Re: boost::math::tools::promote_args not works with C++23 FP-types

From: Matt Borland (matt_at_[hidden])
Date: 2023-02-19 14:58:52

On Sun, Feb 19, 2023 at 06:48, Gero Peterhoff via Boost <boost_at_[hidden]> wrote:

> Hi,
> promote_args from <boost/math/tools/promotion.hpp> not works with new/extended C++23 FP-types
> https://godbolt.org/z/13fq8a3d7
>
> The problem might be the outdated "manual variadic" implementation for C++03. Well, according to the documentation (https://www.boost.org/doc/libs/1_81_0), boost::math requires C++11 anyway; with a switch to C++11 it also works (std::common_type).
>
> This issue also affects some implementations in boost::math::policies.
>
> PS:
> We have (at least) 3 different definitions for the *same* boost::float128(_type/_t) type
> 1) <boost/config.hpp>: boost::float128_type
> 2) <boost/math/cstdfloat/cstdfloat_types.hpp>/<boost/math/cstdfloat/cstdfloat_cmath.hpp>: BOOST_CSTDFLOAT_FLOAT128_NATIVE_TYPE/float128_t
> 3) <boost/math/tools/config.hpp>: BOOST_MATH_FLOAT128_TYPE
> :-(
>
> thx
> Gero
>
> _______________________________________________
> Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

Gero,

Assuming GCC follows its standard release cadence we should have 13 in the April timeframe to test with fixed width floating point types. We are aware of the potential friction these bring. The fixed width floating point types are explicitly not allowed to be typedefs of the existing floating point types so trivial solutions like common type will likely fail. The promotion rules with these types are also much more complex so it’ll take some time to work out the bugs.

Matt