It turns out that PyType_GenericNew, does not do a Py_INCREF on the
passed in type, so if we Py_DECREF the reference to the imported dm module
in dmraid.c theoretically the dm module (holding the dm.table type) could
be destroyed from underneath us. We fix this by not Py_DECREF'ing
the imported module after we are done with it.

Note this does mean we leak a reference and the module will never get
destroyed, but we can live with that. Actually we have been doing that
until very recently, when a patch of mine introduced these DECREF's to
being with.
---
dmraid.c | 4 ----
1 files changed, 0 insertions(+), 4 deletions(-)

diff --git a/dmraid.c b/dmraid.c
index 23e1bbc..b397adc 100644
--- a/dmraid.c
+++ b/dmraid.c
@@ -824,15 +824,12 @@ pydmraid_raidset_get_dm_table(PyObject *self, void *data)
Py_CLEAR(table);
} while ((s = strtok(NULL, "
")));

- Py_DECREF(m);
-
return table_list;

out:
Py_XDECREF(iret);
Py_XDECREF(table);
Py_XDECREF(table_list);
- Py_XDECREF(m);
if (!PyErr_Occurred()) {
if (errno != 0)
PyErr_SetFromErrno(PyExc_SystemError);
@@ -883,7 +880,6 @@ pydmraid_raidset_get_map(PyObject *self, void *data)

out:
Py_XDECREF(iret);
- Py_XDECREF(m);
Py_XDECREF(table);
if (!map && !PyErr_Occurred()) {
if (errno != 0)
--
1.6.1.3

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list