diff --git a/README.rst b/README.rst
index 5e0b4549..2c1e31ca 100644
--- a/README.rst
+++ b/README.rst
@@ -25,7 +25,8 @@ Features
   reported using exceptions.
 * Ease of use: small self-contained code base, no external dependencies,
   permissive BSD `license`_.
-* `Portability`_ and support for older compilers.
+* `Portability`_ with consistent output across platforms and support
+  for older compilers.
 
 Examples
 --------
diff --git a/format.cc b/format.cc
index 94b3aca2..bbc9b423 100644
--- a/format.cc
+++ b/format.cc
@@ -141,7 +141,7 @@ int signbit(double value) {
   if (value < 0) return 1;
   if (value == value) return 0;
   int dec = 0, sign = 0;
-  ecvt(value, 0, &dec, &sign);
+  _ecvt(value, 0, &dec, &sign);
   return sign;
 }
 #endif
@@ -337,6 +337,21 @@ void Formatter::FormatDouble(T value, const FormatSpec &spec, int precision) {
     return;
   }
 
+  if (isinf(value)) {
+    // Format infinity ourselves because sprintf's output is not consistent
+    // across platforms.
+    std::size_t size = 4;
+    const char *inf = upper ? " INF" : " inf";
+    if (!sign) {
+      --size;
+      ++inf;
+    }
+    char *out = FormatString(inf, size, spec);
+    if (sign)
+      *out = sign;
+    return;
+  }
+
   size_t offset = buffer_.size();
   unsigned width = spec.width;
   if (sign) {
diff --git a/format_test.cc b/format_test.cc
index cde7455e..0e3a734d 100644
--- a/format_test.cc
+++ b/format_test.cc
@@ -774,10 +774,6 @@ TEST(FormatterTest, FormatDouble) {
   sprintf(buffer, "%E", 392.65);
   EXPECT_EQ(buffer, str(Format("{0:E}") << 392.65));
   EXPECT_EQ("+0000392.6", str(Format("{0:+010.4g}") << 392.65));
-  double inf = std::numeric_limits<double>::infinity();
-  EXPECT_EQ("inf", str(Format("{}") << inf));
-  EXPECT_EQ("-inf", str(Format("{}") << -inf));
-  EXPECT_EQ("INF", str(Format("{:F}") << inf));
 }
 
 TEST(FormatterTest, FormatNaN) {
@@ -792,6 +788,18 @@ TEST(FormatterTest, FormatNaN) {
   EXPECT_EQ("    nan", str(Format("{:>7}") << nan));
 }
 
+TEST(FormatterTest, FormatInfinity) {
+  double inf = std::numeric_limits<double>::infinity();
+  EXPECT_EQ("inf", str(Format("{}") << inf));
+  EXPECT_EQ("+inf", str(Format("{:+}") << inf));
+  EXPECT_EQ("-inf", str(Format("{}") << -inf));
+  EXPECT_EQ(" inf", str(Format("{: }") << inf));
+  EXPECT_EQ("INF", str(Format("{:F}") << inf));
+  EXPECT_EQ("inf    ", str(Format("{:<7}") << inf));
+  EXPECT_EQ("  inf  ", str(Format("{:^7}") << inf));
+  EXPECT_EQ("    inf", str(Format("{:>7}") << inf));
+}
+
 TEST(FormatterTest, FormatLongDouble) {
   EXPECT_EQ("0", str(Format("{0:}") << 0.0l));
   EXPECT_EQ("0.000000", str(Format("{0:f}") << 0.0l));