Last modified: 2011-02-25 10:33:03 UTC
Automatic flagging after rollback, as implemented by [[bugzilla:14074]], appears to be broken now. At least, history http://ru.wikipedia.org/w/index.php?title=Nokia_5230&offset=20110218111400&limit=3&curid=2041269&action=history&uselang=en and diffs show that the new good revision is not flagged.
The review log http://ru.wikipedia.org/w/index.php?title=Special:Log&type=review&user=&page=Nokia+5230&uselang=en shows that it was the old good revision reviewed in the time of rolling back, nothing about the recent one.
Although 14074 marked as WONTFIX in this database, automatic rollback flagging actually worked in the past (2009–2010, say, before MW 1.17). This history http://ru.wikipedia.org/w/index.php?title=Trine&offset=20101229999999&limit=4&action=history&uselang=en clearly demonstrates the recent revision was "automatically checked".
Any other cases? I can't repro this yet on my test wiki.
Was this actually software rollback (via [rollback]) or some JS script based rollback?
We can not perform other cases, not by Yakiv_Gluck nor by other users.
I think that it was the 1.17 deployment what resulted in [[w:ru:Nokia_5230]] anomaly. An old good revision was stored before it, and vandalism and rollbacking occurred after it. So, this is probably a deployment bug, not a FlaggedRevs bug.
Ordinary action=rollback was used: http://ru.wikipedia.org/w/index.php?title=…&action=rollback&from=…&token=…%2B%5C
I'll let this sit a few days before closing. I can't see why this would happen.
(In reply to comment #4)
> We can not perform other cases, not by Yakiv_Gluck nor by other users.
> I think that it was the 1.17 deployment what resulted in [[w:ru:Nokia_5230]]
Assuming this as a deployment anomaly. I can't see any likely race problems in maybeMakeEditReviewed() or autoReviewEdit() either.
Closing WMF (though that isn't the most exact status since there is an example).