Add unit tests

[lpsinger]

No description provided.

[Courey]

Here is a gist that has a working version of the test in it.

There were a few changes:

1. creating the process and cleaning up the process should be done in a before/after hook. This ensures we can do proper cleanup.
2. I used spawn instead of execa. We're just streaming the output anyway so we don't really need a shell here. It doesn't block the event loop.
3. I'm running it detached.
4. I used beforeEach and afterEach so it will do the setup and cleanup on every test. You could do a before and after if you'd like to run multiple test on each setup.

[lpsinger]

4. I used beforeEach and afterEach so it will do the setup and cleanup on every test. You could do a before and after if you'd like to run multiple test on each setup.

We want to kill the process during the test because the functionality that is under test is that the process dies when it is supposed to.

[Courey]

4. I used beforeEach and afterEach so it will do the setup and cleanup on every test. You could do a before and after if you'd like to run multiple test on each setup.

We want to kill the process during the test because the functionality that is under test is that the process dies when it is supposed to.

Then cut and paste that part into the test. It will have a small pause but still work as expected.

so you'd remove the after hook, and adapt the test so it looks like this:

it('starts and stops in sandbox mode', async () => {
  let result
  for (let i = 0; i < 30 && !result; i++) {
    try {
      const response = await fetch('http://localhost:3333/')
      if (!response.ok) throw new Error(`HTTP ${response.status}`)
      result = await response.json()
    } catch {
      await sleep(1000)
    }
  }
  assert.deepStrictEqual(result?.body.cluster_name, 'elasticsearch')
  if (childProcess.pid) assert.ok(process.kill(-childProcess.pid, signal))

  await new Promise<void>((resolve, reject) => {
    childProcess.on('close', resolve)
    childProcess.on('error', reject)
  })

  childProcess.stdin?.end()
  childProcess.stdout?.destroy()
  childProcess.stderr?.destroy()
})

[lpsinger]

@Courey, I did as you suggested: I switched from execa to child_process.spawn. That seemed to do the trick. I don't know what the deal is with execa. We've definitely had other problems with it.

I also did as you suggested with splitting out the cleanup and teardown code into before and after hooks; it turns out those can contain assertions. It made the code a lot simpler.

This is now passing on my Mac, but it's failing in the Windows CI. Did you and @dakota002 ever test your approach on Windows?

[Courey]

I believe we tested the previous version on Windows.
This failure?

# Error: Test hook "afterEach" at __tests__\\test.js:50:7 generated asynchronous activity after the test ended. This activity created the error "TypeError: Cannot read properties of undefined (reading 'on')" and would have caused the test to fail, but instead triggered an unhandledRejection event.

It looks like it's failing from the actual test suite itself.
It looks like the process is undefined for some reason here.

[lpsinger]

The process is undefined because it couldn't find the executable npx and never started. That's what this error message means:

      location: 'D:\\a\\architect-plugin-search\\architect-plugin-search\\__tests__\\test.js:65:7'
      failureType: 'hookFailed'
      error: 'spawn npx ENOENT'
      code: 'ENOENT'
      stack: |-
        ChildProcess._handle.onexit (node:internal/child_process:286:19)
        onErrorNT (node:internal/child_process:484:16)
        process.processTicksAndRejections (node:internal/process/task_queues:82:21)

@dakota002, help us!

[lpsinger]

Actually, it's working now, even though I'm using execa. Mysterious!

[lpsinger]

@Courey, please review.