Avoid crashing because of an unaligned word access
diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c
index 772707d..a4d210b 100644
--- a/Objects/unicodeobject.c
+++ b/Objects/unicodeobject.c
@@ -6252,7 +6252,15 @@
end = s + size;
while (s < end) {
- Py_UCS4 ch = *(Py_UNICODE*)s;
+ Py_UCS4 ch;
+ /* We copy the raw representation one byte at a time because the
+ pointer may be unaligned (see test_codeccallbacks). */
+ ((char *) &ch)[0] = s[0];
+ ((char *) &ch)[1] = s[1];
+#ifdef Py_UNICODE_WIDE
+ ((char *) &ch)[2] = s[2];
+ ((char *) &ch)[3] = s[3];
+#endif
/* We have to sanity check the raw data, otherwise doom looms for
some malformed UCS-4 data. */
if (