Fix SF bug #1402308, segfault when using mmap(-1, ...)

This didn't crash on Linux, but valgrind complained.
I'm not sure if this test is valid on Windows.

Will backport.

Lib/test/output/test_mmap

@@ -31,6 +31,7 @@ test_mmap
   Modifying copy-on-write memory map.
   Ensuring copy-on-write maps cannot be resized.
   Ensuring invalid access parameter raises exception.
+  Try opening a bad file descriptor...
   Ensuring that passing 0 as map length sets map size to current file size.
   Ensuring that passing 0 as map length sets map size to current file size.
  Test passed

Lib/test/test_mmap.py

@@ -281,6 +281,14 @@ def test_both():
         except OSError:
             pass
 
+    print '  Try opening a bad file descriptor...'
+    try:
+        mmap.mmap(-1, 4096)
+    except mmap.error:
+        pass
+    else:
+        verify(0, 'expected a mmap.error but did not get it')
+
     # Do a tougher .find() test.  SF bug 515943 pointed out that, in 2.2,
     # searching for data with embedded \0 bytes didn't work.
     f = open(TESTFN, 'w+')

Misc/ACKS

@@ -529,6 +529,7 @@ Michael Scharf
 Neil Schemenauer
 David Scherer
 Gregor Schmid
+Ralf Schmitt
 Peter Schneider-Kamp
 Sam Schulenburg
 Stefan Schwarzer

Misc/NEWS

@@ -216,6 +216,8 @@ Core and builtins
 Extension Modules
 -----------------
 
+- Bug #1402308, (possible) segfault when using mmap.mmap(-1, ...)
+
 - Bug #1400822, _curses over{lay,write} doesn't work when passing 6 ints.
   Also fix ungetmouse() which did not accept arguments properly.
   The code now conforms to the documented signature.

Modules/mmapmodule.c

@@ -918,6 +918,7 @@ new_mmap_object(PyObject *self, PyObject *args, PyObject *kwdict)
 #endif
 	m_obj = PyObject_New (mmap_object, &mmap_object_type);
 	if (m_obj == NULL) {return NULL;}
+	m_obj->data = NULL;
 	m_obj->size = (size_t) map_size;
 	m_obj->pos = (size_t) 0;
 	m_obj->fd = dup(fd);