[libc] Fix compilation when long double is used as float128 (#97747)

Small patch that fixes the compilation when float128 is a long double, e.g., rv32. It fixes a static_cast, adds a missing %= operator and changes a cast to use get_val() in a test case instead

Author: mikhailramalho

libc/src/__support/FPUtil/generic/FMod.h

@@ -160,7 +160,8 @@ template <typename T> struct FModDivisionInvMultHelper {
 template <typename T, typename U = typename FPBits<T>::StorageType,
           typename DivisionHelper = FModDivisionSimpleHelper<U>>
 class FMod {
-  static_assert(cpp::is_floating_point_v<T> && cpp::is_unsigned_v<U> &&
+  static_assert(cpp::is_floating_point_v<T> &&
+                    is_unsigned_integral_or_big_int_v<U> &&
                     (sizeof(U) * CHAR_BIT > FPBits<T>::FRACTION_LEN),
                 "FMod instantiated with invalid type.");
 

libc/src/__support/big_int.h

@@ -731,6 +731,10 @@ struct BigInt {
     return *result.div(other);
   }
 
+  LIBC_INLINE constexpr BigInt operator%=(const BigInt &other) {
+    return *this->div(other);
+  }
+
   LIBC_INLINE constexpr BigInt &operator*=(const BigInt &other) {
     *this = *this * other;
     return *this;

libc/test/src/math/smoke/CopySignTest.h

@@ -42,9 +42,9 @@ class CopySignTest : public LIBC_NAMESPACE::testing::FEnvSafeTest {
     StorageType v = 0;
     for (int i = 0; i <= COUNT; ++i, v += STEP) {
       FPBits x_bits = FPBits(v);
-      T x = T(v);
       if (x_bits.is_nan() || x_bits.is_inf())
         continue;
+      T x = x_bits.get_val();
 
       T res1 = func(x, -x);
       ASSERT_FP_EQ(res1, -x);