> I found the cause of this one --- the new code for estimating the
> selectivity of @@ was not guarding against the possibility that the
> given TSQuery is empty, as it is in this example.
Shouldn't be this in regression test?
> I'm still unable to reproduce the "tuple offset out of range" errors.
Maybe this is due to the difference how the index is physically
organized. I will uplaod physical database cluster.
--
Tatsuo Ishii
SRA OSS, Inc. Japan