Output message when t.throws() fails

[timdeschryver]

Fixes #1342

Maybe the test can be written better?

[timdeschryver]

I had to remove the catch because otherwise I couldn't access the AssertionError in the test.

[novemberborn]

The catch is necessary. We don't expose the AssertionError instance to the test author. Instead add a .catch() callback in the pending() implementation here: https://github.com/tdeschryver/ava/blob/4a19a96da89a1533025295a7cd78b74c12528f04/test/assert.js#L15.

[novemberborn]

Looks good! The test does need to be approached differently. The promise returned by t.throws() should never reject.

[novemberborn]

I think it's clearer to just repeat this property.

[novemberborn]

The catch is necessary. We don't expose the AssertionError instance to the test author. Instead add a .catch() callback in the pending() implementation here: https://github.com/tdeschryver/ava/blob/4a19a96da89a1533025295a7cd78b74c12528f04/test/assert.js#L15.

[novemberborn]

Rather than extracting the comparison logic into testFailResults, if you capture lastFailure through the pending() function you could write an eventuallyFailsWith():

function eventuallyFailsWith(t, promise, subset) {
  return promise.then(() => failsWith(t, () => {}, subset))
}

Given that assertion.throws() should return a promise that always resolves, and the error will be captured separately.

[timdeschryver]

Thanks for the tip and reviews!
I got it to work now :), but I had to pass the lastFailure to the failsWith function otherwise it would be assigned to null. Is this OK?

[novemberborn]

My bad, overlooked that failsWith() resets lastFailure.

Instead extract the failure testing code into its own function, e.g.:

function assertFailure() {
  if (!lastFailure) {
  		t.fail('Expected assertion to fail');
 		return;
 	}
 
 	t.is(lastFailure.assertion, subset.assertion);
 	t.is(lastFailure.message, subset.message);
 	t.is(lastFailure.name, 'AssertionError');
 	t.is(lastFailure.operator, subset.operator);
 	if (subset.statements) {
 		t.is(lastFailure.statements.length, subset.statements.length);
 		lastFailure.statements.forEach((s, i) => {
 			t.is(s[0], subset.statements[i][0]);
 			t.match(s[1], subset.statements[i][1]);
 		});
 	} else {
 		t.same(lastFailure.statements, []);
 	}
 	if (subset.values) {
 		t.is(lastFailure.values.length, subset.values.length);
 		lastFailure.values.forEach((s, i) => {
 			t.is(s.label, subset.values[i].label);
 			t.match(s.formatted, subset.values[i].formatted);
 		});
 	} else {
 		t.same(lastFailure.values, []);
  	}
  }
}

And call that from failsWith() and eventuallyFailsWith() (so the latter doesn't call failsWith() anymore). eventuallyFailsWith() should also reset lastFailure.

[timdeschryver]

That's indeed a cleaner solution.
Changes are commited!

[sindresorhus]

Thank you @tdeschryver :)