diff --git a/safemult.h b/safemult.h
index e2a83ca..c63ee62 100644
--- a/safemult.h
+++ b/safemult.h
@@ -15,14 +15,14 @@ extern "C" {
 /* for historical reasons, the mult interface uses 0 for failure and 1
  * for success, while the builtins (and my own addition and
  * subtraction routines in rangecheck.h) do it the other way around. */
-extern inline int umult16(uint16 a,uint16 b,uint16* c) { return !__builtin_mul_overflow(a,b,c); }
-extern inline int imult16( int16 a, int16 b, int16* c) { return !__builtin_mul_overflow(a,b,c); }
+static inline int umult16(uint16 a,uint16 b,uint16* c) { return !__builtin_mul_overflow(a,b,c); }
+static inline int imult16( int16 a, int16 b, int16* c) { return !__builtin_mul_overflow(a,b,c); }
 
-extern inline int umult32(uint32 a,uint32 b,uint32* c) { return !__builtin_mul_overflow(a,b,c); }
-extern inline int imult32( int32 a, int32 b, int32* c) { return !__builtin_mul_overflow(a,b,c); }
+static inline int umult32(uint32 a,uint32 b,uint32* c) { return !__builtin_mul_overflow(a,b,c); }
+static inline int imult32( int32 a, int32 b, int32* c) { return !__builtin_mul_overflow(a,b,c); }
 
-extern inline int umult64(uint64 a,uint64 b,uint64* c) { return !__builtin_mul_overflow(a,b,c); }
-extern inline int imult64( int64 a, int64 b, int64* c) { return !__builtin_mul_overflow(a,b,c); }
+static inline int umult64(uint64 a,uint64 b,uint64* c) { return !__builtin_mul_overflow(a,b,c); }
+static inline int imult64( int64 a, int64 b, int64* c) { return !__builtin_mul_overflow(a,b,c); }
 
 #else