polylith-kaocha acts as a bridge between poly and kaocha. Out of the box poly does its own test discovery and test running. With p-k you delegelate these to kaocha. In other words, if you want poly test to run your tests via kaocha, then p-k is an off-the-shelf solution

If you don't use poly test then it's of no use

Is there any practical difference, apart from having that particular command work?

Like, does limiting tests to only where the changes have an actual effect still work?

Then again, I think I want to have the watch mode anyway, and that's not possible via poly, I think

Yes that was the whole point. I wanted poly's project/brick filtering combined with kaocha's plugins

That's a pretty big thing

Watch mode wouod be something different I'm guessing yea

I don't use that

But to get watch mode, I guess I have to run kaocha directly, and that means duplicating all the source and test paths?

Most likely. You might be able to get those paths from the poly tool though and script the thing

Uf, choices, choices

I think I'll go with the simplest way to get watch mode at this point. Our tests are not yet so slow that the overhead would matter

And we fortunately don't have many components for the same interface yet, so we can delay the project specific configuration

@U8ZQ1J1RR How did you go about getting watch mode working? Looking into enabling kaocha, and this is my favourite feature when I used it previously.

Both watch and fail-fast

Aside from the colourful output, I'm not even sure why you would want kaocha as a test runner without the two killer features mentioned.

I've just added a tests.edn at the root and have added test plans there. Using the root`deps.edn` aliases for the construction of the classpath and test directories as per normal.

We are using plain kaocha without any cooperation with polylith. Simply duplicating the classpath in tests.edn