Channels
#eap
- n
nhaarman
12/21/2016, 9:31 PMIt seems Gradle can't find the
when using the https version oforg.jetbrains.kotlin:native-platform:1.1-M04
. Is that right?maven { url '<http://dl.bintray.com/kotlin/kotlin-eap-1.1>' } - a
alex2069
12/22/2016, 12:49 AMWe had same problem. We discovered that the gradle plugin lib has wrong (old?) pom which includes native-platform as a dependency still http://gradle.artifactoryonline.com/gradle/libs/org/jetbrains/kotlin/kotlin-compiler-runner/1.1-M04/kotlin-compiler-runner-1.1-M04.pom - a
alex2069
12/22/2016, 3:58 AM@dmitry.petrov Any particular reason typealias is now only permitted in top-level? https://github.com/JetBrains/kotlin/commit/4c47d77 We were using it in interfaces and classes quite extensively for simplifying listener and callback declarations; (contrived example)val onClick: ClickListener? = null - d
dmitry.petrov
12/22/2016, 7:02 AMYes, that's a kind of disappointment. It was a very good use case for type aliases, and we really wanted to this in 1.1. But: Problem with type aliases declared in classes and interfaces is that they can capture type parameters from the surrounding scopes without capturing the receiver. Now, we can't introduce type alias constructors for such type aliases, because we don't have such entities in language (they correspond to static members in C# and "species static" in Java 10/Valhalla). Introducing them right now requires some more design effort (mostly related to value types, specialized generics, and "species static" mentioned above), which is still in progress (and, sadly, will not be ready for 1.1). We can't "just allow non-top-level aliases now with limited capabilities, and extend that functionality later", because, sadly, that would be a breaking change in call resolution. So, the only option we are left with now is to allow top-level aliases only, and introduce non-top-level aliases later as specialized generics and "species static" are done. - d
dmitry.petrov
12/22/2016, 7:03 AMTL;DR: design for non-top-level aliases depends on specialized generics, which are a much bigger feature, and are not there yet. - a
alex2069
12/22/2016, 7:13 AMAh that's unfortunate, though good to know they're still on the road map. - d
dmitry.petrov
12/22/2016, 7:13 AMYep, same feelings here. - k
kirillrakhman
12/22/2016, 8:01 AM@dmitry.petrov does this mean nested type aliases will only land when java 10 lands? - d
dmitry.petrov
12/22/2016, 8:04 AMHard to tell. I'd say it's actually more related to Kotlin Native, but @abreslav might have a different view. - k
kirillrakhman
12/22/2016, 8:15 AMhttps://youtrack.jetbrains.com/issue/KT-14967 is still happening in M04 - k
kirillrakhman
12/22/2016, 9:04 AMalso, I can't build my app with kapt3 enabled on my CI server (it works locally) 😐 - i
ilya.chernikov
12/22/2016, 10:38 AM@alex2069 @nhaarman it seems that gradle’s artifactory cached the wrong version of 1.1-M04. We’re tying to resolve the issue, but meanwhile, please use the bintray repo stated in the blog post - http://dl.bintray.com/kotlin/kotlin-eap-1.1👍 1 - y
yan
12/22/2016, 11:03 AM@kirillrakhman What’s the error message? - k
kirillrakhman
12/22/2016, 11:04 AM - k
kirillrakhman
12/22/2016, 11:05 AMLots of Unresolved reference: R - y
yan
12/22/2016, 12:03 PM@kirillrakhman Can you share/send the full build log? - v
vmironov
12/22/2016, 2:04 PMThe following code doesn’t compile with
anymore (used to compile with1.1-M04
):1.0.5Copy code
Is it an expected breaking change?interface Logger { fun log(message: String) } interface WithLogger { val logger: Logger } class WithLoggerMixin(val name: String) : WithLogger { override val logger: Logger = TODO() } abstract class AbstractController : WithLogger by WithLoggerMixin("Hello") class ControllerImpl : AbstractController(), WithLogger by WithLoggerMixin(”Override”) - y
yole
12/22/2016, 2:06 PMyes - j
jw
12/22/2016, 8:51 PMbound references seem to generate unstable class names - j
jw
12/22/2016, 8:52 PMCopy code@@ -20369,2 +20379,2 @@ -com.example.Thing$sam$Func4$239fba2e <init>(Function4) -com.example.Thing$sam$Func4$239fba2e call(Object, Object, Object, Object) +com.example.Thing$sam$Func4$8596a107 <init>(Function4) +com.example.Thing$sam$Func4$8596a107 call(Object, Object, Object, Object) - j
jw
12/22/2016, 8:54 PMis this expected? I would hope they'd be numbered like anonymous classes or lambdas are otherwise - j
jw
12/22/2016, 8:54 PMthe code in this file isn't changing in any way either - k
kirillrakhman
12/23/2016, 9:30 AMCopy code
is this right?abstract class Foo protected constructor() // warning "Protected function call from public-API inline function" inline fun foo(f: () -> Unit) = object: Foo() {} - u
udalov
12/23/2016, 10:48 AM@jw it’s more related to SAM conversions (see
in the class name) and has been this way since Kotlin 1.0, e.g.:…$sam$…Copy code
I think hashes here were supposed to be a quick dirty fix of a name clash problem, which can now (I hope) be solved in a more straightforward way. Please report an issue!fun main(args: Array<String>) { val f = { } println(Runnable(f).javaClass) // “class TestKt$sam$Runnable$cde914b6” } - e
elizarov
12/23/2016, 11:13 AM@kirillrakhman: yes. Put @PublishedApi annotation on your constructor. This is to remind you that you have exposed it via public inline function and if you remove or change it in the future you'll break some previously compiled code that had inlined your public function. - k
kirillrakhman
12/23/2016, 11:14 AM@elizarov it's not actually my class, it's gson'sTypeToken - k
kirillrakhman
12/23/2016, 11:14 AMI have reported the same issue in https://youtrack.jetbrains.com/issue/KT-15410 - e
elizarov
12/23/2016, 11:15 AMIf it's not your class that how are you using its protected constructor? - k
kirillrakhman
12/23/2016, 11:16 AMas I see it, protected constructors on abstract classes are basicly the same as public ones - k
kirillrakhman
12/23/2016, 11:16 AMyou can only call them from a subclass, duh