Add test note about NaN values

Author: mattico

src/float/add.rs

@@ -202,6 +202,12 @@ mod test {
     use qc::{U32, U64};
     use float::Float;
 
+    // NOTE The tests below have special handing for NaN values.
+    // Because NaN != NaN, the floating-point representations must be used
+    // Because there are many diffferent values of NaN, and the implementation
+    // doesn't care about calculating the 'correct' one, if both values are NaN
+    // the values are considered equivalent.
+
     // TODO: Add F32/F64 to qc so that they print the right values (at the very least)
     quickcheck! {
         fn addsf3(a: U32, b: U32) -> bool {
@@ -305,14 +311,12 @@ mod test {
     #[test]
     fn test_float_nan() {
         let r = super::__addsf3(f32::NAN, 1.23);
-        // Need .repr() since NAN != NAN
         assert_eq!(r.repr(), f32::NAN.repr());
     }
 
     #[test]
     fn test_double_nan() {
         let r = super::__adddf3(f64::NAN, 1.23);
-        // Need .repr() since NAN != NAN
         assert_eq!(r.repr(), f64::NAN.repr());
     }