Same basic problem as bug 4433.

Further confusing behaviour. When BarkingFish "rejected" the page move, this caused the dummy edit that gets inserted on a page move to actually become sighted (The opposite of expected behaviour). ( https://secure.wikimedia.org/wikinews/en/w/index.php?title=Special:Log&page=38_killed_in_attack_on_Afghan_bank&hide_patrol_log=1&hide_review_log=0 )

My theory for what happened is:
*Diego moved the page, inserting a dummy edit into the page history as normal. Dummy edit is not sighted as normal
*BarkingFish came along and rejected that change.
*Rejecting (From my understanding anyways, could be mistaken) involves making an edit that restores the previous version, and sighting that edit. Since the previous version is exactly the same as the current version (since it's just a dummy edit), MediaWiki considered this new edit as a null edit, and didn't insert it into the db. However it still tried to flag this revert, but instead ended up flagging diego's original move.

End result, it appears as if BarkingFish sighted the very thing he tried to reject, which seems rather wrong.

(In reply to comment #1)

Same basic problem as bug 4433.

"Undo" (which reject is based off) has this nuisance too in addition to "rollback". It's hard to work with dummy revisions like this programmatically, even if we had a "null edit" flag.

(In reply to comment #2)

End result, it appears as if BarkingFish sighted the very thing he tried to
reject, which seems rather wrong.

Dealt with in r88660. Null rollback/undo autoreview was being fired due to the text of the current and attempted revision being the same.

r85743 is also related.

The confusing aspects of this should be fixed (not live though). The main aspect is the ability to reject non-text change aspects. Changing to "enhancement".