boost multiprecision conversion from binary float to int types fails for large floats

[marc.alexa@…]

It appears the conversion from binary float types such as

typedef number<cpp_bin_float<8,backends::digit_base_2> > quarter_float;

to integer types fails if the float is 2⁸ or larger. For example, using

quarter_float qf(256);
cpp_int i = qf.convert_to<cpp_int>();
unsigned int n = qf.convert_to<unsigned int>();

I get 0 for i and 4294967295 for n.

A try to attach a full code example.

[John Maddock]

Confirmed, investigating, thanks.

[John Maddock]

Fixed in ​https://github.com/boostorg/multiprecision/commit/a4b198e1b79e735825e15e6bf6572b86e7b8dd3c

Thanks for the report.