On 9/9/25 2:15 PM, Tomasz Kaminski wrote:
On Tue, Sep 9, 2025 at 12:45 PM Jonathan Wakely <[email protected]> wrote:
On Fri, 5 Sept 2025 at 08:14, Tomasz Kamiński <[email protected]> wrote:
From: Luc Grosheintz <[email protected]>
This is a partial implementation of P2781R9. It adds std::cw and
std::constant_wrapper, but doesn't modify __integral_constant_like for
span/mdspan.
libstdc++-v3/ChangeLog:
* include/bits/version.def (constant_wrapper): Add.
* include/bits/version.h: Regenerate.
* include/std/type_traits (_CwFixedValue): New class.
(_IndexSequence): New struct.
(_BuildIndexSequence): New struct.
(_ConstExprParam): New concept.
(_CwOperators): New struct.
(constant_wrapper): New struct.
(cw): New global constant.
* src/c++23/std.cc.in (constant_wrapper): Add.
(cw): Add.
* testsuite/20_util/constant_wrapper/adl.cc: New test.
* testsuite/20_util/constant_wrapper/ex.cc: New test.
* testsuite/20_util/constant_wrapper/generic.cc: New test.
* testsuite/20_util/constant_wrapper/instantiate.cc: New test.
* testsuite/20_util/constant_wrapper/op_comma_neg.cc: New test.
* testsuite/20_util/constant_wrapper/version.cc: New test.
Co-authored-by: Tomasz Kamiński <[email protected]>
Signed-off-by: Luc Grosheintz <[email protected]>
Signed-off-by: Tomasz Kamiński <[email protected]>
---
v5 adds data member pointer tests
Some late comments that I should have noticed before approving it, sorry!
diff --git a/libstdc++-v3/include/std/type_traits
b/libstdc++-v3/include/std/type_traits
index 4636457eb5a..26cbbb4fd5b 100644
--- a/libstdc++-v3/include/std/type_traits
+++ b/libstdc++-v3/include/std/type_traits
@@ -41,6 +41,7 @@
#define __glibcxx_want_bool_constant
#define __glibcxx_want_bounded_array_traits
+#define __glibcxx_want_constant_wrapper
#define __glibcxx_want_has_unique_object_representations
#define __glibcxx_want_integral_constant_callable
#define __glibcxx_want_is_aggregate
@@ -4302,6 +4303,376 @@ template<typename _Ret, typename _Fn,
typename... _Args>
};
#endif // C++11
+#ifdef __cpp_lib_constant_wrapper // C++ >= 26
+ template<typename _Tp>
+ struct _CwFixedValue
+ {
+ using _S_type = _Tp;
The "_S_" prefix is for static members, this should be just "__type".
I'll change this.
+
+ template<_CwFixedValue _Tp,
+ typename = typename decltype(_CwFixedValue(_Tp))::_S_type>
+ struct constant_wrapper;
Why do we need decltype(_CwFixedValue(_Tp)) here?
It looks to me like just decltype(_Tp)::_S_type should work ... but it
doesn't. Is that something to do with CTAD being involved?
I think this is a GCC bug, that got a workaround backed in the standard
(the standard specifies the argument this way, this is why it haven't
picked up in review). Just decltype(_Tp) works in clang.
This ran a bell:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117392
+
+ template<_CwFixedValue _X, typename>
+ struct constant_wrapper : _CwOperators
_X is (or was, at some point) used by libc headers on at least one
UNIX variant. I've already pushed a change to make that _Xv instead.
This is the same reason that we use _Tp instead of _T.
I think I'll change the _Tp in the constant_wrapper declaration to
match _Xv, since it's not a type, it's a constant template parameter
(what used to be called non-type template parameter until
https://github.com/cplusplus/draft/pull/7587 renamed it).
