Re: HELP! BUG? pg_dump mucks up grant/revoke

Tom Lane wrote:
>
> Lamar Owen <lamar(dot)owen(at)wgcr(dot)org> writes:
> > A pg_dump that can't be restored should be a release forcer, IMHO.
>
> Let's not panic here. The actual net effect of these bugs will be that
> the pg_dump output script fails to restore the permissions and/or
> comment associated with a view. Which'd be annoying, but hardly
> something to ruin your whole day.

Maybe it won't ruin your whole day, but if you are unaware of the
problem, it can sure ruin a few hours.

I found out about the problem maybe a day before it hit the newsgroup,
but had not identified the cause -- only the effect. It turns out our
live servers refer to a view, and I was stuck diagnosing the problem
with a two year-old update script that was suddenly broken, while trying
to fight off the people who wanted new content uploded. Combine that
with a major upgrade that requires an initdb, and I'm sure you will have
complaints.

IMHO, it speaks well of Redhat and Linux's user-orientation that Lamar
and Trond are seem to be planning to release a patched RPM. Sort of
reaffirms why Linux has had some success making inroads against windows.
And I begin to see why some other unix variants have been accused of
hostility to their users.

It won't matter to me one bit. I have already applied the patch. But If
you want to protect PostgreSQL's growing reputation, a release would
seem to be worthwhile, again IMHO.

--
Karl