[PATCH v1] libstdc++: Rename _S-prefixed identifiers in <mdspan>.

Luc Grosheintz luc.grosheintz@gmail.com
Wed Sep 10 10:08:51 GMT 2025



On 9/10/25 11:24 AM, Jonathan Wakely wrote:
> On Wed, 10 Sept 2025 at 09:44, Jonathan Wakely <jwakely@redhat.com> wrote:
>>
>> On Wed, 10 Sept 2025 at 07:49, Luc Grosheintz <luc.grosheintz@gmail.com> wrote:
>>>
>>> In libstdc++ the prefix _S is used for static members only. In <mdspan>
>>> there's several type aliases that also used the prefix _S. They now use
>>> double underscore instead.
>>
>> Thanks for revisiting these.
>>
>>>
>>> libstdc++-v3/ChangeLog:
>>>
>>>          * include/std/mdspan (_ExtentsStorage::__base): New name for
>>>          _S_base.
>>>          (_ExtentsStorage::__storage_type): New name for _S_storage.
>>>          (extents::__storage_type): New name for _S_storage.
>>>          (layout_stride::mapping::__strides_type): New name for
>>>          _S_stries_t.
>>>          * testsuite/23_containers/mdspan/class_mandate_neg.cc: Update
>>>          test to the new error message.
>>>
>>> Signed-off-by: Luc Grosheintz <luc.grosheintz@gmail.com>
>>> ---
>>>   libstdc++-v3/include/std/mdspan               | 33 ++++++++++---------
>>>   .../23_containers/mdspan/class_mandate_neg.cc |  2 +-
>>>   2 files changed, 18 insertions(+), 17 deletions(-)
>>>
>>> diff --git a/libstdc++-v3/include/std/mdspan b/libstdc++-v3/include/std/mdspan
>>> index 678b2619ebe..f0139a95b11 100644
>>> --- a/libstdc++-v3/include/std/mdspan
>>> +++ b/libstdc++-v3/include/std/mdspan
>>> @@ -148,14 +148,14 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
>>>         class _ExtentsStorage : public _StaticExtents<_Extents>
>>>         {
>>>         private:
>>> -       using _S_base = _StaticExtents<_Extents>;
>>> +       using __base = _StaticExtents<_Extents>;
>>
>> Looking through the other headers, we usually use _Base for this case,
>> and for other typedefs that are private to the class (and its
>> friends).

I'll redo them all (seems odd to only do half).

>>
>> Yes, that's inconsistent with __type that we use in some internal type
>> traits, but in those cases __type is public and we intend it to be
>> used from outside the class, by code using it. e.g. constant_wrapper
>> uses _CwFixedValue::__type.
> 
> I should make some updates to
> https://gcc.gnu.org/onlinedocs/libstdc++/manual/source_code_style.html
> to cover this.
> 
> And turn that numbered list into a proper DocBook <variablelist>, and
> remove number 10 (about prefixing members with 'this->' because of
> Koenig lookup, which is just wrong).
> 

I'd like to suggest the following for the long example:

   - add a using _SomeType = ...; (or _Some_type),
   - an example of attributes like [[nodiscard]] (I think it's on its
   own line),
   - an example with requires,

   - an example of trailing return type (-> ret_type): once for when
   it doesn't have to break and once where it wont fit.

Here's some things I'm not fully convinced are correct:
   - in the example I think the
     if (test)
     {
       nested code
     }
     is wrong because we need to indent the braces by one level.

   - double check the for-loop, if we put the { on the same line,
   I need to do some fixups for <mdspan>.

   - In
       function_name(char* pointer,   // "char *pointer" is wrong.
                     char* argument,
                     const Reference& ref)

     I think we're expected to squash as much onto one line as we can?
       function_name(char* pointer, char* argument, // not: char *pointer
                     const Reference& ref)

   -  // In-class function definitions should be restricted to
      // one-liners.

   I'm seeing lots of violation of that rule.

I remember liking and referring back to the example a lot in the
beginning.


More information about the Libstdc++ mailing list