Tom Lane wrote:
>
> Hiroshi Inoue <Inoue@tpf.co.jp> writes:
> > REINDEX uses the relfilenode mechanism since 7.1 which
> > lets the replacement of index files be under transactional
> > control. I think it's safe enough.
>
> Okay, in that case tcop/utility is being too picky about all three
> cases, no?
Probably we don't have to keep the relhasindex info in the
db any longer and we had better keep some info about REINDEX
in memory local to the backend. In fact in the current
implementation the relhasindex is local to the backend
and any backend couldn't see the column committed to
the status off.
> Hmm ... what that says is that unlinking pg_internal.init in
> setRelfilenode is the wrong place.
Possibly. I couldn't find the appropriate place(way) then
and so #ifdef's are still there.
regards,
Hiroshi Inoue