testcafe-hammerhead: Uncaught exception: socket has been ended by the other party

Hello!

I implemented a runner for the project I’m working on using the TestCafe API. With my coworkers, we noticed that some tests started failing after we upgraded NodeJS past version 10.15.3 (basically, no problem with 10.15.3 or below, but anything newer and we get the issue described here)

We randomly receive the following error:

× Scenario: Successful registration process (single step)
   1) Uncaught exception:
      Error: This socket has been ended by the other party
      at TLSSocket.writeAfterFIN [as write] (net.js:407:14)
      at Socket.ondata (_stream_readable.js:713:22)
      at Socket.emit (events.js:200:13)
      at Socket.EventEmitter.emit (domain.js:471:20)
      at addChunk (_stream_readable.js:294:12)
      at readableAddChunk (_stream_readable.js:275:11)
      at Socket.Readable.push (_stream_readable.js:210:10)
      at TCP.onStreamRead (internal/stream_base_commons.js:166:17)
      Browser: Chrome 84.0.4147.125 / Windows 10

Otherwise, the test itself does not have any reason to fail (one test will be fine for several runs, then fail with that error - and only that error)

I manage to trigger it by chaining a lot of navigateTo to our website. => I usually get a few fails out of 10 identical tests (see the test code)

My investigation so far shows that it seems to be linked to our usage of WebSocket. The issue only happens with Chrome, but we want coverage of all browsers (plus Firefox seems to not always close properly but that’s an issue for another time) Downgrading NodeJS solves the issue but the rest of the project now relies on a newer version, meaning we have to switch back and forth, which is not ideal.

I’m not sure you will have access to the website in the example - it is protected against access out of belgium. I might be able to arrange something if needed. I’ve reproduced the issue with another website (a completely different front-end, that we started from scratch so it does not have much in common with the current site) I was NOT able to reproduce the issue with external websites that use WebSocket (again, we might have a heavier load on the WS than the websites I tried)

fixture`Test the runner`;

test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});
test('Basic V3 test', async (t) => {
  await t.navigateTo('https://www.circus.be').wait(5000);
});

Your Environment details:

• testcafe version: 1.9.1
• node.js version: 12.18.3
• browser name and version: Chrome 84
• platform and version: Windows 10

report.zip