From: Tom Lane <tgl@sss.pgh.pa.us>
To: "David G. Johnston" <david.g.johnston@gmail.com>
cc: Mark Hill <Mark.Hill@sas.com>,
        pgsql-general <pgsql-general@lists.postgresql.org>,
        Phil Rau <Phil.Rau@sas.com>, Michael King <Michael.King@sas.com>,
        Ashley Stephens <Ashley.Stephens@sas.com>
Subject: Re: Error Creating pgcrypto Extension
In-reply-to: 
 <CAKFQuwaCnT854sPzRJGnO83ReCtKoi9DcFxj0cxBe617uZbxaA@mail.gmail.com>
References: 
 <BYAPR05MB6632AA793C3F51F016FE752EEEB82@BYAPR05MB6632.namprd05.prod.outlook.com>
 <CAKFQuwaCnT854sPzRJGnO83ReCtKoi9DcFxj0cxBe617uZbxaA@mail.gmail.com>
Comments: In-reply-to "David G. Johnston" <david.g.johnston@gmail.com>
	message dated "Thu, 17 Sep 2026 16:11:52 -0500"
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-ID: <947661.1789689902.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 17 Sep 2026 20:05:02 -0400
Message-ID: <947662.1789689902@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/947662.1789689902%40sss.pgh.pa.us>
Precedence: bulk

"David G. Johnston" <david.g.johnston@gmail.com> writes:
> On Thu, Sep 17, 2026 at 3:25=E2=80=AFPM Mark Hill <Mark.Hill@sas.com> wr=
ote:
>> We are trying to run this code in Postgres 18.6:
>> *1)  Should pgcrytpo=E2=80=941.3.sql be trying to create gen_random_uui=
d() if/when
>> it already exists?*

> It's not expecting to be installed in the pg_catalog schema since it is =
an
> extension.  Extensions don't usually conditionally install their compone=
nts.

Yeah.  As of now, installing pgcrypto into pg_catalog is unsupported.

>> *2) given what=E2=80=99s in pgcrypto.control should there be a pgcrypto=
=E2=80=941.4.sql
>> file?*

> The pattern of a base full install and then incremental update scripts i=
s
> supported.

Installing 1.3 and then updating to a hypothetical 1.4 wouldn't work
anyway in this scenario, because installing 1.3 would fail.

We could perhaps decide that pgcrypto doesn't need to provide
gen_random_uuid anymore; maybe that backwards-compatibility hack
has served its time.  But I'm not sure if removing the function
would break anybody.  If we did go that way, we'd need to provide
a 1.3-to-1.4 update script that drops the function, and also a
pgcrypto--1.4 install script that rolls up all the prior updates
so that it never tries to create gen_random_uuid in the first place.

On the whole, seeing that installing pgcrypto into pg_catalog has
not worked since v13 and there's been hardly any complaints about
that, I'm not sure that we should risk breaking working setups to
allow it.

			regards, tom lane