`test_pthread_print_override_modularize` is failing on macOS after node upgrade to v16

[sbc100]

Ever since we updated the version of node running on the emscripten-releases builder or (https://chromium-review.googlesource.com/c/emscripten-releases/+/4634060) this test has been failing consistenly, but only on macOS and only on the emscripten-releases waterfall (not on the github CI mac bots).

See https://logs.chromium.org/logs/emscripten-releases/buildbucket/cr-buildbucket/8777589317403048353/+/u/Emscripten_testsuite__other_/stdout

FAIL: test_pthread_print_override_modularize (test_other.other)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/opt/s/w/ir/cache/vpython/store/cpython-ste21jq03jlov0apfh16dstb30/contents/lib/python3.8/unittest/case.py", line 60, in testPartExecutor
    yield
  File "/opt/s/w/ir/cache/vpython/store/cpython-ste21jq03jlov0apfh16dstb30/contents/lib/python3.8/unittest/case.py", line 676, in run
    self._callTestMethod(testMethod)
  File "/opt/s/w/ir/cache/vpython/store/cpython-ste21jq03jlov0apfh16dstb30/contents/lib/python3.8/unittest/case.py", line 633, in _callTestMethod
    method()
  File "/opt/s/w/ir/x/w/install/emscripten/test/common.py", line 227, in decorated
    f(self, *args, **kwargs)
  File "/opt/s/w/ir/x/w/install/emscripten/test/test_other.py", line 5960, in test_pthread_print_override_modularize
    self.assertContained('hello, earth!', output)
  File "/opt/s/w/ir/x/w/install/emscripten/test/common.py", line 1112, in assertContained
    self.fail("Expected to find '%s' in '%s', diff:\n\n%s\n%s" % (
  File "/opt/s/w/ir/cache/vpython/store/cpython-ste21jq03jlov0apfh16dstb30/contents/lib/python3.8/unittest/case.py", line 753, in fail
    raise self.failureException(msg)
AssertionError: Expected to find 'hello, earth!
' in 'received "loaded" command from terminated worker: 1
received "callHandler" command from terminated worker: 1
', diff:

--- expected
+++ actual
@@ -1 +1,3 @@
-hello, earth!
+received "loaded" command from terminated worker: 1
+received "callHandler" command from terminated worker: 1
+

It seems like some kind of race between the thread/program exiting and postMessages being received from the workers.. but I can't seem to reproduce it.

[dschuff]

I had some luck awhile back getting some of the other flaky failures to reproduce on Linux by pinning a bunch of processes to a few CPUs to slow everything down.

[sbc100]

This flake looks like the same issue with races on pthread shutdown that we see in other flaky tests.

[jkukkula]

We hit this in our tests.
Very hasty debug session makes it appear that the promise returned by worker.terminate(); is not awaited to resolve before the worker's onmessage handler is made to throw an error, so it's possible to get messages from worker still hit the message handler.

The behaviour of worklet.terminate() has changed in node v12.5.0, so the implementation probably was ok until that: https://nodejs.org/api/worker_threads.html#workerterminate

This function now returns a Promise. Passing a callback is deprecated, and was useless up to this version, as the Worker was actually terminated synchronously. Terminating is now a fully asynchronous operation.