gh-123836: workaround fmod(x, y) bug on Windows

Modules/_math.h

@@ -23,3 +23,21 @@
 }
 
 #define m_log1p _Py_log1p
+
+static double
+_Py_fmod(double x, double y)
+{
+    /* Some platforms (e.g. Windows 10 with MSC v.1916) loose sign
+       for zero result.  But C99+ says: "if y is nonzero, the result
+       has the same sign as x".
+     */
+    double r = fmod(x, y);
+
+    if (r == 0.0 && y != 0.0) {
+        r = copysign(r, x);
+    }
+
+    return r;
+}
+
+#define m_fmod _Py_fmod

Modules/mathmodule.c

@@ -2347,7 +2347,7 @@ math_fmod_impl(PyObject *module, double x, double y)
     if (isinf(y) && isfinite(x))
         return PyFloat_FromDouble(x);
     errno = 0;
-    r = fmod(x, y);
+    r = m_fmod(x, y);
     if (isnan(r)) {
         if (!isnan(x) && !isnan(y))
             errno = EDOM;