Re: [BUGS] BUG #11867: Strange behaviour with composite types after resetting database tablespace

On Mon, 2014-11-03 at 22:08 -0500, Tom Lane wrote:
> <spock>Fascinating.</spock>

:-)

> I believe what is happening is:

[ . . . ]

> > This is not mission-critical for me but I'd be grateful for suggestions for
> > work-arounds.
>
> I don't see any workaround that's much easier than fixing the bug.
> But what's your use case that involves flipping databases from one
> tablespace to another and then back again within the life expectancy of
> unused shared-buffers pages? It doesn't seem terribly surprising that
> nobody noticed this before ...

It's a completely unreal use case. I am working on a diff tool, and
this was found as part of my test suite: build db x, drop db x, build
db y, apply diff y->x, compare with original x, apply diff x->y, compare
with original y.

So this is a fairly minor inconvenience for me.

__
Marc