Issue #18513: Add workaround for OS X 10.8 cexp bug that leads to wrong cmath.rect(0.0,-0.0) results.

Misc/NEWS

@@ -28,6 +28,9 @@ Library
 
 - Issue #18455: multiprocessing should not retry connect() with same socket.
 
+- Issue #18513: Fix behaviour of cmath.rect w.r.t. signed zeros on OS X 10.8 +
+  gcc.
+
 - Issue #18101: Tcl.split() now process Unicode strings nested in a tuple as it
   do with byte strings.
 

Modules/cmathmodule.c

@@ -1006,6 +1006,13 @@ cmath_rect(PyObject *self, PyObject *args)
         else
             errno = 0;
     }
+    else if (phi == 0.0) {
+        /* Workaround for buggy results with phi=-0.0 on OS X 10.8.  See
+           bugs.python.org/issue18513. */
+        z.real = r;
+        z.imag = r * phi;
+        errno = 0;
+    }
     else {
         z.real = r * cos(phi);
         z.imag = r * sin(phi);