This is a blast from the past: ```Caused by: ksp.o...

# kotlin-inject

[ralf]

This is a blast from the past:

Caused by: ksp.org.jetbrains.kotlin.analysis.api.lifetime.KaInvalidLifetimeOwnerAccessException: Access to invalid ksp.org.jetbrains.kotlin.analysis.api.platform.lifetime.KotlinAlwaysAccessibleLifetimeToken@33adf999: PSI has changed since creation
        at ksp.org.jetbrains.kotlin.analysis.api.fir.symbols.KaFirNamedFunctionSymbol.getPsi(KaFirNamedFunctionSymbol.kt:247)
        at com.google.devtools.ksp.impl.symbol.kotlin.UtilKt.toContainingFile(util.kt:237)
        at com.google.devtools.ksp.impl.symbol.kotlin.AbstractKSDeclarationImpl$containingFile$2.invoke(AbstractKSDeclarationImpl.kt:79)
        at com.google.devtools.ksp.impl.symbol.kotlin.AbstractKSDeclarationImpl$containingFile$2.invoke(AbstractKSDeclarationImpl.kt:78)
        at kotlin.SynchronizedLazyImpl.getValue(LazyJVM.kt:83)
        at com.google.devtools.ksp.impl.symbol.kotlin.AbstractKSDeclarationImpl.getContainingFile(AbstractKSDeclarationImpl.kt:78)
        at me.tatarka.kotlin.ast.KSAstProvider.addOriginatingElement(KSAst.kt:94)
        at me.tatarka.inject.compiler.KmpComponentCreateGenerator.generate(KmpComponentCreateGenerator.kt:27)
        at me.tatarka.inject.compiler.ksp.ProcessKmpComponentCreateKt.generateKmpComponentCreateFiles(ProcessKmpComponentCreate.kt:40)
        at me.tatarka.inject.compiler.ksp.InjectProcessor.finish(InjectProcessor.kt:113)

I can reproduce this consistently in unit tests for all platforms where I define the function in commonTest

@KmpComponentCreate
expect fun createTestComponent(
  coroutineScope: CoroutineScope,
  coroutineDispatcher: CoroutineDispatcher,
): TestComponent

This happens only with KSP 2, there’s no issue with KSP 1 obviously.

[eygraber]

That was the error related to caching KSP objects I think. Odd that it would pop up again.

[ralf]

Yep, I fixed one or two in kotlin-inject myself. I'm not using this annotation a lot, so it seems to be just another case that was missed before.

[eygraber]

Are you using the snapshot that has KSP2 support?

[ralf]

No, last stable, which is quite old. The slow releases are painful.

[ralf]

I used to maintain a fork internally, but this is about an open source project.

[eygraber]

The snapshot should have a bunch of fixes for that issue

[eygraber]

A release would be nice...

[Daniel Weidensdörfer]

Will there be such a fix for the kotlin-inject without anvil?

[eygraber]

It's fixed in kotlin-inject, but a release hasn't been made with those changes yet.

[Daniel Weidensdörfer]

Alright, can you give some planned date?

[eygraber]

I can't, I'm not the maintainer 😅

[Daniel Weidensdörfer]

If you know him, let him know 🙂

[Daniel Weidensdörfer]

Is kotlin-inject-anvil a super set of kotlin-inject?

[eygraber]

@ralf can probably explain better, but I believe kotlin-inject-anvil is built on top of kotlin-inject

[ralf]

@Daniel Weidensdörfer Create an issue on Github?

I’m maintaining kotlin-inject-anvil. No, it’s not a superset, but an extension. But given the multiple problems of kotlin-inject, we may consider forking the core at some point.