如何复现TCP服务器的Broken pipe错误以测试解决方案？

Got it, let’s walk through reliable, easy-to-follow ways to trigger that SIGPIPE error so you can test your fix thoroughly. Here are the most common scenarios that cause this issue:

1. Manual Client Disconnect (Quick & Dirty) #

This is the simplest way to replicate the error, perfect for initial testing:

• Start up your TCP server as you normally would.
• Use a basic terminal client like nc (netcat) or telnet to connect to your server’s listening port:

  nc localhost <your-server-port>
  

• Immediately terminate the client (press Ctrl+C in the nc window, or just close the terminal entirely) without waiting for the server to send any data.
• Now, trigger your server to send data to this disconnected client connection. The moment the server calls send() or write() on that broken socket, the SIGPIPE signal will be thrown—exactly the error you’re seeing.

2. Programmable Client (Controlled Testing) #

If you need a repeatable, scripted test case, write a tiny client program to simulate the disconnect:
Here’s a quick Python example that connects then immediately closes the socket:

import socket
import time

# Replace <your-server-port> with your actual server port number
SERVER_PORT = <your-server-port>

# Establish connection to server
client_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
client_socket.connect(("localhost", SERVER_PORT))

# Close the socket right away—no data sent or received
client_socket.close()

# Wait a second to ensure the TCP FIN packet reaches the server
time.sleep(1)

Run this script, then make your server attempt to send data to the now-closed connection. You’ll get the SIGPIPE error every time.

3. Verify Connection State First (Optional) #

To make sure you’re targeting a truly broken connection, you can check the socket state before triggering the send:

• After the client connects, run this command to list active TCP connections:

  ss -tulpn | grep <your-server-port>
  

  You’ll see the connection marked as ESTABLISHED.
• After disconnecting the client, re-run the command—you’ll see the state change to TIME_WAIT or CLOSE_WAIT. Any send() call to this socket will trigger SIGPIPE.

A quick note: By default, Unix systems send SIGPIPE when writing to a broken socket, and if your server isn’t handling this signal, it will crash. If you’ve already added signal handling code temporarily comment it out so you can confirm the error is being triggered correctly.

内容的提问来源于stack exchange，提问作者pm40