Issue #4916: fix little-endian UTF-16 decoding bug on big-endian UCS-4 builds, introduced by r68483.
diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c
index bc1612d..c668d0c 100644
--- a/Objects/unicodeobject.c
+++ b/Objects/unicodeobject.c
@@ -2791,16 +2791,24 @@
_p[3] = 0;
#endif
#endif
- ((unsigned char *) _p)[1] = _q[0];
- ((unsigned char *) _p)[0] = _q[1];
- ((unsigned char *) _p)[1 + Py_UNICODE_SIZE] = _q[2];
- ((unsigned char *) _p)[0 + Py_UNICODE_SIZE] = _q[3];
-#if (SIZEOF_LONG == 8)
- ((unsigned char *) _p)[1 + 2 * Py_UNICODE_SIZE] = _q[4];
- ((unsigned char *) _p)[0 + 2 * Py_UNICODE_SIZE] = _q[5];
- ((unsigned char *) _p)[1 + 3 * Py_UNICODE_SIZE] = _q[6];
- ((unsigned char *) _p)[0 + 3 * Py_UNICODE_SIZE] = _q[7];
+ /* Issue #4916; UCS-4 builds on big endian machines must
+ fill the two last bytes of each 4-byte unit. */
+#if (!defined(BYTEORDER_IS_LITTLE_ENDIAN) && Py_UNICODE_SIZE > 2)
+# define OFF 2
+#else
+# define OFF 0
#endif
+ ((unsigned char *) _p)[OFF + 1] = _q[0];
+ ((unsigned char *) _p)[OFF + 0] = _q[1];
+ ((unsigned char *) _p)[OFF + 1 + Py_UNICODE_SIZE] = _q[2];
+ ((unsigned char *) _p)[OFF + 0 + Py_UNICODE_SIZE] = _q[3];
+#if (SIZEOF_LONG == 8)
+ ((unsigned char *) _p)[OFF + 1 + 2 * Py_UNICODE_SIZE] = _q[4];
+ ((unsigned char *) _p)[OFF + 0 + 2 * Py_UNICODE_SIZE] = _q[5];
+ ((unsigned char *) _p)[OFF + 1 + 3 * Py_UNICODE_SIZE] = _q[6];
+ ((unsigned char *) _p)[OFF + 0 + 3 * Py_UNICODE_SIZE] = _q[7];
+#endif
+#undef OFF
_q += SIZEOF_LONG;
_p += SIZEOF_LONG / 2;
}