Unit test command line usage is more limited than the old SpecRunner

I understand why SpecRunner was removed (to avoid maintaining it in ThirdParty separate from npm, but the command line testing workflow is more limiting.

One thing I do quite often in SpecRunner is to filter for tests. That way when I’m working on code, I can select a single test or many tests using ?spec=<search query> or the “Select a unit test to run”. This is a lot easier to work with than fit/fdescribe.

I’ve been trying out the new --includeName and it’s not finding tests from nested describe blocks as well as the ?spec=<search query> in the SpecRunner.html does.

(on branch extension-revisions) I wanted to run the tests from the describe block Scene/MetadataClassProperty > "normalize and unnormalize" but when I run --includeName 'normalize and unnormalize' (being very careful to use strong quotes in bash to avoid escaping the spaces) or even just --includeName normalize, 0 tests execute so I think this test filter doesn’t look at nested describe blocks. Meanwhile in the SpecRunner, ?spec=normalize and unnormalize works fine and gives me the 10 specs I want to run.

…I suppose I could always do fdescribe(), though I’ve gotten used to the search query workflow that doesn’t involve fiddling with the code (and accidentally committing it in the process)

Talking about this with @ebogo1, one of the limitations is that --includeName only has access to the top-level tests, not the inner describe blocks. That’s unfortunate.

@j9liu you had a few more limitations of using npm run test, could you add the details in a comment here?

[ebogo1]

When I put the Karma tab in the background, some tests about the render loop timeout.

This has been an issue for a while with the SpecRunner too, where some specs time out if the cursor moves. I couldn’t find an open issue for this but I don’t think it’s new behavior.

[IanLilleyT]

https://github.com/CesiumGS/cesium/pull/10238 is working pretty well for me, thanks! There’s a couple things that would make it even more seamless, but maybe for another PR:

• Have Karma watch files automatically instead of doing npm run build whenever there are code changes. That way I can keep the Karma window open and refresh the page whenever I make changes.
• A way to avoid clicking the Debug button when using --debug?