Implicit conversion from float -100.21739130434781 to int loses precision