Section 7.4 requires (for single-precision floating point numbers) an accuracy of 2.5 ulp for reciprocal and division. However, the fmod, modf, remainder, remquo functions require 0 ulp. This seems...