Improve `test` BSP service

[oyvindberg]

• add TestRun and TestRuns structures to aggregate all produced TestSuiteEvent.Results.
• propagate all test results back up to service
• use test results to return bsp.StatusCode.Error on failed tests (Fixing #960)
• ensure that the compilation before test finished with StatusCode.Ok

[oyvindberg]

I agree there should be tests. I'm thinking to duplicate the structure for the compile tests and adapt the structure for test end point. or something like that. It'll be a day or two before I get to it

[oyvindberg]

added a test now

[oyvindberg]

Just an update.

I've had an hour here and there over the last week, and numerous times I've jumped in to fix this test. I'll probably get there, but it's incredibly slow and difficult work.

[tgodzik]

Just an update.

I've had an hour here and there over the last week, and numerous times I've jumped in to fix this test. I'll probably get there, but it's incredibly slow and difficult work.

Do you need help with that? What is slowing you down most?

[oyvindberg]

It would help a lot. I'm working towards an MVP of bleep and really need this feature, but I have much less time than I would have wanted.

It's a deep call stack, and some state handling I haven't had time to understand.

• I'm currently stuck debugging why my tests within the test harness are not run because compileAnalysis is empty (see TestTask.scala:83)
• I commented out that check, and then it's very puzzling that I don't see all the test events this PR needs when running the tests. I hope that is not an junit vs scalatest issue. if it is, the main code in this PR needs to adapt to whatever cues sbt uses to determine failed tests

[oyvindberg]

Re slow, it's a combination of many things. I'm unable to run sbt in arm64 mode (is there even an issue for that?), so everything with the build is just incredibly slow. roundtrip for compiling frontend plus tests and then testOnly the test I want to run is like half a minute. I needed to change the default timeout value for the tests from 5 to 10 seconds for things to work at all. I know much of this responsibility lies outside of bloop, and I know the build itself is beeing looked at, just wanted to air some frustration :)

[tgodzik]

I will try and take a look, but that might be next week unfortunately.

[oyvindberg]

That would be perfect. Gives me some time to try to figure it out myself first :)

[oyvindberg]

thanks @tgodzik !