Re: BUG #18457: Possible data loss needs to be evaluated

On Tuesday, May 7, 2024, PG Bug reporting form <noreply(at)postgresql(dot)org>
wrote:

> The following bug has been logged on the website:
>
> Bug reference: 18457
> Logged by: Julius Ernesti
> Email address: julius(dot)ernesti(at)kpi4me(dot)de
> PostgreSQL version: 14.10
> Operating system: Rocky Linux 9
> Description:
>
> For the sake of simplicity, here is another example: There are two
> connections A and B
>
> Via A: insert into test (id, val) values (2, 'this will be a unique key
> constraint error'); --> The Rollback statement will not be executed due to
> an unknown error
> Via B: insert into test (id, val) values (4, 'no error'); --> Successful
> Transaction
> Via A: insert into test (id, val) values (5, 'no error, but the is
> currently
> a running transaction') --> Failed. Maybe a rollback is executed, I dont
> know
> Via B; select * from test where id = 4; --> Empty query
>

If you actually perform that algorithm manually using two psql sessions you
will find, if done properly, that the row with id=4 exists. If it does not
then show that work because you will indeed have found a bug. But as of
now you have not demonstrated a bug in PostgreSQL.

David J.