🐛 Bug: process.exit(0) ends mocha silently

[BridgeAR]

Bug Report Checklist

• I have read and agree to Mocha's Code of Conduct and Contributing Guidelines
• I have searched for related issues and issues with the faq label, but none matched my issue.
• I have 'smoke tested' the code to be tested by running it outside the real test suite to get a better sense of whether the problem is in the code under test, my usage of Mocha, or Mocha itself.
• I want to provide a PR to resolve this

Expected

Mocha should patch process.exit() to throw an error in case that method is called, since it ends the whole test suite immediately.

That might be alright for non zero exit codes since the test suite would then be detected as erroring, while I believe other tests should in fact continue, similar to an uncaught exception or an unhandled rejection letting other tests continue to run.

Actual

Calling process.exit(0) in a test silently ends the process and no other test continues.

We have a big test suite and a library we are using was misconfigured. Due to that, that library called process.exit(0), which immediately ended the test suite and it swallowed other issues. This is a pretty big concern.

A mitigation is to monkey patch process.exit in a beforeAll hook and to reset that in an afterAll hook.

Minimal, Complete and Verifiable Example

'use strict'

const { describe, it, afterEach } = require('mocha')

describe('code origin', () => {
  afterEach(() => {
    console.log('afterEach')
  })

  describe('entryTags', () => {
    it('should collect expected entry tags if not given a function', () => {
      console.log('should collect expected entry tags if not given a function')
    })

    it('should collect expected entry tags if given a function', () => {
      console.log('should collect expected entry tags if given a function')
      process.exit(0)
    })

    it('should find a user frame if the stack trace limit is set to 0', () => {
      console.log('should find a user frame if the stack trace limit is set to 0')
    })
  })
})

Versions

11.7.5

Additional Info

No response

[mark-wiemer]

Simpler repro:

describe("suite", () => {
  it("works", () => {
    process.exit(0);
  });
});

No fun, I'll see what we can do here. This is also on main branch, commit 430960c

A mitigation is to monkey patch process.exit in a beforeAll hook and to reset that in an afterAll hook.

Glad this workaround does indeed work. In general I don't expect "well-configured" test suites to rely on process.exit at the top-level, of course, but you never know. Fixing this might require a lot of work, unfortunately, and since there's a workaround we may end up closing this as not planned. Curious for thoughts from @mochajs/maintenance-crew

[JoshuaKGoldberg]

I agree this is something Mocha should handle. It's somewhere in the murky areas between "bug" and "feature request". By design, Mocha runs everything in the same process, and if code exits the process... we'll, that's the "design". But obviously that behavior isn't desirable. 🤷

major: v11

That being said, I think we should land this only in v12, not in v11 (i.e. only on the current main, not backport to the 11.x branch). My experience from libraries like https://github.com/JoshuaKGoldberg/console-fail-test has been that plenty of folks do weird things around built-ins and globals. Any sort of global modification to the console, process, etc. will probably break somebody.