Undici ConnectTimeoutError tests

This repository tests Undici for situations that lead to UND_ERR_CONNECT_TIMEOUT errors.

This error can be reproduced by blocking the event loop for at least 10 seconds (the default Undici connectTimeout) after starting a fetch request.

To reproduce, run:

pnpm test

To see debug information, you can use NODE_DEBUG options such as NODE_DEBUG=*.

This results in a log that looks like:

Starting test!
 - Using undici for fetch
 - Undici connectTimeout is set to the default of 10000 ms
--------------------
Starting fetch to http://www.example.org/?_=0
FETCH 99062: connecting to www.example.org using http:undefined
NET 99062: createConnection [
  {
    highWaterMark: 65536,
    path: undefined,
    localAddress: null,
    port: 80,
    host: 'www.example.org'
  },
  null,
  [Symbol(normalizedArgs)]: true
]
NET 99062: pipe false undefined
NET 99062: connect: find host www.example.org
NET 99062: connect: dns options { family: undefined, hints: 1024 }
NET 99062: connect: autodetecting
TIMER 99062: no 10000 list was found in insert, creating a new one
Started long running task
Long running task took 12907 ms
TIMER 99062: process timer lists 13136
TIMER 99062: timeout callback 10000
TIMER 99062: 10000 list empty
NET 99062: connect/multiple: only one address found, switching back to single connection
NET 99062: connect: attempting to connect to 93.184.215.14:80 (addressType: 4)
NET 99062: destroy
NET 99062: close
NET 99062: close handle
FETCH 99062: connection to www.example.org using http:undefined errored - Connect Timeout Error
FETCH 99062: request to GET http://www.example.org//?_=0 errored - Connect Timeout Error
TypeError: fetch failed
    at fetch (/Users/mknichel/code/tmp/undici-connect-timeout-errors/node_modules/.pnpm/[email protected]/node_modules/undici/index.js:112:13)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at fetchExample (/Users/mknichel/code/tmp/undici-connect-timeout-errors/index.ts:21:20)
    at main (/Users/mknichel/code/tmp/undici-connect-timeout-errors/index.ts:66:3) {
  [cause]: ConnectTimeoutError: Connect Timeout Error
      at onConnectTimeout (/Users/mknichel/code/tmp/undici-connect-timeout-errors/node_modules/.pnpm/[email protected]/node_modules/undici/lib/core/connect.js:190:24)
      at /Users/mknichel/code/tmp/undici-connect-timeout-errors/node_modules/.pnpm/[email protected]/node_modules/undici/lib/core/connect.js:133:46
      at Immediate._onImmediate (/Users/mknichel/code/tmp/undici-connect-timeout-errors/node_modules/.pnpm/[email protected]/node_modules/undici/lib/core/connect.js:174:9)
      at process.processImmediate (node:internal/timers:478:21) {
    code: 'UND_ERR_CONNECT_TIMEOUT'
  }
}

It looks like what is happening is:

1. A new connection is created but the attempt to connect is delayed until another round of the event loop. lookup has to make a call to GetAddrInfoReq which causes the emitLookup callback to be async and take place on the next event loop.
2. A task takes place that takes a long time.
3. The onConnectTimeout timer is run because the previous task took longer than the timeout. This calls setImmediate to destroy the socket.
4. The lookup callback (emitLookup in node:net) is run. The connection attempt begins (internalConnect is called) but that is also asynchronous, so it won't finish in this round of the event loop.
5. The setImmediate function is run, destroying the socket.
6. The error is thrown.

Making this pass

node-fetch does not exhibit this problem:

pnpm test -- --node-fetch

The default Undici connect timeout can also be increased to be longer than the blocked event loop:

pnpm test -- --connect-timeout-ms 30000