diff --git a/include/fmt/format.h b/include/fmt/format.h
index ed90f077..9fd3fa44 100644
--- a/include/fmt/format.h
+++ b/include/fmt/format.h
@@ -278,10 +278,8 @@ struct dummy_int {
 };
 typedef std::numeric_limits<internal::dummy_int> fputil;
 
-// Dummy implementations of system functions such as signbit and ecvt called
-// if the latter are not available.
-inline dummy_int signbit(...) { return dummy_int(); }
-inline dummy_int _ecvt_s(...) { return dummy_int(); }
+// Dummy implementations of system functions called if the latter are not
+// available.
 inline dummy_int isinf(...) { return dummy_int(); }
 inline dummy_int _finite(...) { return dummy_int(); }
 inline dummy_int isnan(...) { return dummy_int(); }
@@ -321,7 +319,7 @@ namespace std {
 // Standard permits specialization of std::numeric_limits. This specialization
 // is used to resolve ambiguity between isinf and std::isinf in glibc:
 // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48891
-// and the same for isnan and signbit.
+// and the same for isnan.
 template <>
 class numeric_limits<fmt::internal::dummy_int> :
     public std::numeric_limits<int> {
@@ -345,19 +343,6 @@ class numeric_limits<fmt::internal::dummy_int> :
       return isnan(x) != 0;
     return _isnan(static_cast<double>(x)) != 0;
   }
-
-  // Portable version of signbit.
-  static bool isnegative(double x) {
-    using namespace fmt::internal;
-    if (const_check(sizeof(signbit(x)) != sizeof(fmt::internal::dummy_int)))
-      return signbit(x) != 0;
-    if (x < 0) return true;
-    if (!isnotanumber(x)) return false;
-    int dec = 0, sign = 0;
-    char buffer[2];  // The buffer size must be >= 2 or _ecvt_s will fail.
-    _ecvt_s(buffer, sizeof(buffer), x, 0, &dec, &sign);
-    return sign != 0;
-  }
 };
 }  // namespace std
 
@@ -2862,9 +2847,9 @@ void basic_writer<Range>::write_double(T value, const format_specs &spec) {
   internal::handle_float_type_spec(handler.type, handler);
 
   char sign = 0;
-  // Use isnegative instead of value < 0 because the latter is always
+  // Use signbit instead of value < 0 because the latter is always
   // false for NaN.
-  if (internal::fputil::isnegative(static_cast<double>(value))) {
+  if (std::signbit(value)) {
     sign = '-';
     value = -value;
   } else if (spec.flag(SIGN_FLAG)) {
diff --git a/test/format-impl-test.cc b/test/format-impl-test.cc
index 16694a4a..4679436c 100644
--- a/test/format-impl-test.cc
+++ b/test/format-impl-test.cc
@@ -129,7 +129,7 @@ TEST(FormatTest, ArgConverter) {
 
 TEST(FormatTest, FormatNegativeNaN) {
   double nan = std::numeric_limits<double>::quiet_NaN();
-  if (fmt::internal::fputil::isnegative(-nan))
+  if (std::signbit(-nan))
     EXPECT_EQ("-nan", fmt::format("{}", -nan));
   else
     fmt::print("Warning: compiler doesn't handle negative NaN correctly");
@@ -138,7 +138,8 @@ TEST(FormatTest, FormatNegativeNaN) {
 TEST(FormatTest, StrError) {
   char *message = nullptr;
   char buffer[BUFFER_SIZE];
-  EXPECT_ASSERT(fmt::safe_strerror(EDOM, message = nullptr, 0), "invalid buffer");
+  EXPECT_ASSERT(fmt::safe_strerror(EDOM, message = nullptr, 0),
+                "invalid buffer");
   EXPECT_ASSERT(fmt::safe_strerror(EDOM, message = buffer, 0),
                 "invalid buffer");
   buffer[0] = 'x';
@@ -151,7 +152,7 @@ TEST(FormatTest, StrError) {
 #endif
 
   int result = fmt::safe_strerror(error_code, message = buffer, BUFFER_SIZE);
-  EXPECT_EQ(0, result);
+  EXPECT_EQ(result, 0);
   std::size_t message_size = std::strlen(message);
   EXPECT_GE(BUFFER_SIZE - 1u, message_size);
   EXPECT_EQ(get_system_error(error_code), message);