Implicit conversion from float -26.616541353383454 to int loses precision