"David G. Johnston" <david.g.johns...@gmail.com> writes:
> On Tue, May 27, 2025 at 3:08 PM Tom Lane <t...@sss.pgh.pa.us> wrote:
>> Yeah. I'd say the fundamental problem is that pgbouncer is abusing
>> the server_version parameter, which will break plenty of things
>> besides psql --- pg_dump for instance
> They proxy over the real server information during actual bouncer work -
> it's just their local administrative database connection - one that isn't
> proxied - that is at issue here.
Okay, but they're still abusing the wire protocol: server_version is
supposed to be a Postgres version, not some random number. My advice
to them would be to make up their mind which PG version they intend
their "local administrative database" to act like, and put that in
server_version.
regards, tom lane
