don't try to create non-standard denormalization signal

Underflow exception is only raised when the result is
invalid, but fmod is always exact. x87 has a denormalization
exception, but that's nonstandard. And the superflous *1.0
will be optimized away by any compiler that does not honor
signaling nans.

src/math/fmod.c

@@ -17,7 +17,7 @@
 
 #include "libm.h"
 
-static const double one = 1.0, Zero[] = {0.0, -0.0,};
+static const double Zero[] = {0.0, -0.0,};
 
 double fmod(double x, double y)
 {
@@ -140,7 +140,6 @@ double fmod(double x, double y)
 			lx = hx>>(n-32); hx = sx;
 		}
 		INSERT_WORDS(x, hx|sx, lx);
-		x *= one;  /* create necessary signal */
 	}
 	return x;  /* exact output */
 }

src/math/fmodf.c

@@ -20,7 +20,7 @@
 
 #include "libm.h"
 
-static const float one = 1.0, Zero[] = {0.0, -0.0,};
+static const float Zero[] = {0.0, -0.0,};
 
 float fmodf(float x, float y)
 {
@@ -99,7 +99,6 @@ float fmodf(float x, float y)
 		n = -126 - iy;
 		hx >>= n;
 		SET_FLOAT_WORD(x, hx|sx);
-		x *= one;          /* create necessary signal */
 	}
 	return x;  /* exact output */
 }

src/math/fmodl.c

@@ -48,7 +48,7 @@ typedef uint32_t manh_t;
 
 #define MANL_SHIFT      (LDBL_MANL_SIZE - 1)
 
-static const long double one = 1.0, Zero[] = {0.0, -0.0,};
+static const long double Zero[] = {0.0, -0.0,};
 
 /*
  * fmodl(x,y)
@@ -153,7 +153,6 @@ long double fmodl(long double x, long double y)
 	} else {
 		ux.bits.exp = iy + BIAS;
 	}
-	x = ux.e * one;   /* create necessary signal */
+	return ux.e;       /* exact output */
-	return x;         /* exact output */
 }
 #endif