test: fix flaky test-async-wrap-uncaughtexception

[Trott]

It's possible for beforeExit to be emitted more than once if work is
scheduled by its listeners. Accommodate this fact in the test.

Fixes: #16210

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

Affected core subsystem(s)

test async-wrap

[Trott]

ping @joaocgreis @addaleax @joyeecheung I want to make sure this is the correct fix and not working around a legitimate bug in core that needs fixing. See #16210 for background/details.

[Trott]

CI: https://ci.nodejs.org/job/node-test-pull-request/11083/

[Trott]

@nodejs/testing

[joyeecheung]

The process.removeAllListeners('uncaughtException'); in L14 can be removed now?

[refack]

Not convinced this is the right fix.

[refack]

I'm not sure this is the right fix.

1. This seems like a noop change. If there is a second call to 'uncaughtException' then the process will err-exit anyway.
2. I'm not sure I understand why this is non-deterministic.

[addaleax]

I kind of have to side with @refack here – if this is a fix, then it’s not obvious where the second uncaught exception is coming from. From the looks of it, that would not be a bug in the test but rather a bug

[joyeecheung]

@refack @addaleax Maybe we should just remove the removeAllListeners('uncaughtException') line? If there are more uncaught exceptions(i.e. a bug), the mustCall wrapping the listener would fail. If there are not, not removing the listeners does not block the process from exiting anyway.

[refack]

Some stress tests:
master: https://ci.nodejs.org/job/node-stress-single-test/1475/nodes=ubuntu1604-64/console 🔴
PR as is: https://ci.nodejs.org/job/node-stress-single-test/1476/nodes=ubuntu1604-64/console ✔️
PR rebased: https://ci.nodejs.org/job/node-stress-single-test/1477/nodes=ubuntu1604-64/console 🔴

[refack]

@joyeecheung testing (refack/node@master...refack:pr/16412)
https://ci.nodejs.org/job/node-stress-single-test/1478/nodes=ubuntu1604-64/ ✔️ [wrong params]
f7a34bdd02ec14cff680105900eca217957f0530 - https://ci.nodejs.org/job/node-stress-single-test/nodes=ubuntu1604-64/1481/console ✔️
rerun e567402 - https://ci.nodejs.org/job/node-stress-single-test/1482/nodes=ubuntu1604-64/console 🔴
99eae86aa0ed5f87e00522e1d60df3945dc0b79c - https://ci.nodejs.org/job/node-stress-single-test/1483/nodes=ubuntu1604-64/console ✔️
87efc69ecdd3673c3553d57137902db953743915 - https://ci.nodejs.org/job/node-stress-single-test/1484/nodes=ubuntu1604-64/console

[refack]

OK so there is definatly a change/regression. Seeing this locally with officialy build binaries:

8.6.0 ✔️

d:\code\node-github-desktop$ node8.6.0 D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js
uncaughtException Error: ah crap
    at RandomBytes.require.randomBytes.common.mustCall (D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js:51:9)
    at RandomBytes.ondone (D:\code\node-github-desktop\test\common\index.js:525:15)
beforeExit
    at process.on (D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js:17:11)
    at emitOne (events.js:115:13)
    at process.emit (events.js:210:7)
RANDOMBYTESREQUEST(5): trigger: 1 execution: 1
before:  5
  Immediate(6): trigger: 5 execution: 5
after:   5
before:  6
after:   6
destroy: 5
destroy: 6

8.7.0 🔴

d:\code\node-github-desktop$ node8.7.0 D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js
uncaughtException Error: ah crap
    at RandomBytes.require.randomBytes.common.mustCall (D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js:51:9)
    at RandomBytes.ondone (D:\code\node-github-desktop\test\common\index.js:525:15)
beforeExit
    at process.on (D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js:17:11)
    at emitOne (events.js:115:13)
    at process.emit (events.js:210:7)
RANDOMBYTESREQUEST(5): trigger: 1 execution: 1
before:  5
  Immediate(6): trigger: 5 execution: 5
after:   5
before:  6
after:   6
destroy: 5
destroy: 6

beforeExit
    at process.on (D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js:17:11)
    at emitOne (events.js:115:13)
    at process.emit (events.js:210:7)
RANDOMBYTESREQUEST(5): trigger: 1 execution: 1
before:  5
  Immediate(6): trigger: 5 execution: 5
after:   5
before:  6
after:   6
destroy: 5
destroy: 6

D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js:20
  if (++count > 1) throw e;
                   ^

Error
    at process.on (D:\code\node-github-desktop\test\parallel\test-async-wrap-uncaughtexception.js:17:11)
    at emitOne (events.js:115:13)
    at process.emit (events.js:210:7)

[Trott]

@refack git bisect to the rescue?

[Trott]

I think I found the commit that is the likely culprit. See #16210 (comment)

Closing this for now.