Sigh. I know <@U0FMXU16E> warned us, but IntelliJ ...

# minutest

[dmcg]

Sigh. I know @christophsturm warned us, but IntelliJ now defaulting to Gradle to run tests is a royal pain. Not only do the names get rendered badly, but if you have org.gradle.caching=true then tests won't run at all if their results are cached.

[christophsturm]

is there a youtrack for the second problem? there should probably be issues for both

[dmcg]

You can at least revert

[dmcg]

https://youtrack.jetbrains.com/issue/IDEA-210779

[mkobit]

the caching thing is definitely not great

for the names, there is https://github.com/gradle/gradle/issues/5975 which probably factors into it

[danny]

I thought they only delegated build/run by default, not test?

[dmcg]

So build and run are through Gradle now? Test definitely changed to Gradle when I finally upgraded this morning, at least until I found the setting above.
I’m actually in favour of the change overall - Gradle seems to be smart enough these days for me not to want a different IDE and CI build system. It’s just a shame that it doesn’t work well with JUnit dynamic tests. Maybe it’s time to shift over to our own test engine and deprecate the dynamic tests

[danny]

The 'Run tests using' setting is unchanged - you'd have to set that to the Gradle test runner to actually have Gradle run the tests

[danny]

Nice thing about the Gradle runner is it can run your tests in parallel

[dmcg]

IntelliJ helpfully set Run Tests Using to Gradle on upgrade. I’m pretty sure it wasn’t that before.

[dmcg]

I might try the parallel running - I’ve relied on parallel module building up to now, but it doesn’t help with a single module like Minutest. Does IntelliJ descramble the console output?

[danny]

Yeah - those tests are off running in individual forks and Gradle's reporting the output back

[danny]

I usually just do -

tasks.withType(Test) {
        maxParallelForks = gradle.startParameter.maxWorkerCount
}

[danny]

maxWorkerCount is the number of cores on your machine

[christophsturm]

junit also has its own parallel running support and that is even faster because it does not fork