Implicit conversion from float -12.333931777378808 to int loses precision