whatsnew: PyStructSequence_InitType2, Victor's MemoryError fixes.

I'm documenting this new function, but I'm wondering if this will trigger
someone to think about whether or not this actually makes sense :)

Doc/c-api/tuple.rst

@@ -129,6 +129,14 @@ type.
    Initializes a struct sequence type *type* from *desc* in place.
 
 
+.. c:function:: void PyStructSequence_InitType2(PyTypeObject *type, PyStructSequence_Desc *desc)
+
+   The same as ``PyStructSequence_InitType``, but returns ``0`` on success and ``-1`` on
+   failure.
+
+   .. versionadded:: 3.4
+
+
 .. c:type:: PyStructSequence_Desc
 
    Contains the meta information of a struct sequence type to create.

Doc/whatsnew/3.4.rst

@@ -1421,6 +1421,10 @@ Other Build and C API Changes
   specifications for ``%s``, ``%A``, ``%U``, ``%V``, ``%S``, and ``%R``.
   (Contributed by Ysj Ray and Victor Stinner in :issue:`7330`.)
 
+* New function :c:func:`PyStructSequence_InitType2` supplements the
+  existing :c:func:`PyStructSequence_InitType` function.  The difference
+  is that it returns ``0`` on success and ``-1`` on failure.
+
 
 .. _other-improvements-3.4:
 
@@ -1466,6 +1470,10 @@ Other Improvements
   uses of :ref:`__prepare__ <prepare>`.  (Contributed by Benjamin Peterson in
   :issue:`17853`.)
 
+* A number of MemoryError-related crashes were identified and fixed by Victor
+  Stinner using his :pep:`445`-based ``pyfailmalloc`` tool (:issue:`18408`,
+  :issue:`18520`).
+
 
 Significant Optimizations
 -------------------------