Message-ID: <gh-pgjdbc-pgjdbc-3975@github.com>
From: "vlsi (@vlsi)" <noreply+vlsi@github.com>
To: "pgjdbc/pgjdbc" <noreply+pgjdbc-pgjdbc@github.com>
Date: Thu, 19 Mar 2026 10:55:33 +0000
Subject: [pgjdbc/pgjdbc] PR #3975: fix: fix flaky replication restart tests by waiting for confirmed_flush_lsn
List-Id: <gh-pgjdbc-pgjdbc.github.com>
X-GitHub-Author-Id: 213894
X-GitHub-Author-Login: vlsi
X-GitHub-Issue: 3975
X-GitHub-Labels: building-and-testing
X-GitHub-Repo: pgjdbc/pgjdbc
X-GitHub-State: merged
X-GitHub-Type: pull_request
X-GitHub-Url: https://github.com/pgjdbc/pgjdbc/pull/3975
Content-Type: text/plain; charset=utf-8

The replicationRestartFromLastFeedbackPosition* tests were flaky because they called forceUpdateStatus() and then immediately closed the replication connection. forceUpdateStatus() only flushes the status update to the TCP socket — if the walsender has not read it before the connection is closed, the slot's confirmed_flush_lsn is not advanced and the slot replays from an earlier position on restart.

Fix: after forceUpdateStatus(), poll pg_replication_slots until confirmed_flush_lsn matches the expected value before closing the connection. This ensures the server has processed the feedback.