diff --git a/flang/include/flang/Evaluate/integer.h b/flang/include/flang/Evaluate/integer.h
index 31768c21daae6..7395645701265 100644
--- a/flang/include/flang/Evaluate/integer.h
+++ b/flang/include/flang/Evaluate/integer.h
@@ -27,6 +27,10 @@
 #include <string>
 #include <type_traits>
 
+// Some environments, viz. glibc 2.17, allow the macro HUGE
+// to leak out of <math.h>.
+#undef HUGE
+
 namespace Fortran::evaluate::value {
 
 // Implements an integer as an assembly of smaller host integer parts
diff --git a/flang/include/flang/Evaluate/real.h b/flang/include/flang/Evaluate/real.h
index 5266bd0ef64bf..d0da9634651f3 100644
--- a/flang/include/flang/Evaluate/real.h
+++ b/flang/include/flang/Evaluate/real.h
@@ -18,6 +18,10 @@
 #include <limits>
 #include <string>
 
+// Some environments, viz. glibc 2.17, allow the macro HUGE
+// to leak out of <math.h>.
+#undef HUGE
+
 namespace llvm {
 class raw_ostream;
 }
diff --git a/flang/lib/Evaluate/fold-implementation.h b/flang/lib/Evaluate/fold-implementation.h
index 6b3c9416724cb..9dd8c3843465d 100644
--- a/flang/lib/Evaluate/fold-implementation.h
+++ b/flang/lib/Evaluate/fold-implementation.h
@@ -39,6 +39,10 @@
 #include <type_traits>
 #include <variant>
 
+// Some environments, viz. glibc 2.17, allow the macro HUGE
+// to leak out of <math.h>.
+#undef HUGE
+
 namespace Fortran::evaluate {
 
 // Utilities