[PATCH] libstdc++: fix C++17 regression in concept __heterogeneous_key (2nd)

[Tomasz Kaminski]

On Fri, Feb 6, 2026 at 11:54 AM Jonathan Wakely <jwakely.gcc@gmail.com>
wrote:

>
>
> On Fri, 6 Feb 2026, 10:34 Tomasz Kaminski, <tkaminsk@redhat.com> wrote:
>
>>
>>
>> On Fri, Feb 6, 2026 at 10:14 AM Nathan Myers <ncm@cantrip.org> wrote:
>>
>>> The commit 3f7905550483408a2c4c5096a1adc8d7e863eb12 defined a
>>> concept __heterogeneous_key using a name not defined in C++17.
>>> This is fixed by guarding the definition behind a name defined
>>> in C++23 the earliest Standard that uses the definition.
>>>
>>> https://gcc.gnu.org/pipermail/gcc-patches/2026-February/707814.html
>>>
>>> libstdc++-v3/ChangeLog
>>>         * include/bits/stl_function.h (concept __heterogeneous_key):
>>>         Guard with __glibcxx_associative_heterogeneous_erasure
>>> ---
>>>
>> LGTM, defining the concept only if we provide extensions that need them
>> is reasonable. We can always change that if more places will depend on
>> that
>> and conditions will become unwieldy.
>>
>> I think you can commit that as an obvious fix, without need for Jonathan
>> approval.
>>
>
> Yes, I think it was already pushed.
>
I spoke to early, and assume the change was actually tested, but the tests
fails
because we need to update guards concepts:
diff --git a/libstdc++-v3/include/bits/hashtable.h
b/libstdc++-v3/include/bits/hashtable.h
index f7b8905d8ab..48695c013f3 100644
--- a/libstdc++-v3/include/bits/hashtable.h
+++ b/libstdc++-v3/include/bits/hashtable.h
@@ -3012,7 +3012,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
       = __enable_if_t<!__or_<is_integral<_Hash>,
__is_allocator<_Hash>>::value>;
 #endif

-#ifdef __cpp_concepts
+#ifdef __glibcxx_associative_heterogeneous_erasure // C++ >= 23
 template <typename _Kt, typename _Container>
   concept __heterogeneous_hash_key =
     __transparent_comparator<typename _Container::hasher> &&
diff --git a/libstdc++-v3/include/bits/stl_tree.h
b/libstdc++-v3/include/bits/stl_tree.h
index 7a733241ec5..5d361b55028 100644
--- a/libstdc++-v3/include/bits/stl_tree.h
+++ b/libstdc++-v3/include/bits/stl_tree.h
@@ -3315,7 +3315,7 @@ namespace __rb_tree
     };
 #endif // C++17

-#ifdef __cpp_concepts
+#ifdef __glibcxx_associative_heterogeneous_erasure // C++ >= 23
 template <typename _Kt, typename _Container>
   concept __heterogeneous_tree_key =
     __transparent_comparator<typename _Container::key_compare> &

Testing the change, and will push it once all tests passed.

>
>
>
>>  libstdc++-v3/include/bits/stl_function.h | 5 ++---
>>>  1 file changed, 2 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/libstdc++-v3/include/bits/stl_function.h
>>> b/libstdc++-v3/include/bits/stl_function.h
>>> index 657f40bca2e..1312564216b 100644
>>> --- a/libstdc++-v3/include/bits/stl_function.h
>>> +++ b/libstdc++-v3/include/bits/stl_function.h
>>>
>> The file does not include bits/version.h, so we assume that it is
>> included before,
>> and we already use __glibcxx_transparent_operators in the file, so this
>> is not a problem
>> of this change. I have looked, and we bring bits version via type_traits
>> inclusion.
>>
>>
>>> @@ -1525,7 +1525,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
>>>  #endif
>>>  #endif
>>>
>>> -#ifdef __cpp_concepts
>>> +#ifdef __glibcxx_associative_heterogeneous_erasure  // C++23
>>>  template <typename _Kt, typename _Container>
>>>    concept __not_container_iterator =
>>>      (!is_convertible_v<_Kt&&, typename _Container::iterator> &&
>>> @@ -1533,8 +1533,7 @@ template <typename _Kt, typename _Container>
>>>
>>>  template <typename _Kt, typename _Container>
>>>    concept __heterogeneous_key =
>>> -    (!is_same_v<typename _Container::key_type,
>>> -               typename remove_cvref<_Kt>::type>) &&
>>> +    (!is_same_v<typename _Container::key_type, remove_cvref_t<_Kt>>) &&
>>>      __not_container_iterator<_Kt, _Container>;
>>>  #endif
>>>
>>> --
>>> 2.52.0
>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://gcc.gnu.org/pipermail/libstdc++/attachments/20260206/4b937115/attachment.htm>