[libcxx-commits] [PATCH] D67900: [libc++] Use builtin type traits whenever possible

[Eric Fiselier via libcxx-commits]

We should revert the change for `is_arithmetic` and `is_floating_point`.

Marshall and I have discussed the __float128 issue previously, but we
couldn't reach an agreement that it should be handled as an extension.
For that reason, I think we should go back to the old behavior.

/Eric

On Wed, Apr 1, 2020 at 1:00 AM Zoe Carver via Phabricator <
reviews at reviews.llvm.org> wrote:

> zoecarver added a comment.
>
> @tcwang here's you're reproducer: https://godbolt.org/z/RJtTBG
>
> Looks like we don't have any `__float128` tests not only in numeric limits
> but in libc++ altogether.
>
> Anyway, to address the problem: `is_arithmetic_v<__float128>` is true
> after this patch and false before this patch. It seems like returning true
> is the correct behavior so the change should probably happen in numeric
> limits. It's not entirely clear what that change should be, though. Do we
> want to support `__float128` in `numeric_limits`? We don't seem to support
> it anywhere else. Maybe add a `static_assert` so that future bugs are
> easier to track down?
>
> In the meantime, `numeric_limits` just returned `0` for `numeric_limits<
> __float128 >::min()` so maybe you can replace the current use with that?
>
>
> Repository:
>   rG LLVM Github Monorepo
>
> CHANGES SINCE LAST ACTION
>   https://reviews.llvm.org/D67900/new/
>
> https://reviews.llvm.org/D67900
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/libcxx-commits/attachments/20200401/564fd9ab/attachment-0001.html>