Greetings,
* Tom Lane (tgl@sss.pgh.pa.us) wrote:
> Bruce Momjian <bruce@momjian.us> writes:
> > On Wed, Dec 9, 2020 at 12:19:18PM -0500, Tom Lane wrote:
> >> Ah, I'm sorry, I pointed you at the wrong catalog entirely. It's
> >> not pg_default_acl that controls this, it's pg_init_privs. I believe
> >> what pg_dump is doing is emitting GRANT commands that replicate
> >> the difference between pg_pltemplate's current actual privileges and
> >> what is shown for it in pg_init_privs. So you need to make those
> >> two things match, in whichever way is easiest.
>
> > Should pg_dump or pg_upgrade be detecting and reporting these things,
> > rather than requiring this analysis by the user?
>
> It's a little premature to be considering that when we still haven't
> identified exactly what the problem is.
Pretty sure we did get to the bottom of it, the OP was just in the wrong
database.
Anastasia, as I recall, had a patch that worked to avoid dumping out
privileges and such for things which disappeared but I don't think
anyone ended up finding time to review and commit it. I do think that,
in general, that was a good effort though and it'd be nice to get it (or
something like it) committed to avoid having this issue for users who
have GRANT'd rights to pg_catalog tables that are later changed between
versions.
Thanks,
Stephen