Regression in multiprecision::cpp_int comparison

[chfast@…]

The comparison of 2 values of different cpp_int types in the following program produces incorrect result:

#include <boost/version.hpp>
#include <boost/multiprecision/cpp_int.hpp>
#include <iostream>

using bigint = boost::multiprecision::number<boost::multiprecision::cpp_int_backend<>>;
using u128 =  boost::multiprecision::number<boost::multiprecision::cpp_int_backend<128, 128, boost::multiprecision::unsigned_magnitude, boost::multiprecision::unchecked, void>>;

int main()
{
        u128 a = 14;
        bigint b = bigint("115792089237316195423570985008687907853269984665640564039457584007913129639948");
        auto correct = a < b;

        std::cout << "Boost " << BOOST_VERSION << " : " << correct << std::endl;
        return !correct;
}

Versions affected: 1.58+.

[Paweł Bylica <chfast@…>]

Test program

[John Maddock]

Confirmed, investigating.

[John Maddock]

Fixed in ​https://github.com/boostorg/multiprecision/commit/d711ce23df7f17e38dd218a91fa3dd905c91f881

[anonymous]

How this change is going to be released? In boost 1.59?

[John Maddock]

How this change is going to be released? In boost 1.59?

Yes, that's the plan.