Message-ID: <gh-pgjdbc-pgjdbc-4080@github.com>
From: "davecramer (@davecramer)" <noreply+davecramer@github.com>
To: "pgjdbc/pgjdbc" <noreply+pgjdbc-pgjdbc@github.com>
Date: Wed, 20 May 2026 13:08:14 +0000
Subject: [pgjdbc/pgjdbc] PR #4080: Refactor pgstream to use threading
List-Id: <gh-pgjdbc-pgjdbc.github.com>
X-GitHub-Author-Id: 406518
X-GitHub-Author-Login: davecramer
X-GitHub-Issue: 4080
X-GitHub-Repo: pgjdbc/pgjdbc
X-GitHub-State: open
X-GitHub-Type: pull_request
X-GitHub-Url: https://github.com/pgjdbc/pgjdbc/pull/4080
Content-Type: text/plain; charset=utf-8

fix: make fastpath, copy, and notify protocols work with async message reader
    
    When asyncReading is enabled, the AsyncMessageReader thread owns the
    socket and feeds messages into a MessageQueue. Several protocol methods
    were still reading directly from pgStream, causing deadlocks:
    
    - receiveFastpathResult: used by Large Object operations
    - processCopyResults/initCopy: used by COPY protocol
    - processNotifies: used by async notification listening
    
    Convert all direct pgStream reads in these methods to use the
    queue-aware helpers (nextMessageType, readInt4, readBytes, etc).
    
    Add helper methods:
    - hasMessagePending(): checks queue or stream based on mode
    - peekNextMessageType(): non-blocking peek for copy protocol
    - readCopyDataLength(): handles length field differences between modes
    - readBytes(): reads raw bytes from payload or stream
    
    Make nextMessageType() use timed poll (1s) with connection-closed
    detection and socket timeout respect, preventing indefinite hangs
    when the connection is broken externally.
    
    Add processNotifies async mode path that uses timed queue polling
    instead of socket timeouts for notification wait behavior.