Implicit conversion from float -91.2111801242236 to int loses precision