So, I've just switched to the <final release of Ko...

# scripting

[Sebastian Schuberth]

So, I've just switched to the final release of Kotlin 2.0.0 in my project, but now I'm getting strange failure in the test for my custom scripts: It seems that some test cases use the input for scripts from previous test cases. Has something changed e.g. WRT script caching that migrators should be aware of?

[Sebastian Schuberth]

I've created https://youtrack.jetbrains.com/issue/KT-68336/K2-does-not-seem-to-pass-the-right-constructor-arguments-to-custom-scripts for this.

[Sebastian Schuberth]

Ping @Piotr Krzemiński also from here 😉

[ilya.chernikov]

I do not expect any scripting-related changes that affect cases like yours.
I tried to debug the test briefly, and without understanding much of the internals, I see that the ortResult object that is passed to the script constructor here is exactly the same object that is created in the beginning of the test.
Why do you think it is a wrong one?
And also there are only a few failing tests. so some fundamental parameter passing bugs would probably break most of them.

[Sebastian Schuberth]

Thanks for looking into this @ilya.chernikov! I expect the incompatibleOrtResult to get passed, see here, but instead that tests seems to use the ortResult data from here (same file, earlier test case). Could there be some weird conflict between my global ortResult variable (from org.ossreviewtoolkit.evaluator.ortResult) and my DSL builder for the data class that is also called ortResult and defined here?

[ilya.chernikov]

I can imagine that some general K2 changes, e.g. in resolution, can have such effect.
But formally, what I see in the debugger is the exactly what you describe as a desired state, at least speaking about objects, not the contents.
If I'm not mistaken, of course. Lets check my screenshots - the incompatibleOrtResult is marked.
First frame at the test body, and second - before constructing the script class

[Sebastian Schuberth]

Screenshot-wise that looks indeed good. But with K2 the test fails with debug lines like

12:13:49.202 [pool-1-thread-1] INFO  org.ossreviewtoolkit.evaluator.Rule - Evaluating rule 'OSADL_MATRIX_COMPATIBILITY' for dependency 'Maven:org.ossreviewtoolkit:package-excluded:1.0' (project=Maven:org.ossreviewtoolkit:project-excluded:1.0, scope=compile, level=0).
12:13:49.203 [pool-1-thread-1] INFO  org.ossreviewtoolkit.evaluator.Rule - 	!(isExcluded()) == false
12:13:49.203 [pool-1-thread-1] INFO  org.ossreviewtoolkit.evaluator.Rule - 	Rule skipped.

where Maven:org.ossreviewtoolkit:project-excluded:1.0 is a project that's not even present in incompatibleOrtResult, but only in the ortResult global test data. Reverting to Kotlin 1.9.24 makes the test pass again.

[Sebastian Schuberth]

And please don't get me wrong, it could totally be that I'm doing something shady (without knowing) that just happened to work in Kotlin 1.9.24, but actually shouldn't, and was fixed in K2, so I'm now facing it. I'm just completely clueless what could cause this kind of behavior.

[Sebastian Schuberth]

But you can reproduce the test failing with K2, right @ilya.chernikov?

[ilya.chernikov]

So, I figured out. The problem is changed resolution scheme for scripts, where imported declarations are preferred. You have the org.ossreviewtoolkit.evaluator.ortResult imported by default, and the access is resolved to it instead of the constructor parameter.
Actually the usage of the constructor parameters is obsolete, we should probably issue a warning there. The recommended approach is to use providedProperties. The patch below fixes it properly (and it should work for K1 as well).
But thank you for the issue - we should definitely address it in some way.

[ilya.chernikov]

Untitled

Untitled

[Sebastian Schuberth]

Thanks a lot for the speedy investigation @ilya.chernikov, that's highly appreciated!

[Sebastian Schuberth]

Unfortunately, I'm running into the next problem: providedProperties from script compilation and script evaluation do not add up, but replace each other. Is this expected?

[Sebastian Schuberth]

Ah, no scratch that. I was calling providedProperties for evaluation twice.

[Sebastian Schuberth]

One more feedback yet: After migrating constructorArgs to providedProperties, IntelliJ does not seem to be able to resolve them in scripts anymore... e.g. my ortResult is now always marked in red, although the scripts runs fine:

[ilya.chernikov]

Ok, I see. Yes, the IDE support is far from ideal, but we're working on it. Please submit an issue about it too.

[Sebastian Schuberth]

Oh, I was expecting there surely already is an issue about this already, as it's a quite obvious limitation. I'll check first to be on the safe side. Interestingly, it seems to work in Fleet, though.

[Sebastian Schuberth]

Is it maybe https://youtrack.jetbrains.com/issue/KTIJ-1194?

[ilya.chernikov]

No, it is a different different problem.

[Sebastian Schuberth]

Thanks for confirming, I've created a new issue: https://youtrack.jetbrains.com/issue/KT-68472/No-proper-syntax-highlighting-for-providedProperties-of-custom-scripts

[ilya.chernikov]

Thanks! We'll try to take care of it.