Implicit conversion from float -23.850415512465375 to int loses precision