Re: [HACKERS] pg_upgrade and missing loadable libraries

On Sun, Jun  4, 2017 at 02:04:37PM -0400, Tom Lane wrote:
> Bruce Momjian <bruce@momjian.us> writes:
> > The problem is that in some cases extensions are improperly removed or
> > the extension has bugs that leaves pg_proc entries around that aren't
> > dumped, but are seen by pg_upgrade and generate an error.  In these
> > cases, and I have seen a few recently, we don't give the user any way to
> > find the cause except ask for assistance, i.e. we don't show them the
> > query we used to find the problem libraries.
> 
> Meh.  I think that sort of situation is one in which non-experts are
> going to need help in any case.  It's unlikely that pg_upgrade can,
> or should try to, offer them advice sufficient to fix the problem.
> 
> Also, I completely reject the idea that pg_upgrade's output should
> be optimized for that situation rather than the typical "you forgot
> to install these extensions in the new installation" case.

I didn't want to optimize for it --- I wanted a way to detect when DROP
EXTENSION has no hope of working, and give more details.  I assume the
problem with that is the the object names are inside SQL scripts that
cannot be easily interrogated.  Are the pg_proc entries tied to the
extension in some verifiable way that we could identify orphaned pg_proc
lines?

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com

+ As you are, so once was I.  As I am, so you will be. +
+                      Ancient Roman grave inscription +