test: Skip nexttoward_1_snan on clang soft ld targets

When the target has no hardware support for long double, but does
have hardware for float or double, compiler.rt is inconsistent
about reporting FE_INVALID.

Signed-off-by: Keith Packard <[email protected]>

Author: keith-packard

test/math_errhandling_tests.c

@@ -489,7 +489,10 @@ static FLOAT_T makemathname(test_nexttoward_negmin_neg1)(void) { return makemath
 static FLOAT_T makemathname(test_nexttoward_qnan_1)(void) { return makemathname(nexttoward)(makemathname(qnanval), makelname(one)); }
 static FLOAT_T makemathname(test_nexttoward_snan_1)(void) { return makemathname(nexttoward)(makemathname(snanval), makelname(one)); }
 static FLOAT_T makemathname(test_nexttoward_1_qnan)(void) { return makemathname(nexttoward)(makemathname(one), makelname(qnanval)); }
+#if !defined(__clang__) || !defined(PICOLIBC_LONG_DOUBLE_NOEXCEPT)
+/* compiler.rt is inconsistent when dealing with long double snan on hardware with only non-ld floats */
 static FLOAT_T makemathname(test_nexttoward_1_snan)(void) { return makemathname(nexttoward)(makemathname(one), makelname(snanval)); }
+#endif
 static FLOAT_T makemathname(test_nexttoward_max_inf)(void) { return makemathname(nexttoward)(makemathname(max_val), makelname(infval)); }
 static FLOAT_T makemathname(test_nexttoward_negmax_neginf)(void) { return makemathname(nexttoward)(-makemathname(max_val), -makelname(infval)); }
 static FLOAT_T makemathname(test_nexttoward_min_0)(void) { return makemathname(nexttoward)(makemathname(min_val), makelname(zero)); }
@@ -1287,7 +1290,9 @@ const struct {
         TEST(nexttoward_qnan_1, (FLOAT_T)NAN, 0, 0),
         TEST(nexttoward_snan_1, (FLOAT_T)NAN, FE_INVALID, 0),
         TEST(nexttoward_1_qnan, (FLOAT_T)NAN, 0, 0),
+#if !defined(__clang__) || !defined(PICOLIBC_LONG_DOUBLE_NOEXCEPT)
         TEST(nexttoward_1_snan, (FLOAT_T)NAN, FE_INVALID, 0),
+#endif
         TEST(nexttoward_max_inf, (FLOAT_T)INFINITY, FE_OVERFLOW, ERANGE),
         TEST(nexttoward_negmax_neginf, -(FLOAT_T)INFINITY, FE_OVERFLOW, ERANGE),
         TEST(nexttoward_min_0, (FLOAT_T)0, FE_UNDERFLOW, ERANGE),