🚀 Feature: Allow hooks to easily set "allowUncaught" flag

[indexzero]

Both suite.allowUncaught and test.allowUncaught are completely ignored by the mocha.Runner instance being used at any time. (See code: https://github.com/mochajs/mocha/blob/master/lib/runner.js#L413-L416).

  if (this.allowUncaught) {
    test.allowUncaught = true;
    return test.run(fn);
  }

Which really should be something like:

  if (this.allowUncaught || test.parent.allowUncaught || test.allowUncaught) {
    test.allowUncaught = true;
    return test.run(fn);
  }

Now I'm not really sure the best way to traverse up the tree of suites to find all of the parents so the test.parent.allowUncaught is pretty naive, but the second test.allowUncaught check allows for a (somewhat) elegant work-around:

describe('Anything', function () {
  this.on('test', function (test) {
    test.allowUncaught = true;
  });
});

The super-hack work-around until I have the time to make a pull-request for this is:

//
// This is an awful and fragile hack that
// needs to be changed ASAP.
//
var _runTest = mocha.Runner.prototype.runTest;
mocha.Runner.prototype.runTest = function () {
  this.allowUncaught = true;
  _runTest.apply(this, arguments);
};

describe('Anything', function () {

  after(function () {
    mocha.Runner.prototype.runTest = _runTest;
  });
});

Worth noting that the --allowUncaught CLI option also appears to be completely broken right now. My use-case is slightly different, however, since I only want to set allowUncaught to true for some test suites (not all).

[kevinoid]

Per-test or per-suite support for .allowUncaught would be great for my uses as well.

[TimBeyer]

I could also really use this right now.
I need to assert that my server does the correct shutdown procedure and logs the errors when encountering an uncaught exception.

[mlucool]

👍

[ghost]

👍

[spalger]

I'm currently setting this on the runner and it works great:

mocha.setup(...)
requireAllZeTests()
const runner = mocha.run()
runner.allowUncaught = true

[doublerebel]

Related: #1801

[boneskull]

@spalger has a workaround, and here's another:

beforeEach(function () {
  this.runnable().allowUncaught = true;
});

it('should be uncaught', function () {
  throw new Error();
});

It's not possible to have allowUncaught be true on a per-test-case basis, because we need to know whether or not to allow the uncaught exception before we run the test. So it has to be in a beforeEach() (this.runnable() when run in a before() is undefined).

[boneskull]

I'm calling this a "bug" insofar as it's a "missing feature".

However, I don't have enough understanding of the use case(s) for allowUncaught, and why it's intended to only be run in the browser. How does it help debugging in Chrome, specifically?

[qrohlf]

@boneskull I use 'stop on uncaught errors' to automatically stop execution so I can debug a failing test where the exception occurs, with all local variables intact. It's quite helpful.

[boneskull]

@qrohlf What I was trying to get to was "why is this only a browser thing"?

[qrohlf]

@boneskull woops, missed that dimension to your question. I agree that allowUncaught should be usable in CLI/node environments.

[boneskull]

That would seem reasonable 😄, but there's some reason the intent of the feature was limited to browsers. Basically someone needs to get to the bottom of it by inspecting issues and/or changeset history.

[boneskull]

--allow-uncaught in Node.js was added by #2793; this is now a feature request to make it easier to set the flag.

I believe it is possible to set this flag from within a test, but not without some code changes.
Right now, it's only possible to use in beforeEach.

Basically the result should be:

it('should set allowUncaught', function () {
  this.allowUncaught();
  throw new Error('boom crash exit process');
});