Parallel invocation of AnalysisFacade fails spuriously in 1.20.0-RC1

[davidburstrom]

Expected Behavior

It should be possible to run tests and other scenarios in parallel where AnalysisFacade is invoked.

Observed Behavior

io.github.detekt.tooling.api.UnexpectedError
	at app//io.gitlab.arturbosch.detekt.core.tooling.AnalysisFacade$runAnalysis$1.invoke(AnalysisFacade.kt:55)
	at app//io.gitlab.arturbosch.detekt.core.tooling.AnalysisFacade$runAnalysis$1.invoke(AnalysisFacade.kt:47)
	at app//io.gitlab.arturbosch.detekt.core.tooling.ProcessingSpecSettingsBridgeKt.withSettings(ProcessingSpecSettingsBridge.kt:26)
	at app//io.gitlab.arturbosch.detekt.core.tooling.AnalysisFacade.runAnalysis$detekt_core(AnalysisFacade.kt:47)
	at app//io.gitlab.arturbosch.detekt.core.tooling.AnalysisFacade.run(AnalysisFacade.kt:25)
        ...
Caused by: java.lang.RuntimeException: Duplicate registration for EP 'org.jetbrains.kotlin.com.intellij.treeCopyHandler': first in FakeIdForTests, second in FakeIdForTests
	at org.jetbrains.kotlin.com.intellij.openapi.components.ComponentManager.createError(ComponentManager.java:167)
	at org.jetbrains.kotlin.com.intellij.openapi.extensions.impl.ExtensionsAreaImpl.checkThatPointNotDuplicated(ExtensionsAreaImpl.java:235)
	at org.jetbrains.kotlin.com.intellij.openapi.extensions.impl.ExtensionsAreaImpl.registerExtensionPoint(ExtensionsAreaImpl.java:240)
	at org.jetbrains.kotlin.com.intellij.openapi.extensions.impl.ExtensionsAreaImpl.doRegisterExtensionPoint(ExtensionsAreaImpl.java:210)
	at org.jetbrains.kotlin.com.intellij.openapi.extensions.impl.ExtensionsAreaImpl.doRegisterExtensionPoint(ExtensionsAreaImpl.java:189)
	at org.jetbrains.kotlin.com.intellij.openapi.extensions.impl.ExtensionsAreaImpl.doRegisterExtensionPoint(ExtensionsAreaImpl.java:183)
	at org.jetbrains.kotlin.com.intellij.openapi.extensions.impl.ExtensionsAreaImpl.registerExtensionPoint(ExtensionsAreaImpl.java:160)
	at io.github.detekt.parser.DetektPomModel.<init>(DetektPomModel.kt:27)
	at io.github.detekt.parser.KotlinEnvironmentUtilsKt.createKotlinCoreEnvironment(KotlinEnvironmentUtils.kt:56)
	at io.gitlab.arturbosch.detekt.core.settings.EnvironmentFacade$environment$2.invoke(EnvironmentAware.kt:38)
	at io.gitlab.arturbosch.detekt.core.settings.EnvironmentFacade$environment$2.invoke(EnvironmentAware.kt:31)
	at kotlin.SynchronizedLazyImpl.getValue(LazyJVM.kt:74)
	at io.gitlab.arturbosch.detekt.core.settings.EnvironmentFacade.getEnvironment(EnvironmentAware.kt:31)
	at io.gitlab.arturbosch.detekt.core.ProcessingSettings.getEnvironment(ProcessingSettings.kt)
	at io.gitlab.arturbosch.detekt.core.KtTreeCompiler.<init>(KtTreeCompiler.kt:14)
	at io.gitlab.arturbosch.detekt.core.tooling.ParsingStrategyKt$inputPathsToKtFiles$1.invoke(ParsingStrategy.kt:24)
	at io.gitlab.arturbosch.detekt.core.tooling.ParsingStrategyKt$inputPathsToKtFiles$1.invoke(ParsingStrategy.kt:23)
	at io.gitlab.arturbosch.detekt.core.tooling.Lifecycle$analyze$filesToAnalyze$1.invoke(Lifecycle.kt:38)
	at io.gitlab.arturbosch.detekt.core.tooling.Lifecycle$analyze$filesToAnalyze$1.invoke(Lifecycle.kt:38)
	at io.gitlab.arturbosch.detekt.core.util.PerformanceMonitor.measure(PerformanceMonitor.kt:42)
	at io.gitlab.arturbosch.detekt.core.tooling.Lifecycle$DefaultImpls.measure(Lifecycle.kt:34)
	at io.gitlab.arturbosch.detekt.core.tooling.Lifecycle$DefaultImpls.analyze(Lifecycle.kt:38)
	at io.gitlab.arturbosch.detekt.core.tooling.DefaultLifecycle.analyze(Lifecycle.kt:61)
	at io.gitlab.arturbosch.detekt.core.tooling.AnalysisFacade$runAnalysis$1.invoke(AnalysisFacade.kt:48)
	... 47 more

This is quite similar to #2433

Steps to Reproduce

Reproduction could be provided if requested.

Context

I'm trying to execute parallel JUnit tests where AnalysisFacade is being invoked, and it hints that parallel production execution also would fail similarly

Your Environment

• Version of Detekt used: 1.20.0-RC1
• Version of Gradle used (if applicable): 7.4
• Gradle scan link (add --scan option when running the gradle task): N/A
• Operating System and version: macOS
• Link to your project (if it's a public repository): N/A

[BraisGabin]

This is a regression, right?

[davidburstrom]

Certainly is :)

[cortinico]

I think this might be the root cause:

• Fix AutoCorrection crashing with Missing extension point #4545

Reverting that will cause the autoCorrect feature to don't work at all for anyone. I was not able to find a better solution sadly.

[davidburstrom]

Why wouldn't you be able to reintroduce the synchronized? To me it looks like that would solve this issue and I don't see why that would cause secondary issues.

[cortinico]

Why wouldn't you be able to reintroduce the synchronized?

That could work. My gut feeling was that the source of all the pain was a call to getRootArea(). I frankly don't know what's the root cause of the stacktrace you reported. Wrapping with synchronized could work (it was originally removed as we mirrored our setup to the KtLint one).

[davidburstrom]

I think in my case, the problem stems from the fact that multiple threads are running DetektPomModel.<init> in parallel.

[BraisGabin]

@davidburstrom do you have time to create a PR with your proposal to fix this?

[davidburstrom]

Yeah, I can take a stab at it, especially since I can repro it locally. How much time do I have before the GA release?

[BraisGabin]

I don't think we can launch the stable release until this is fixed. So there's not an eta.

[davidburstrom]

As far as I can tell, the bug is fixed in 1.20.0-RC2.