RE: Index tuple count != heap tuple count problem identified]

> -----Original Message-----
> From: Bruce Momjian [mailto:pgman(at)candle(dot)pha(dot)pa(dot)us]
>
> > > -----Original Message-----
> > > From: pgsql-hackers-owner(at)hub(dot)org
> [mailto:pgsql-hackers-owner(at)hub(dot)org]On
> > > Behalf Of Tom Lane
> > >
> > > You'll probably recall reports of messages like this out of VACUUM:
> > > NOTICE: Index ind1: NUMBER OF INDEX' TUPLES (2002) IS NOT THE
> > > SAME AS HEAP' (3003).
> > > I've figured out the cause (or at least a cause) of this condition.
> > >
> > > The CREATE INDEX operation has only bothered to index the non-dead
> > > tuples. So, VACUUM's little sanity check fails.
> >
> >
> > Is it wrong to change the implementation of CREATE INDEX ?
> > I have a fix.
> > It needs the change of duplicate check(tuplesort->btbuild) and
> > I've thougth that it would be better to change it after the release
> > of 7.0.
>
> Well, it seems we better do something about it before 7.0 is released.
> Now it seems we have to decide to change CREATE INDEX, or modify VACUUM.
>

It's difficult for me to provide a fix for CREATE INDEX before 7.0 is
released.
It's not sufficiently checked and I don't remember details now.
I'm a little busy now and don't have enough time to look at it again.

Regards.

Hiroshi Inoue
Inoue(at)tpf(dot)co(dot)jp