[kankankanp]

What kind of change does this PR introduce?

A test-only change to make a NULL password visible in the PgBouncer SQL regression output. Related to #1568.

What is the current behavior?

The expired-user check calls select pgbouncer.get_auth(...). That displays the function's two fields as one composite value:

(test_expired_user_password,)

The empty password field is easy to miss. Setting \pset null alone does not change it, because psql sees the composite value as one non-null column.

What is the new behavior?

The check uses select * from pgbouncer.get_auth(...) to display username and password as separate columns, and sets \pset null '<null>' for the test:

          username          | password 
----------------------------+----------
 test_expired_user_password | <null>

The SQL test and its matching expected output are updated together. The authentication function and the other checks are unchanged.

Additional context

• Review focus: The expected output should make the expired password's NULL value explicit without changing what the test checks.
• Local verification: PostgreSQL 14 psql -a -q -X produced the expected output for a row with a null password.
• CI: External nixbot/nix-eval and nixbot/nix-build checks passed. GitHub Actions workflows are awaiting maintainer approval for this fork PR and have not run yet. The full Nix regression suite was not run locally because Nix is not installed here.
• No migration or production behavior change.

[kankankanp]

Hi @samrose, could you please approve the GitHub Actions workflow runs for this fork PR when convenient? They are currently awaiting maintainer approval. For context, #1567 added the pgbouncer.get_auth regression test, and this PR makes the NULL password visible in its output, as requested in #1568. Thank you for your time and consideration.

[samrose]

@kankankanp I am no longer the maintainer of this repo. Those who are can review when they are ready to