boost/multiprecision/number.hpp:52:17: error: no viable overloaded '='

[Jeff Hammond <jeff.science@…>]

Summary

The following simple program attempts to make use of simple features in Boost Multiprecision and Math. I have no problem using constants::pi with double, extended double, quad precision, or even a user-defined version of double double precision. However, when I try to get a single precision version of this constant, it fails.

Source Code

#include <iostream>
#include <utility>

#include <boost/math/constants/constants.hpp>
#include <boost/multiprecision/cpp_bin_float.hpp>

using boost::multiprecision::cpp_bin_float_single;
using boost::multiprecision::cpp_bin_float_double;

typedef boost::multiprecision::cpp_bin_float_single f32;
typedef boost::multiprecision::cpp_bin_float_double f64;

int main(void)
{
#if BUGTEST_PRECISION == 32
    //f32 pi32 = boost::math::constants::pi<f32, boost::math::policies::policy<boost::math::policies::digits2<32> > >();
    f32 pi32 = boost::math::constants::pi<f32>();
    std::cout << pi32 << std::endl;
#elif BUGTEST_PRECISION == 64
    //f64 pi64 = boost::math::constants::pi<f64, boost::math::policies::policy<boost::math::policies::digits2<64> > >();
    f64 pi64 = boost::math::constants::pi<f64>();
    std::cout << pi64 << std::endl;
#else
#error Set BUGTEST_PRECISION to 32 or 64.
#endif
    return 0;
}

Compilers Tested

Mac Yosemite (10.10.5):

• Apple LLVM version 7.0.0 (clang++)
• Clang 3.6.2 from Homebrew
• GCC 4.2.1 ({g++})
• GCC 5.2.0 from Homebrew
• Intel 16.0.0

Representative Output

Clang and GCC show nearly identical errors. The Intel one is much shorter but seemingly equivalent.

jrhammon-mac01:boost jrhammon$ clang++ -g -Wall -std=c++11 -stdlib=libc++ -DBUGTEST_PRECISION=64 bug.cc -o bug.x
jrhammon-mac01:boost jrhammon$ clang++ -g -Wall -std=c++11 -stdlib=libc++ -DBUGTEST_PRECISION=32 bug.cc -o bug.x
In file included from bug.cc:5:
In file included from /usr/local/include/boost/multiprecision/cpp_bin_float.hpp:9:
In file included from /usr/local/include/boost/multiprecision/cpp_int.hpp:12:
/usr/local/include/boost/multiprecision/number.hpp:52:17: error: no viable overloaded '='
      m_backend = canonical_value(v);
      ~~~~~~~~~ ^ ~~~~~~~~~~~~~~~~~~
/usr/local/include/boost/math/constants/constants.hpp:289:3: note: in instantiation of function template specialization
      'boost::multiprecision::number<boost::multiprecision::backends::cpp_bin_float<24,
      boost::multiprecision::backends::digit_base_type::digit_base_2, void, short, -126, 127>,
      boost::multiprecision::expression_template_option::et_off>::number<float>' requested here
  BOOST_DEFINE_MATH_CONSTANT(pi, 3.141592653589793238462643383279502884e+00, "3.141592653589793238462643383279502884197169...
  ^
/usr/local/include/boost/math/constants/constants.hpp:246:13: note: expanded from macro 'BOOST_DEFINE_MATH_CONSTANT'
   { return BOOST_JOIN(x, F); }\
            ^
/usr/local/include/boost/config/suffix.hpp:544:28: note: expanded from macro 'BOOST_JOIN'
#define BOOST_JOIN( X, Y ) BOOST_DO_JOIN( X, Y )
                           ^
/usr/local/include/boost/config/suffix.hpp:545:31: note: expanded from macro 'BOOST_DO_JOIN'
#define BOOST_DO_JOIN( X, Y ) BOOST_DO_JOIN2(X,Y)
                              ^
/usr/local/include/boost/config/suffix.hpp:546:32: note: expanded from macro 'BOOST_DO_JOIN2'
#define BOOST_DO_JOIN2( X, Y ) X##Y
                               ^
<scratch space>:14:1: note: expanded from here
3.141592653589793238462643383279502884e+00F
^
/usr/local/include/boost/multiprecision/cpp_bin_float.hpp:107:19: note: candidate function not viable: no known conversion from
      'typename boost::disable_if<is_same<typename detail::canonical<float, cpp_bin_float<24,
      boost::multiprecision::backends::digit_base_type::digit_base_2, void, short, -126, 127> >::type, float>, typename
      detail::canonical<float, cpp_bin_float<24, boost::multiprecision::backends::digit_base_type::digit_base_2, void, short,
      -126, 127> >::type>::type' (aka 'double') to 'const boost::multiprecision::backends::cpp_bin_float<24,
      boost::multiprecision::backends::digit_base_type::digit_base_2, void, short, -126, 127>' for 1st argument
   cpp_bin_float& operator=(const cpp_bin_float &o) BOOST_NOEXCEPT_IF(noexcept(std::declval<rep_type&>() = std::declval<co...
                  ^
/usr/local/include/boost/multiprecision/cpp_bin_float.hpp:303:19: note: candidate function not viable: no known conversion from
      'typename boost::disable_if<is_same<typename detail::canonical<float, cpp_bin_float<24,
      boost::multiprecision::backends::digit_base_type::digit_base_2, void, short, -126, 127> >::type, float>, typename
      detail::canonical<float, cpp_bin_float<24, boost::multiprecision::backends::digit_base_type::digit_base_2, void, short,
      -126, 127> >::type>::type' (aka 'double') to 'const char *' for 1st argument
   cpp_bin_float& operator=(const char *s);
                  ^
/usr/local/include/boost/multiprecision/cpp_bin_float.hpp:116:19: note: candidate template ignored: could not match
      'cpp_bin_float<Digits, DigitBase, type-parameter-0-2, type-parameter-0-3, MinExponent, MaxExponent>' against 'double'
   cpp_bin_float& operator=(const cpp_bin_float<D, B, A, E, MinE, MaxE> &o)
                  ^
/usr/local/include/boost/multiprecision/cpp_bin_float.hpp:129:74: note: candidate template ignored: substitution failure
      [with Float = double]: no type named 'type' in 'boost::enable_if_c<false, boost::multiprecision::backends::cpp_bin_float<24,
      boost::multiprecision::backends::digit_base_type::digit_base_2, void, short, -126, 127> &>'
      && (std::numeric_limits<Float>::radix == 2), cpp_bin_float&>::type operator=(const Float& f)
                                                                    ~~~~ ^
/usr/local/include/boost/multiprecision/cpp_bin_float.hpp:271:30: note: candidate template ignored: disabled by 'enable_if'
      [with I = double]
   typename boost::enable_if<is_integral<I>, cpp_bin_float&>::type operator=(const I& i)
                             ^
1 error generated.

[Jeff Hammond <jeff.science@…>]

source file that fails

[Jeff Hammond <jeff.science@…>]

log of failed build with Clang

[Jeff Hammond <jeff.science@…>]

log of failed build with Clang 3.6

[Jeff Hammond <jeff.science@…>]

log of failed build with Intel

[Jeff Hammond <jeff.science@…>]

log of failed build with GCC 4.2.1

[Jeff Hammond <jeff.science@…>]

log of failed build with GCC 5.2.0

[John Maddock]

Thanks for the report, looks like I got the conversions and what is and is not permitted slightly wrong. Fixed in ​https://github.com/boostorg/multiprecision/commit/7dfea115f8ddd78ef19a1d60bff04ecc93df7253