Implicit conversion from float -5.338078291814952 to int loses precision