Implicit conversion from float -61.234567901234556 to int loses precision