Wikimedia Bugzilla is closed!

Looking at queries run on the Toolserver's replica of the en.wiki database:

mysql> SELECT * FROM categorylinks WHERE cl_to = "Africa-related_stubs";
+---------+----------------------+----------------+---------------------+
| cl_from | cl_to                | cl_sortkey     | cl_timestamp        |
+---------+----------------------+----------------+---------------------+
|  891746 | Africa-related_stubs | Hamani Diori   | 2005-05-31 04:11:48 | 
| 1547553 | Africa-related_stubs | Patty Schnyder | 2005-02-26 17:30:45 | 
+---------+----------------------+----------------+---------------------+
2 rows in set (0.04 sec)

mysql> SELECT * FROM page WHERE page_id = 891746;
Empty set (1.22 sec)


It's clear that there are old rows in the categorylinks table that for some reason were not purged when the page was deleted. The rows reference deleted pages. This causes inaccuracies when things like COUNT(cl_to) are run in database queries.

A maintenance script should be written and run on the categorylinks tables of WMF wikis to purge these old rows.