The PowerLinux buildslave fails in test_dbm:test_keys() because of a problem with the "in" operator.

>>> import dbm
>>> d = dbm.open('t','c')
>>> a = [('a', 'b'), ('12345678910', '019237410982340912840198242')]
>>> for k,v in a:
...     d[k] = v
... 
>>> print d
<dbm.dbm object at 0x3fff93073110>
>>> print d.keys()
['a', '12345678910']
>>> print 'a' in d
False      <--- This apparently should be True
>>> print 'a' in d.keys()
True

Can you try the other dbm methods? e.g. has_key(), get()...

Forget it, can you just try the following patch?

diff --git a/Modules/dbmmodule.c b/Modules/dbmmodule.c
--- a/Modules/dbmmodule.c
+++ b/Modules/dbmmodule.c
@@ -168,11 +168,13 @@
 dbm_contains(register dbmobject *dp, PyObject *v)
 {
     datum key, val;
+    char *ptr;
+    Py_ssize_t size;
 
-    if (PyString_AsStringAndSize(v, (char **)&key.dptr,
-                                 (Py_ssize_t *)&key.dsize)) {
+    if (PyString_AsStringAndSize(v, &ptr, &size))
         return -1;
-    }
+    key.dptr = ptr;
+    key.dsize = size;
 
     /* Expand check_dbmobject_open to return -1 */
     if (dp->di_dbm == NULL) {

My example and test_dbm succeeds on Python2.7 with your patch applied.
Thanks!

Ok. This is a classic example of why a big-endian buildbot is useful :)

New changeset 53da3bad8554 by Antoine Pitrou in branch '2.7':
Issue #17926: Fix dbm.__contains__ on 64-bit big-endian machines.
http://hg.python.org/cpython/rev/53da3bad8554

Committed, thank you.

See also issue9687.