gcc/cp/method.cc
> > > > +++ b/gcc/cp/method.cc
> > > > @@ -2236,6 +2236,37 @@ ref_xes_from_temporary (tree to, tree from, bool direct_init_p)
> > > >      return ref_conv_binds_directly (to, val, direct_init_p).is_false ();
> > > >    }
> > > > +/* Return true if FROM can be converted to TO using implicit conversions,
> > > > +   or both FROM and TO are possibly cv-qualified void.  NB: This doesn't
> > > > +   implement the "Access checks are performed as if from a context unrelated
> > > > +   to either type" restriction.  */
> > > > +
> > > > +bool
> > > > +is_convertible (tree from, tree to)
> > > 
> > > You didn't want to add conversion to is*_xible?
> > 
> > No, it didn't look like a good fit.  It does things we don't need, and
> > also has if VOID_TYPE_P -> return error_mark_node; which would be wrong
> > for __is_convertible.
> > 
> > I realized I'm not testing passing an incomplete type to the built-in,
> > but since that is UB, I reckon we don't need to test it (we issue
> > "error: invalid use of incomplete type").
> 
> But your patch does test that, in the existing call to check_trait_type from
> finish_trait_expr?

Yes, it eventually checks complete_type_or_else.

> The patch is OK.

Thanks,

Marek
  

On Fri, 23 Sept 2022 at 15:43, Jonathan Wakely wrote:
>
> On Fri, 23 Sept 2022 at 15:34, Marek Polacek wrote:
> >
> > On Thu, Sep 22, 2022 at 06:14:44PM -0400, Jason Merrill wrote:
> > > On 9/22/22 09:39, Marek Polacek wrote:
> > > > To improve compile times, the C++ library could use compiler built-ins
> > > > rather than implementing std::is_convertible (and _nothrow) as class
> > > > templates.  This patch adds the built-ins.  We already have
> > > > __is_constructible and __is_assignable, and the nothrow forms of those.
> > > >
> > > > Microsoft (and clang, for compatibility) also provide an alias called
> > > > __is_convertible_to.  I did not add it, but it would be trivial to do
> > > > so.
> > > >
> > > > I noticed that our __is_assignable doesn't implement the "Access checks
> > > > are performed as if from a context unrelated to either type" requirement,
> > > > therefore std::is_assignable / __is_assignable give two different results
> > > > here:
> > > >
> > > >    class S {
> > > >      operator int();
> > > >      friend void g(); // #1
> > > >    };
> > > >
> > > >    void
> > > >    g ()
> > > >    {
> > > >      // #1 doesn't matter
> > > >      static_assert(std::is_assignable<int&, S>::value, "");
> > > >      static_assert(__is_assignable(int&, S), "");
> > > >    }
> > > >
> > > > This is not a problem if __is_assignable is not meant to be used by
> > > > the users.
> > >
> > > That's fine, it's not.
> >
> > Okay then.  libstdc++ needs to make sure then that it's handled right.
>
> That's fine, the type traits in libstdc++ are always "a context
> unrelated to either type", unless users do something idiotic like
> declare std::is_assignable as a friend.
>
> https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1339r1.pdf
> wants to explicitly say that's idiotic.

And I just checked that a variable template like std::is_assignable_v
also counts as "a context unrelated to either type", even when
instantiated inside a member function of the type:

#include <type_traits>

template<typename T, typename U>
constexpr bool is_assignable_v = __is_assignable(T, U);

class S {
  operator int();
  friend void g(); // #1
};

void
g ()
{
  // #1 doesn't matter
  static_assert(std::is_assignable<int&, S>::value, "");
  static_assert(std::is_assignable_v<int&, S>, "");
  static_assert(__is_assignable(int&, S), "");
}

The first two assertions are consistent, and fail, which is what we
want.  The direct use of the built-in succeeds, but we don't care.
  

On Fri, Sep 23, 2022 at 05:34:21PM +0100, Jonathan Wakely wrote:
> On Fri, 23 Sept 2022 at 15:43, Jonathan Wakely wrote:
> >
> > On Fri, 23 Sept 2022 at 15:34, Marek Polacek wrote:
> > >
> > > On Thu, Sep 22, 2022 at 06:14:44PM -0400, Jason Merrill wrote:
> > > > On 9/22/22 09:39, Marek Polacek wrote:
> > > > > To improve compile times, the C++ library could use compiler built-ins
> > > > > rather than implementing std::is_convertible (and _nothrow) as class
> > > > > templates.  This patch adds the built-ins.  We already have
> > > > > __is_constructible and __is_assignable, and the nothrow forms of those.
> > > > >
> > > > > Microsoft (and clang, for compatibility) also provide an alias called
> > > > > __is_convertible_to.  I did not add it, but it would be trivial to do
> > > > > so.
> > > > >
> > > > > I noticed that our __is_assignable doesn't implement the "Access checks
> > > > > are performed as if from a context unrelated to either type" requirement,
> > > > > therefore std::is_assignable / __is_assignable give two different results
> > > > > here:
> > > > >
> > > > >    class S {
> > > > >      operator int();
> > > > >      friend void g(); // #1
> > > > >    };
> > > > >
> > > > >    void
> > > > >    g ()
> > > > >    {
> > > > >      // #1 doesn't matter
> > > > >      static_assert(std::is_assignable<int&, S>::value, "");
> > > > >      static_assert(__is_assignable(int&, S), "");
> > > > >    }
> > > > >
> > > > > This is not a problem if __is_assignable is not meant to be used by
> > > > > the users.
> > > >
> > > > That's fine, it's not.
> > >
> > > Okay then.  libstdc++ needs to make sure then that it's handled right.
> >
> > That's fine, the type traits in libstdc++ are always "a context
> > unrelated to either type", unless users do something idiotic like
> > declare std::is_assignable as a friend.
> >
> > https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1339r1.pdf
> > wants to explicitly say that's idiotic.
> 
> And I just checked that a variable template like std::is_assignable_v
> also counts as "a context unrelated to either type", even when
> instantiated inside a member function of the type:
> 
> #include <type_traits>
> 
> template<typename T, typename U>
> constexpr bool is_assignable_v = __is_assignable(T, U);
> 
> class S {
>   operator int();
>   friend void g(); // #1
> };
> 
> void
> g ()
> {
>   // #1 doesn't matter
>   static_assert(std::is_assignable<int&, S>::value, "");
>   static_assert(std::is_assignable_v<int&, S>, "");
>   static_assert(__is_assignable(int&, S), "");
> }
> 
> The first two assertions are consistent, and fail, which is what we
> want.  The direct use of the built-in succeeds, but we don't care.

Great, thanks. 

Marek