hmm why does it breaks travis?

I can reproduce the error locally.

  throw new assert.AssertionError({
        ^
AssertionError: no errors
    at /home/travis/build/kof/node-qunit/test/testrunner.js:49:11
    at /home/travis/build/kof/node-qunit/lib/testrunner.js:182:24

It seems the assert library is a bit useless when it comes to error reporting. Adding the following exposes the actual error.

diff --git a/test/testrunner.js b/test/testrunner.js
index fdb3eb0..221f9f7 100644
--- a/test/testrunner.js
+++ b/test/testrunner.js
@@ -46,6 +46,9 @@ chain.add('base testrunner', function() {
             failed: 2,
             passed: 5
         };
+        if (err) {
+            throw err;
+        }
         a.equal(err, null, 'no errors');
         a.ok(res.runtime > 0, 'Date was modified');
         delete res.runtime;

Error: Called start() outside of a test context too many times
    at Object.extend.start (node-qunit/node_modules/qunitjs/qunit/qunit.js:277:11)
    at _require (node-qunit/lib/child.js:66:11)
    at node-qunit/lib/child.js:177:5
    at Array.forEach (native)
    at Object.<anonymous> (node-qunit/lib/child.js:176:15)

The actual test in question:

test('myAsyncMethod test', function() {
    ok(true, 'myAsyncMethod started');

    stop();
    expect(3);

    myAsyncMethod(function(data) {
        equal(data, 123, 'myAsyncMethod returns right result');
        equal(data, 321, 'this should trigger an error');
        start();
    });
});

The stacktrace points at: https://github.com/kof/node-qunit/blob/master/lib/child.js#L66

That _require() method is called multiple times, so this looks like a bug in node-qunit that newer QUnit versions help to uncover.

yeah QUnit.start(); is called multiple times, maybe in the older version it wasn't an issue ...

FYI, this behavior "tightening" arrived in v1.16.0. The previous behavior was considered a bug, so we did not honor its backward compatibility.

Discussion: qunitjs/qunit#653 (comment)

Which hopefully implies that the bug fix needed here will be backwards compatible.

I ran into problems just trying to update node-qunit to QUnit v1.11.0: #110

Please fix this. Users are unable to use node-qunit with newer test suites because APIs introduced in the last 2 years of QUnit development (e.g. assert.expect()) fail on node-qunit which is pinned to QUnit 1.10.0 still where that method didn't exist.

@Krinkle totally forgot this one, can we already update to an even newer version?

1.20.0 was just released, but that shouldn't really matter, since the ^ includes all versions below 2.0.0.

That said, you should definitely test with 1.20.0, since it might uncover other bugs in node-qunit. For example, calling start() with a non-numeric argument will cause tests to fail (somewhat similar to the other issue discussed above).

@Krinkle would you like to become a collaborator in this project?

also @jzaefferer ?

Sure, also @leobalter, since he's the (new) QUnit project lead.

done

welcome in team @jzaefferer @Krinkle @leobalter

@kof thanks. So, are you going to put some time into fixing this or other issues, or are you hoping for one of us to take over?

@jzaefferer I would be glad to get some help on this

Okay. I've got some things to discuss with @leobalter anyway, maybe we can also do a triage of issues here.

Thanks @kof!