Use Py_ssize_t for _Py_RefTotal.

I tried to handle Win64 properly, but please review.

Include/object.h

@@ -568,7 +568,7 @@ environment the global variable trick is not safe.)
  * #ifdefs (we used to do that -- it was impenetrable).
  */
 #ifdef Py_REF_DEBUG
-PyAPI_DATA(long) _Py_RefTotal;
+PyAPI_DATA(Py_ssize_t) _Py_RefTotal;
 PyAPI_FUNC(void) _Py_NegativeRefcount(const char *fname,
 					    int lineno, PyObject *op);
 #define _Py_INC_REFTOTAL	_Py_RefTotal++

Python/pythonrun.c

@@ -29,6 +29,16 @@
 #include "windows.h"
 #endif
 
+#ifndef Py_REF_DEBUG
+#  define PRINT_TOTAL_REFS()
+#else /* Py_REF_DEBUG */
+#  if defined(MS_WIN64)
+#    define PRINT_TOTAL_REFS() fprintf(stderr, "[%zd refs]\n", _Py_RefTotal);
+#  else /* ! MS_WIN64 */
+#    define PRINT_TOTAL_REFS() fprintf(stderr, "[%ld refs]\n", _Py_RefTotal);
+#  endif /* MS_WIN64 */
+#endif
+
 extern char *Py_GetPath(void);
 
 extern grammar _PyParser_Grammar; /* From graminit.c */
@@ -382,9 +392,7 @@ Py_Finalize(void)
 	dump_counts();
 #endif
 
-#ifdef Py_REF_DEBUG
-	fprintf(stderr, "[%ld refs]\n", _Py_RefTotal);
-#endif
+	PRINT_TOTAL_REFS()
 
 #ifdef Py_TRACE_REFS
 	/* Display all objects still alive -- this can invoke arbitrary
@@ -674,9 +682,7 @@ PyRun_InteractiveLoopFlags(FILE *fp, const char *filename, PyCompilerFlags *flag
 	}
 	for (;;) {
 		ret = PyRun_InteractiveOneFlags(fp, filename, flags);
-#ifdef Py_REF_DEBUG
-		fprintf(stderr, "[%ld refs]\n", _Py_RefTotal);
-#endif
+		PRINT_TOTAL_REFS()
 		if (ret == E_EOF)
 			return 0;
 		/*

Python/sysmodule.c

@@ -604,7 +604,7 @@ sys_getrefcount(PyObject *self, PyObject *arg)
 static PyObject *
 sys_gettotalrefcount(PyObject *self)
 {
-	return PyInt_FromLong(_Py_RefTotal);
+	return PyInt_FromSsize_t(_Py_RefTotal);
 }
 
 #endif /* Py_TRACE_REFS */