Upgrade qunitjs to 2.x

[mucaho]

This is an in-progress implementation that fixes #122 .

• Update bridge between phantomjs and qunit
  • Looks fine, except for undocumented option QUnit.config.autorun = false.
    It's not mentioned in upgrade guide, so it should continue to work.
• Update tests as stated in upgrade guide
• Change devDependency qunitjs from version 2.0.0-rc1 to 2.3.0
• Update documentation to reflect the change
• Things I may have missed

[jzaefferer]

Good start, thanks @mucaho!

@leobalter should we add autorun to the QUnit.config docs or mention it in the upgrade guide? How should we deal with that here?

[leobalter]

autorun is not part of QUnit 2.0. It wasn't a documented feature on 1.x as well. We discussed this on the removals for 2.0

QUnit.config.autostart should work and it's documented.

[jzaefferer]

@mucaho can you remove/replace any usage of autorun?

[mucaho]

Ok, removed autorun.

However, the test reporter seems to rely on serial execution order of tests. If QUnit will call interleaved callbacks for test cases, that could mess up reporting. (Probably most noticeable with verbose set and many asynchronous tests).
Specifically, tasks/qunit.js#failedAssertions (here, here, here, here).

One simple approach would be to match the QUnit.log callbacks to the appropriate QUnit.testStart & QUnit.testDone callbacks by comparing the name and module fields. However, this may in rare cases not work with asynchronous tests that have the same name.

[jzaefferer]

You removed the "Run tests serially, not in parallel." comment, is that where you got the idea that something is running in parallel? I'm pretty sure QUnit runs all tests in sequence, not in parallel. Or is there something else?

[stevenbenner]

QUnit 2.0.0 has been released and is up on npm now. 2.0.0 release notes

[jzaefferer]

@mucaho could you update this?

[matzetronic]

Hi,
can you give us an update of this issue? I wrote my tests already in QUnit 2.0 and I also have the issue #129 so I thought this could be related to this and how QUnit test results are reported from QUnit 2.0 tests to the 1.20 QUnit used by grunt-contrib-qunit..

[Arkni]

@matzetronic the current version of this plugin (v1.2.0) should work just fine with any version of QUnit. There is only one requirement of using versions of QUnit >= 1.23.0 in order to use the CLI flags --seed (not released yet) and --modules.

As for #129, could you please provide the steps + source code to reproduce (please put them in #129)

[vogdb]

Hiii! I'm sorry to bother but what is the status here?

[jsf-clabot]

All committers have signed the CLA.

[mucaho]

Updated the qunitjs devDependency to ^2.3.0.
Had to adapt the noglobals test file since it's no longer possible to overwrite the error reporting function pushFailure of QUnit and thus I didn't find a way to prevent QUnit from reporting a failure. If you know of a more elegant workaround, please let me know.

[leobalter]

cc @trentmwillis

[leobalter]

@Arkni it looks good to me.

[Arkni]

@mucaho
Thanks a lot :)

And sorry for forgotten about this PR!

[mgol]

@leobalter this plugin needs a major version bump now, doesn't it?

[trentmwillis]

Yes, next release should be a major version.

[leobalter]

yes, please

[Arkni]

We also need to remove this part of code phantomjs/bridge.js#L36-L39 before releasing. We don't need that BC check anymore.

[Arkni]

@leobalter
Should we also switch to use the js-reporter events (using QUnit.on(...)) instead of the the callback QUnit.log(...), ... ?
If we do that, we will need to set the min supported version of QUnit to 2.3.0

[leobalter]

@Arkni it's not necessary for now, it's still valid to use QUnit.log but it would be great if you want to try QUnit.on. The api is a bit different, thou.

[Arkni]

OK then. I'll try to remove the lines I referenced in my previous comment #123 (comment) and do a release tomorrow (in case no one beat me to it :-) ).

I'll play with it and test it on some of my projects before changing any thing in this plugin.

Thanks!

[Arkni]

Published to npm as 2.0.0 :)