Cannot catch EADDRINUSE

[c7hm4r]

Hello! I want to implement that a different port is tried automatically when the currently tried out port is in use. However, it is impossible to catch the EADDRINUSE exception when running await node.start();. NodeJS terminates before node.start() returns a Promise.

Following is output:

node:events:491
      throw er; // Unhandled 'error' event
      ^

Error: listen EADDRINUSE: address already in use :::8000
    at Server.setupListenHandle [as _listen2] (node:net:1485:16)
    at listenInCluster (node:net:1533:12)
    at Server.listen (node:net:1621:7)
    at file:///project/node_modules/it-ws/dist/src/server.js:14:25
    at new Promise (<anonymous>)
    at Server.listen (file:///project/node_modules/it-ws/dist/src/server.js:12:22)
    at WebSocketListener.listen (file:///project/node_modules/@libp2p/websockets/dist/src/listener.js:72:27)
    at DefaultTransportManager.listen (file:///project/node_modules/libp2p/dist/src/transport-manager.js:155:37)
    at DefaultTransportManager.start (file:///project/node_modules/libp2p/dist/src/transport-manager.js:44:20)
    at file:///project/node_modules/@libp2p/components/dist/src/index.js:67:29
Emitted 'error' event on WebSocketServer instance at:
    at Server.emit (node:events:513:28)
    at emitErrorNT (node:net:1512:8)
    at process.processTicksAndRejections (node:internal/process/task_queues:82:21) {
  code: 'EADDRINUSE',
  errno: -98,
  syscall: 'listen',
  address: '::',
  port: 8000
}

Node.js v18.9.0

To me, the issue has a medium severity, because the workaround to check the port usage before is tedious, platform dependent, and cannot be reliable because it is not atomic.

Versions:

• Node.js: 18.9.0
• libp2p: 0.39.1
• @libp2p/websockets: 3.0.3

Possibly related:

• Catch errors thrown when Websocket throws Connection Refused #77

[achingbrain]

You could use 0 for the port number, then read the relevant addr from getMultiaddrs on the node after startup?

Might be a nicer approach than trying (potentially) lots of new sockets until you find a free one?

[c7hm4r]

Thank you. It would be a partial solution for my problem at hand. But in general, it would be nice to be able to at least print a human-friendly error message instead of confronting the user with some technical stack trace.

[achingbrain]

I think this should be validated - if a port is in used the error should be catchable, it should not take the process down.

From the error message the WebSocketServer needs a handler listening for the error event.

[mpetrunic]

alanshaw/it-ws#34 should fix this but it's not working for some reason