Tom Lane wrote:
> inoue@postgresql.org writes:
> > prevent open failure of VIEW
>
> I do not like this fix. We should be moving away from having smgr
> access relcache entries, not adding more relcache knowledge to it.
> So checks for view-ness should be at a higher level.
>
> What case were you seeing a failure in?
>
I see the failure in xlogutils.c.
Oops I'm mistaken,another is smgrnblocks call
in RelationGetNumberOfBlocks() and I didn't
change it at all.
I don't like this fix either and I intend to change
at higher level first. However it seems to me
that checking relkind every time is not only ugly
but also unreliable.
I don't stick to my change.
Please change as you prefer.
Regards.
Hiroshi Inoue