A modern programming language that makes developers happier.

kotlinlang

Is this a bug, or an expected limitation?
```suspend fun foo(): String = "foo"

abstract class Bar(val foo: String)

suspend fun main() {
    val bar = object: Bar(foo()) {} // ❌ Suspension functions can be called only within coroutine body
}```
Obviously it works fine if I instead write
```suspend fun main() {
    val foo = foo()
    val scope = object: Bar(foo) {}
}```

maybe the constructor function is not suspend and it is not inline too

you can create an inline invoke function in the companion to workaround this (I haven’t tried personally)

Isn't the call to the super constructor (`Bar(foo())`) actually compiled into the constructor of the anonymous class in this case? The constructor of the anonymous class is obviously not a `suspend` function, and as such would not be able to call other `suspend` functions.

E.g. the class of the anonymous class you're creating could also be expressed like this:
```class AnonymousClass : Bar(foo()) {
}

suspend fun main() {
    val bar = AnonymousClass()
}```
Which wouldn't compile because the constructor of `AnonymousClass` is not and cannot be a `suspend` function.

Interesting, I hadn't considered that there's actually a constructor generated for the anonymous class! That makes sense, though the result is delightfully unintuitive

It's definitely ambiguous in this sense - I give you that :smile:

Okay, so now I understand why it doesn’t work. Thanks for the insight! I still can’t decide if it’s a bug or not, though. I feel like there’s nothing in the rules of the language that say this _shouldn’t_ work. It’s more of an implementation detail that requires knowing how the compiler will transform this code. :thinking_face:

It's an interesting example of how capturing works for anonymous classes.

Instead of the result of `foo()` (temp variable) being captured, the call to `foo()` itself s captured.

I agree with Javier and I'd report this, since it is, as you said, very unintuitive.

When I went to create an issue I found that it already exists :+1: <https://youtrack.jetbrains.com/issue/KT-22984>