On Mon, Feb 11, 2019 at 12:29 PM Steve Wampler <swampler@nso.edu> wrote:
> Thanks - but I thought the search_path update was a PG 10 change and so shouldn't reflect on 9.5.15 behavior. Did
it
> get back-ported?
Yes, it was deemed a security vulnerability and thus back-patched.
Release notes will indicate when that happened.
> In any event I'm surprised that pg_dump for 9.5.15 can produce a dump that can't be restored by either pg_restore
> (when -Fc is used on both ends) or with psql (without -Fc used on pg_dump). I would have expected some message
> from pg_dump if it ran into issues preventing this.
pg_dump doesn't look into function bodies to make determinations - to
it those are just strings. There are number of susceptible areas that
cannot be reliably detected during backup. That is why it is
imperative that backups are restored before being considered valid (at
the least those backups that occur subsequent to a schema change).
David J.