Skip f64::mul_add tests on MinGW

Per [1], MinGW has an incorrect fma implementation. This showed up in
tests run with cranelift after adding float math operations to `core`.
Presumably we hadn't noticed this when running tests with LLVM because
LLVM was constant folding the result away.

[1]: https://sourceforge.net/p/mingw-w64/bugs/848/

Author: tgross35

library/core/src/num/f64.rs

@@ -1758,6 +1758,8 @@ pub fn fract(x: f64) -> f64 {
 /// ```
 /// #![feature(core_float_math)]
 ///
+/// # // FIXME(mingw): mingw has an incorrect fma https://sourceforge.net/p/mingw-w64/bugs/848/
+/// # #[cfg(all(target_os = "windows", target_env = "gnu", not(target_abi = "llvm")))] {
 /// use core::f64;
 ///
 /// let m = 10.0_f64;
@@ -1775,6 +1777,7 @@ pub fn fract(x: f64) -> f64 {
 /// assert_eq!(f64::mul_add(one_plus_eps, one_minus_eps, minus_one), -f64::EPSILON * f64::EPSILON);
 /// // Different rounding with the non-fused multiply and add.
 /// assert_eq!(one_plus_eps * one_minus_eps + minus_one, 0.0);
+/// # }
 /// ```
 ///
 /// _This standalone function is for testing only. It will be stabilized as an inherent method._

library/coretests/tests/floats/f64.rs

@@ -394,6 +394,8 @@ fn test_next_down() {
     assert_f64_biteq!(nan2.next_down(), nan2);
 }
 
+// FIXME(mingw): mingw has an incorrect fma https://sourceforge.net/p/mingw-w64/bugs/848/
+#[cfg_attr(all(target_os = "windows", target_env = "gnu", not(target_abi = "llvm")), ignore)]
 #[test]
 fn test_mul_add() {
     let nan: f64 = f64::NAN;