diff --git a/mfbt/FloatingPoint.h b/mfbt/FloatingPoint.h
index d0177b31e91..b80ce266f1a 100644
--- a/mfbt/FloatingPoint.h
+++ b/mfbt/FloatingPoint.h
@@ -139,7 +139,12 @@ IsNegativeZero(double d)
   return bits == DoubleSignBit;
 }
 
-/** Returns the exponent portion of the double. */
+/**
+ * Returns the exponent portion of the double.
+ *
+ * Zero is not special-cased, so ExponentComponent(0.0) is
+ * -int_fast16_t(DoubleExponentBias).
+ */
 static MOZ_ALWAYS_INLINE int_fast16_t
 ExponentComponent(double d)
 {
diff --git a/mfbt/tests/TestFloatingPoint.cpp b/mfbt/tests/TestFloatingPoint.cpp
index 7ae9961fff6..844313f312a 100644
--- a/mfbt/tests/TestFloatingPoint.cpp
+++ b/mfbt/tests/TestFloatingPoint.cpp
@@ -8,6 +8,8 @@
 #include <math.h>
 
 using mozilla::DoublesAreIdentical;
+using mozilla::DoubleExponentBias;
+using mozilla::ExponentComponent;
 using mozilla::IsFinite;
 using mozilla::IsInfinite;
 using mozilla::IsNaN;
@@ -102,6 +104,22 @@ TestDoublesAreIdentical()
   ShouldNotBeIdentical(UnspecifiedNaN(), NegativeInfinity());
 }
 
+static void
+TestExponentComponent()
+{
+  MOZ_ASSERT(ExponentComponent(0.0) == -int_fast16_t(DoubleExponentBias));
+  MOZ_ASSERT(ExponentComponent(-0.0) == -int_fast16_t(DoubleExponentBias));
+  MOZ_ASSERT(ExponentComponent(0.125) == -3);
+  MOZ_ASSERT(ExponentComponent(0.5) == -1);
+  MOZ_ASSERT(ExponentComponent(1.0) == 0);
+  MOZ_ASSERT(ExponentComponent(1.5) == 0);
+  MOZ_ASSERT(ExponentComponent(2.0) == 1);
+  MOZ_ASSERT(ExponentComponent(7) == 2);
+  MOZ_ASSERT(ExponentComponent(PositiveInfinity()) == DoubleExponentBias + 1);
+  MOZ_ASSERT(ExponentComponent(NegativeInfinity()) == DoubleExponentBias + 1);
+  MOZ_ASSERT(ExponentComponent(UnspecifiedNaN()) == DoubleExponentBias + 1);
+}
+
 static void
 TestPredicates()
 {
@@ -152,5 +170,6 @@ int
 main()
 {
   TestDoublesAreIdentical();
+  TestExponentComponent();
   TestPredicates();
 }