On upgrading to Kotlin 1.5, I’m getting

e: java.lang.IllegalStateException: not identifier: <init>

errors while compiling (example). Is this a known issue?

Hey everyone. 👋 I joined here in the hope to reach some of the Kotlin compiler experts. I'm following up on my post here. Long story short: is there a way I can analyze why the Kotlin compiler takes so long to compile the same codebase which compiles in part of the the time on other compilers like C#? It seems quite long that my codebase compiles in 30s for TypeScript (👍), 1:15min for C# (👍) but 4:48min for Kotlin (👎). I just upgraded to Kotlin 1.5 in the hope to get a speed up but it's basically the same speed as with 1.4.x. Example compilation here.

is there any expected performance penalty for enabling

-Xextended-compiler-checks

?

Hello there! I've heard that the plan is to eventually disallow suppressing certain compiler errors (like "INVISIBLE_MEMBER" and "INVISIBLE_REFERENCE" for example) which is absolutely understandable, but would compiler plugins still be able to suppress certain compiler errors one way or another? Currently, I'm generating some sources in an AnalysisHandler (which I know will eventually be deprecated don't worry but I'm just using it for now and the logic should be the same for the future) based on existing sources and shadowing them using

@Suppress("REDECLARATION")

but I'm wondering if maybe there's a better way to do that (like a custom call resolved I guess) and if eventually this trick will be completely disallowed

Different question. Are there any current plans to eventually use JVM IR for inlining? or just any plans to include serialised JVM IR in published jars? Because that can be immensely useful for compiler plugins that need to use IR from external sources. For example, I'm currently performing inlining manually on the IR level, and so on the JVM I have to do a few hacks with finding the sources jar for a file and compiling that source to use its IR, but if I had access to serialised IR, I can just deserialise that and inline it directly.

Hey all 🙂 I’m new to Kotlin compiler plugins and hope this is the right place to ask. I’m playing with the new IR compiler plugins. I created a simple test, that adds a function to an annotated class. The function just calls

println("I'm here");

I can compile the class and when I try to use

kotlinc

to compile another class that calls the generated function, I get:

kotlinc -classpath . test.kt
test.kt:2:10: error: unresolved reference: generatedFunction
  Main().generatedFunction()

But when I do the same with

javac

, everything is ok. I can also run the class and I get the desired output

java Test
I'm here

Any pointers would be highly appreciated 🙂 p.s. Is there any documentation about writing compiler plugins with IR out there?

Hi! Why is type inference failing here? To my eyes it has all the info it needs to infer the type...

class Builder {
  fun <R> build(action: (R) -> R): Builder = this
}

fun main() {
  val value = 3
  Builder()
    .build { value } // compiler error: not enough information to infer type variable R
}

That is: lambda is returning

Int

therefore

R

is

Int

UPD: this is kotlin 1.5.0

Is there a way, via gradle configuration or a compiler plugin, to get the compiler to generate a klib for JVM? I want to use the IR after compilation (creating bindings for a compiler plugin to use), and while I could use a plugin to write those bindings to an external file, if I can just use the klib I won't have to worry about incremental compilation.

apologies if this is the wrong place to ask. I've searched google and YouTrack and found nothing remotely related. Am upgrading a JVM/JS project to 1.5.0, and getting the below stack trace when the JS is compiling in one of my modules. It's tough to even know where to start, because it doesn't give a hint about the offending file, and the compiler debug breakpoints aren't hitting. Any debug hints appreciated

Hey, I debugged KT-29088 a bit (trying to learn how the compiler works) and I think the root cause is that CompareTo only implements comparison of

Char

,

Byte

,

Short

,

Int

,

Long

,

Float

and

Double

. And for all other types, it will fallback to Int. However

CompareTo

is registered for all possible combinations of primitives, hence also for Boolean::compareTo(Boolean). I'm unsure if there is a smart trick/optimization behind using

kotlin.jvm.internal.Intrinsics::compare(Int, Int)

also for

Boolean::compareTo(Boolean)

?

Hi - just read this article about the need to add

@JvmSynthetic

to

internal

members to properly hide them from Java: https://medium.com/scalereal/hide-internal-members-of-kotlin-module-from-jvm-c7730507fb17 which informed me of

@JvmSynthetic

https://kotlinlang.org/api/latest/jvm/stdlib/kotlin.jvm/-jvm-synthetic/ which in turn led me to

ACC_SYNTHETIC

https://docs.oracle.com/javase/specs/jvms/se15/html/jvms-4.html Is there a reason why the compiler with the JVM backend doesn’t just add the

ACC_SYNTHETIC

access_flag to

internal

methods without needing to add the noise of an extra annotation?

I saw in the Kotlin roadmap that the compiler frontend is being reworked. Does this imply that the API is going to have significant breaking changes?

Has anyone considered a compiler plugin for mocking?

Hey guys, I've got a fairly basic math language I've expressed with antlr, it works and runs, but now I want to do some transcoding into a lisp-like language (specifically, my custom math language that supports things like

x*y^z + cos(f)

I want to translated into smt-lib2). I've read the compilers book in school and have used ANTLR professionally for some time, and I can probably hack this out with an antlr visitor and creating classes as the occur to me along with a good whack of string mangling, but I'm willing to bet somebody smarter than me has some nice compiler/code-generator tutorials/tools that would make my life a fair bit easier. Has anybody here done something similar and have some suggested reading?

I'm having trouble (

NoClassDefFoundError

errors) trying to use a library in a native compiler plugin (i.e. using

kotlin-compiler

instead of

kotlin-compiler-embeddable

. Said library also uses

kotlin-compiler

, and I'm depending on it using

implementation

, which works fine for the non-native compiler plugin, but when I try to use something from the library in the native plugin I get

NoClassDefFoundError

. Do I need to shade it in myself?

and, yes, i'm probably mucking in things I shouldn't and I'm currently reworking it all on top of the antlr4 grammars so I don't need to rely on compiler internals but I was hoping that with the plugins that the compiler offers there'd be a nicer way to do what I'm trying to do.

Does using

internal

modifier cause some optimized compilation in a way that compiler knows it isn't used from other dependent modules and skips the check "in them"?

Hello everyone!, I was looking at

org.jetbrains.kotlin.fir.extensions.FirExtensionRegistrar

and looking into how to integrate with plugins that depend on

org.jetbrains.kotlin.compiler.plugin.ComponentRegistrar

and

org.jetbrains.kotlin.compiler.plugin.CommandLineProcessor

. I see

AllOpen

and others are connected to the new FIR frontend and I was hoping to run some compiler plugin tests and work on the FIR support for arrow meta. I have a couple of questions that came up while I was working on this. Here it goes in case anyone can help. 1. The

FirExtensionRegistrar

unlike the

ComponentRegistrar

is not an interface but an abstract class and it requires a separate lifecycle. Once instantiated for regular CLI plugins and another for the FIR style of subscription. Did I miss-understood this or is there a way to integrate this with existing compiler plugins based on the

ComponentRegistrar

? 2. Even when declaring both of those extensions and enabling FIR with

configuration.put(CommonConfigurationKeys.USE_FIR, true)

the override for the FIR plugin`override fun ExtensionRegistrarContext.configurePlugin(): Unit` never gets called, Is there something else needed to enable FIR beside that flag? If anyone has had success testing the FIR extension points or could provide any guidance or example as to how to enable them I’d very much appreciate it. Thanks!

Hey folks. I know that there are people how are interested in new FIR compiler and want to discover how it works, so I want to share with you some information about FIR itself and kotlin project environment setup so you can inspect, debug and even modify FIR locally for your experiments. Actually, this is only way to interact with FIR right now, since compiler API for FIR is postponed (see thread above) Details in :thread-please:

fun interface Foo : () -> Unit

fun interface Foo1 : Foo
fun interface Foo2 : Foo

fun test(foo: Foo2) {
    println(foo)
}

fun main() {
    val instance = object : Foo1 {
        override fun invoke() {
        }
    }
    test(instance) // is this allowed by design?
}

Maybe I'm not smart enough to understand why this is allowed, it makes me feel confused, is there any particular reason to support this behavior?

Expected an compiler error, but it didn't came 😉 Is this expected behavior?

class Dummy {

    // Define as read-only access map
    private val buffers: Map<String, String>

    init {
        buffers = mutableMapOf()

        // Unexpected, no error
        buffers["test"] = "hello"
    }

    fun dummy2() {

        // Not allowed, as expected
       buffers["test"] = "hello"
    }

}

I hope this is the correct place to ask for Gradle issues. I am using the

kotlin("js")

plugin, 1.5.10, Kotlin DSL, and I want to change the default timeout.

js {
	browser {
		testTask {
			useKarma {
				useChromiumHeadless()
				timeout.set(??)
			}
		}
	}
}

According to the Gradle documentation:

task myTask {
	// java.time.Duration 
	timeout = Duration.ofMinutes(10)
}

However if I add write:

timeout.set(java.time.Duration.ofMinutes(10))

Then IntelliJ & Gradle complain with

Unresolved reference: time

timeout

doesn't have any overloads that would accept

kotlin.Duration

or even

groovy.Duration

. Is it even possible currently to set the timeout with the Kotlin DSL?

[Source, IR BE]: Is there any reason why some of the Impls of the IR elements are in ir.tree module and others in ir.tree.impl? The latter seem to only differ in that they contain and override IrFactory (also dunno why). In FIR, all Impls seem to lay in the tree module (but it is auto-gen, I know).

Hi 👋 what use case does this error report on? Seems to be related to annotated inline lambdas with annotations using non source retention (any annotations that survive until the output binaries), maybe related to closure capturing but not 100% sure.

Hi, is there a way to read the full method signature from a KtNamedFunction? Was wondering if there is a utils method or something like that. Thanks!

Hi all 🙂 We’re trying to upgrade to Kotlin 1.5, but are a bit stuck. We have a number of places that create a

private typealias

, and then use that alias in a public method. This seems to work in Gradle, but not in our Bazel setup. My main question is, is this valid code and a bug in our Bazel setup, or should we not expose a private typealias in a public api?

Hey guys. I’m running into a particular issue after updating the Kotlin plugin in AS to 1.5.20. The surrounding code is probably irrelevant, it’s the same error appearing in multiple occurrences of such code. I got a data class that looks like this:

data class MapObjects(
    val map: com.google.android.gms.maps.GoogleMap,
    val markerManager: com.google.maps.android.collections.MarkerManager,
    val markerCollection: com.google.maps.android.collections.MarkerManager.Collection,
)

The

MarkerCollection.Collection

is a Java nested/inner class, not a static one. After the update to Kotlin 1.5.20 (keeping runtime dependencies on 1.5.0), it compiles, but IDE complains about code like this:

val mapObjs: MapObjects = mapObjects // lateinit var access
mapObjs.markerCollection.clear()

Basically anything accessing the

markerCollection

shows

Internal Error occurred while analyzing this expression

. I reported it from within IDE already. Funny detail though: when I update runtime dependency to 1.5.20, it outright crashes while compiling:

org.jetbrains.kotlin.backend.common.BackendException: Backend Internal error: Exception during file facade code generation

. Based on that I do not believe it’s related to Android Studio in particular. Please send help, and if I should, include where/how do I file this in the Kotlin’s issue tracker at https://youtrack.jetbrains.com/issues/KT.

it seems like inline classes don't always get boxed when used as generics (contrary to what docs say). Is this a bug or expected behavior?

inline class InlinePayload(val id: String)

fun <T> test(t: T): T {
  return t
}

val (p1, p2) = listOf(test(3))
  .zip(
    listOf(test(InlinePayload("a"))), // (1)
    ::Pair)
  .first()

println(p2.id) // (2) EXCEPTION!

It seems that

(1)

creates a

List<String>

rather than

List<InlinePayload>

and then

(2)

crashes with

class java.lang.String cannot be cast to class ru.example.app.InlinePayload (java.lang.String is in module java.base of loader 'bootstrap'; ru.example.app.InlinePayload is in unnamed module of loader 'app')

Is anyone else having issues attaching sources for kotlin compiler? I understand kotlin-compiler-embeddable doesn't contain sources, but even when I attach kotlin-compiler-sources.jar I dont get good results. Example: while debugging if I Go to Type Source for a

DeserializedTypeAliasDescriptor

, it brings me to the decompiled class file. Meanwhile a symbol search gives the correct Kotlin source file

DeserializedMemberDescriptor.kt

. Any ideas? I have

compileOnly(kotlin("compiler-embeddable"))

in my gradle file, and

relocate("com.intellij", "<http://org.jetbrains.kotlin.com|org.jetbrains.kotlin.com>.intellij")

when creating a JAR, but I dont believe that is relevant here as the package name is the same in both source and class file.