[PATCH v1 0/1] Prepare for and double check LWG4272.
Luc Grosheintz
luc.grosheintz@gmail.com
Tue Sep 9 07:27:05 GMT 2025
On 9/9/25 9:03 AM, Tomasz Kaminski wrote:
> Hi,
>
> On the LWG reflector the issues was found with the original PR where for
> the padded layouts,
> we have had:
>
> !is_convertible_v<typename LayoutRightPaddedMapping::extents_type,
> extents_type> &&
> rank_ > 1 &&
> (padding_value != dynamic_extent ||
> LayoutRightPaddedMapping::padding_value == dynamic_extent)
>
> Where it should be instead:
>
> !is_convertible_v<typename LayoutLeftPaddedMapping::extents_type,
> extents_type> ||
> rank_> 1 &&
> (padding_value != dynamic_extent ||
> LayoutLeftPaddedMapping::padding_value == dynamic_extent)
>
> Would you mind rebasing this patch on top of left padded layout to confirm
> that this is now working correctly?
I saw this yesterday. Thanks to the person that found it and thank you
for updating the issue.
That seems like a good idea (and was on the list of things to maybe do
next).
>
> On Thu, Aug 21, 2025 at 6:18 PM Tomasz Kaminski <tkaminsk@redhat.com> wrote:
>
>> The LWG4274 was prioritized on reflector, and got assigned priority P2,
>> however was not yet approved. I will recheck/review this patch next week,
>> and
>> will start another pool on reflector, with mention of implementation
>> experience.
>>
>> On Tue, Jun 10, 2025 at 9:21 PM Luc Grosheintz <luc.grosheintz@gmail.com>
>> wrote:
>>
>>> This patch implements LWG4272. LWG4272 has not been discussed
>>> or approved. This patch serves to demonstrate that LWG4272 leads
>>> to uniform behaviour of ctors when rank == 0.
>>>
>>> This isn't patch doesn't add any _GLIBCXX_RESOLVE_LIB_DEFECTS
>>> markers, and isn't necessarily intended to be merged (as is).
>>>
>>> Luc Grosheintz (1):
>>> libstdc++: Convertibility of rank == 0 layouts, LWG4272.
>>>
>>> libstdc++-v3/include/std/mdspan | 6 +-
>>> .../23_containers/mdspan/layouts/ctors.cc | 60 ++++++++++++++++++-
>>> 2 files changed, 62 insertions(+), 4 deletions(-)
>>>
>>> --
>>> 2.49.0
>>>
>>>
>
More information about the Libstdc++
mailing list