Can't convert cpp_dec_float to cpp_int

[radj307]

I'm using cpp_int and cpp_dec_float_100 in a project, but I'm encountering an error when trying to convert from float to int on both MSVC 19.34.31947 and gcc 10, using boostmp 1.84 standalone.

#include <boost/multiprecision/cpp_int.hpp>
#include <boost/multiprecision/cpp_dec_float.hpp>

namespace mp = boost::multiprecision;

int main()
{
    using mp::cpp_int;
    using cpp_float = mp::cpp_dec_float_100;

    cpp_float r1{ static_cast<cpp_float>(cpp_int{ 5 }) }; //< works
    cpp_int r2{ static_cast<cpp_int>(cpp_float{ 5.5 }) }; //< ERROR: (see below)

    return 0;
}

This is the error message in GCC 10 (formatted for readability):

operands to ‘?:’ have different types
‘std::enable_if<true, boost::multiprecision::detail::expression<boost::multiprecision::detail::function, boost::multiprecision::detail::number_kind_floating_pointfloor_funct<boost::multiprecision::backends::cpp_dec_float<100> >, boost::multiprecision::number<boost::multiprecision::backends::cpp_dec_float<100> >, void, void> >::type’
  {aka ‘boost::multiprecision::detail::expression<boost::multiprecision::detail::function, boost::multiprecision::detail::number_kind_floating_pointfloor_funct<boost::multiprecision::backends::cpp_dec_float<100> >, boost::multiprecision::number<boost::multiprecision::backends::cpp_dec_float<100> >, void, void>’}
and
‘std::enable_if<true, boost::multiprecision::detail::expression<boost::multiprecision::detail::function, boost::multiprecision::detail::number_kind_floating_pointceil_funct<boost::multiprecision::backends::cpp_dec_float<100> >, boost::multiprecision::number<boost::multiprecision::backends::cpp_dec_float<100> >, void, void> >::type’
  {aka ‘boost::multiprecision::detail::expression<boost::multiprecision::detail::function, boost::multiprecision::detail::number_kind_floating_pointceil_funct<boost::multiprecision::backends::cpp_dec_float<100> >, boost::multiprecision::number<boost::multiprecision::backends::cpp_dec_float<100> >, void, void>’}
(...)\boost\multiprecision\detail\functions\trunc.hpp

The error does not occur when using Boost::math alongside it.

[ckormanyos]

I'm encountering an error when trying to convert

Hi @radj307 great catch and thanks for the report.

I can reproduce this situation exactly with modular-boost on today's develop-branch as well.

The phenomenon seems to arise from detail/functions/trunc.hpp line 29.

The operator ? mentioned is the ternary operator at that line where the arguments (i.e., the compiler's decuced result types from floorand ceil) differ. One is the backend type alone whereby the other is the fully number-wrapped backend type.

I was not, sadly, alone able to find the root cause.

But if it helps investigations, rewriting impl::trunc() in the following way eliminates the problem. But this is not a great solution suggestoin. I'd like to see why the overloads have differences. But alone I couldn't figure it out (yet).

Another detail, this truncis called from lltrunc in the same header around line 59. So that's probably closer to the place where the arguments for the overloads are determined, probably even higher upstream in the call chain actually.

template <typename T>
inline T trunc BOOST_PREVENT_MACRO_SUBSTITUTION (const T arg)
{
    using std::floor;
    using std::ceil;

    T t { };

    if (arg > T(0))
    {
        t = floor(arg);
    }
    else
    {
        t = ceil(arg);
    }

    return t;
}

Cc: @mborland and @jzmaddock

[ckormanyos]

See also #587, which might not be the last word on this, but offers a preliminary (perhaps legitimate, long-lasting) fix.

Cc: @jzmaddock and @mborland

[ckormanyos]

Fixed in #587