On 30/10/13 12:04, Alan DeKok wrote: > The bug in rlm_sql was that SELECT returning no values was treated as > an empty string, not as a failed query. So in an alternation, the empty > string is a valid string, and the second part wasn't used. Ah, I get it.