`@JvmOverloads` leads to binary-incompatible advice on consumer module

[SimonMarquis]

Plugin version
latest (3.16.1)

Gradle version
latest (9.6.1)

JDK version
25

(Optional) Kotlin and Kotlin Gradle Plugin (KGP) version
KGP 2.3.21

(Optional) Android Gradle Plugin (AGP) version
9.2.1

(Optional) reason output for bugs relating to incorrect advice

./gradlew :features:VehicleAgreement:impl:reason --id :libraries:StandardLibraryExtensions

> Task :features:VehicleAgreement:impl:reason

------------------------------------------------------------
You asked about the dependency ':libraries:StandardLibraryExtensions'.
You have been advised to remove this dependency from 'testImplementation'.
------------------------------------------------------------

There is no path from :features:VehicleAgreement:impl to :libraries:StandardLibraryExtensions for releaseCompileClasspath


Shortest path from :features:VehicleAgreement:impl to :libraries:StandardLibraryExtensions for releaseRuntimeClasspath:
:features:VehicleAgreement:impl
\--- :libraries:VehicleUiComponent
      \--- :libraries:StandardLibraryExtensions

Shortest path from :features:VehicleAgreement:impl to :libraries:StandardLibraryExtensions for releaseUnitTestCompileClasspath:
:features:VehicleAgreement:impl
\--- :libraries:StandardLibraryExtensions

Shortest path from :features:VehicleAgreement:impl to :libraries:StandardLibraryExtensions for releaseUnitTestRuntimeClasspath:
:features:VehicleAgreement:impl
\--- :libraries:StandardLibraryExtensions

Source: release, main
---------------------
(no usages)

Source: release, test
---------------------
* Is binary-incompatible, and should be removed from the classpath:
  Expected METHOD fr/leboncoin/libraries/standardlibraryextensions/RandomKt.nextNumericalString(Lkotlin/random/Random;Ljava/lang/Integer;)Ljava/lang/String;, but was fr/leboncoin/libraries/standardlibraryextensions/RandomKt.nextNumericalString(Lkotlin/random/Random;)Ljava/lang/String;

Describe the bug

Random.nextNumericalString() extension is defined like this

@JvmOverloads
fun Random.nextNumericalString(size: Int? = null): String = ...

Without the @JvmOverloads no advice is generated, which is what we would expect.
Note: this annotation was added when we used to have a mix of java and kotlin code.

To Reproduce

I'm trying to reproduce the bug in a test project, but it does not seem to trigger.
SimonMarquis#2

I'm probably forgetting something 🤔

[autonomousapps]

Thanks for the issue. If you can get the repro to work, let me know.

[SimonMarquis]

Are there specific intermediate outputs I should look for to debug this further? (Or attach a debugger somewhere?)

I can't share the project setup unfortunately.

[autonomousapps]

ComputeUsagesTask has a function:

  private fun checkBinaryCompatibility(
    coordinates: Coordinates,
    binaryClassCapability: BinaryClassCapability,
    context: GraphViewVisitor.Context,
  ) {
    val result = BinaryCompatibilityChecker(coordinates, binaryClassCapability, context).result ?: return

    if (!result.isBinaryCompatible) {
      reportBuilder[coordinates, Kind.DEPENDENCY] = Reason.BinaryIncompatible(
        result.relevantMemberAccesses, result.nonMatchingBinaryClasses
      )
    }
  }

BinaryCompatibilityChecker is very complex, but that's where I'd start.

I would note that it includes this guard:

private fun compute(): Result? {
  // Can't be incompatible if the code compiles in the context of no duplication
  if (context.duplicateClasses.isEmpty()) return null

So even in the presence of @JvmOverloads, DAGP shouldn't be telling you that you have binary-incompatible usages unless you also have duplicate classes on the classpath. Do you see duplicate class warnings? Are you suppressing duplicate class warnings?