with changing

launch

to

async

and

join

to

await

everything start to work as expect, I see all three catch (as described in https://github.com/Kotlin/kotlinx.coroutines/issues/61). But with

launch

its very strange that

method-catch

doesn’t call

You can’t catch an exception from coroutine with launch. But you can with async/await. Examples: https://github.com/codyoss/kotlin-coroutine-examples/blob/master/src/main/kotlin/com/github/codyoss/coroutine/example/exceptions/01_LaunchAndFail.kt#L8 https://github.com/codyoss/kotlin-coroutine-examples/blob/master/src/main/kotlin/com/github/codyoss/coroutine/example/exceptions/03_Async.kt#L14

I know that with async/await it’s possible to do it (i have mentioned it in question). The question was about

launch

, about why it’s don’t catched from not suspend method

when you launch something you have to think about it as it off on another world. So if an exception happens over on Mars you would never know about it on Earth. The await function provides a bridge to join the two worlds so they can interact. That analogy fell apart rather quickly…

launch

is equivalent of running a thread in coroutines world. When you write

val thread = thread { throw Exception() }
thread.join()

You don’t expect for

join

to throw, do you? Same about

launch

. But

async

returns

Deferred

, which is analogue of the

Future

in coroutines world. And

future.get

throws an exception if related computation failed with exception

@codyoss, @Vsevolod Tolstopyatov [JB] thanks for explanation, and especially for the analogy with the thread. Now it’s much clearer.

Does a `launch`ed child with an unhandled exception in today's implementation cancel it's parent with that exception, as it used to be the behaviour?

@uli it cancels child with that exception. Though this mechanism is slightly non-deterministic and being reworked under https://github.com/Kotlin/kotlinx.coroutines/issues/333 to provide consistent behaviour for users. As soon as it’s ready I will add guide section about exceptions mental model

Thanks @Vsevolod Tolstopyatov [JB] for the update