Implicit conversion from float -134.7058823529412 to int loses precision