The same way VLAs in C11 are optional - the `constexpr goto` and non
mandatory `constexpr` standard library implementation (which should exists
in the first place as duplicate of the non `constexpr` one) will be too,
the mandatory `constexpr` library would include the methods required as
`constexpr` until now.

Then the next standard version - you could check what implementations have
implemented as optional and require that - but also for every new feature
added to the language add a `constexpr` version as optional feature.

On Thu, May 25, 2023 at 12:04 PM sasho648 <sasho648_at_[hidden]> wrote:

> Yeah but in my opinion is a lot of duplicated work - instead of just
> mandating that each and every C++ feature can be constexpr and only keeping
> up to date the minimum required set.
>
> Which means - introducing `constexpr goto` from the beginning and now
> instead of guessing what an implementation may be required to do to allow
> it - just looking at existing implementations that have allowed it.
>
> Same for the standard library - having each and every method by default a
> constexpr duplicate - only mandating a minimum required methods to be
> available. And then slowly extending this.
>
> In other words adding new features to the standard should require to add a
> thought for the same feature as a constexpr - and then you could talk all
> you want what the minimum set should be - but for the next iteration you
> will have actual data of implementations.
>
> On Thu, May 25, 2023 at 2:35 AM Jason McKesson via Std-Proposals <
> std-proposals_at_[hidden]> wrote:
>
>> On Wed, May 24, 2023 at 12:00 PM sasho648 via Std-Proposals
>> <std-proposals_at_[hidden]> wrote:
>> >
>> > Weird how that should even be concern of the committee. Just give the
>> syntax and allow implementations to decide what can they implement as
>> `constexpr` or not.
>> >
>> > Maybe only impose "minimal requirements" for what needs to be able to
>> be evaluated at compile time.
>>
>> ... what exactly would be the point of this? I mean, right now,
>> compilers can evaluate *anything* they want at compile-time. Nothing
>> is stopping them.
>>
>> `constexpr` is not for "maybe my compiler can do something with this".
>> It's for "when I call this in a place that *requires* a constant
>> expression, evaluate it at compile-time". It's not a statement of a
>> preference; it's a *requirement*, one that places burdens on the
>> compiler.
>>
>> If you cannot rely on `constexpr` actually being a constant expression
>> when you need it to be, then what is the point of the feature?
>> --
>> Std-Proposals mailing list
>> Std-Proposals_at_[hidden]
>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>
>