math: excess precision fix modf, modff, scalbn, scalbnf

old code was correct only if the result was stored (without the
excess precision) or musl was compiled with -ffloat-store.
now we use STRICT_ASSIGN to work around the issue.
(see note 160 in c11 section 6.8.6.4)

1 files changed, 6 insertions, 8 deletions

diff --git a/src/math/scalbn.c b/src/math/scalbn.c
index 42a2b294..003141e3 100644
--- a/src/math/scalbn.c
+++ b/src/math/scalbn.c
@@ -2,8 +2,6 @@
 
 double scalbn(double x, int n)
 {
-	/* make sure result is stored as double on overflow or underflow */
-	volatile double z;
 	double scale;
 
 	if (n > 1023) {
@@ -13,8 +11,8 @@ double scalbn(double x, int n)
 			x *= 0x1p1023;
 			n -= 1023;
 			if (n > 1023) {
-				z = x * 0x1p1023;
-				return z;
+				STRICT_ASSIGN(double, x, x * 0x1p1023);
+				return x;
 			}
 		}
 	} else if (n < -1022) {
@@ -24,12 +22,12 @@ double scalbn(double x, int n)
 			x *= 0x1p-1022;
 			n += 1022;
 			if (n < -1022) {
-				z = x * 0x1p-1022;
-				return z;
+				STRICT_ASSIGN(double, x, x * 0x1p-1022);
+				return x;
 			}
 		}
 	}
 	INSERT_WORDS(scale, (uint32_t)(0x3ff+n)<<20, 0);
-	z = x * scale;
-	return z;
+	STRICT_ASSIGN(double, x, x * scale);
+	return x;
 }