Last modified: 2013-01-14 16:00:53 UTC
The poll for changes stuff and saving recent change entries could use more checks to make sure duplicate entries do not get saved. This could be: 1) at the recent change save point 2) in poll for changes, programatically know which change id to start at, maybe with a hook point before pollforchanges queries the changes table. or both. This is in event the poll for changes script has to be restarted. Right now, it can use --since and work at a specified time interval with a cron job or run continuously as a daemon.
https://gerrit.wikimedia.org/r/#/c/37227/ - we have check now at the save point. https://gerrit.wikimedia.org/r/#/c/37819/ - pollforchanges tracks the last change ID handled in a state file