Add initial test for multiplex worker corruption

[restingbull]

[Bencodes]

@restingbull I ran some tests in our codebase with the worker java implementation and wasn't able to reproduce the issue. I was however able to reproduce it almost instantly switching back to the coroutines implementation.

[rockwotj]

There are a bunch of thread local variables in the compiler using the IR backend: https://github.com/JetBrains/kotlin/blob/54cca882797de3d7861fcddae0214ad802937669/compiler/ir/ir.tree/src/org/jetbrains/kotlin/ir/util/SymbolTable.kt#L360-L374

I'm not sure if something like that could be the source of this error, but it seems like it if switching to a stable thread caller vs coroutine that may switch threads - thread local seems like a likely cause of these issues. Maybe if you use a coroutine dispatch that always switches threads at yield points may be a good way to force and see if that is the issue? (although I'm not sure how to do that off the top of my head)

[rockwotj]

Looks like you could do that with a dispatcher that always returns true for isDispatchNeeded and then have one that for dispatch creates a new thread for every task...

Docs: https://kotlin.github.io/kotlinx.coroutines/kotlinx-coroutines-core/kotlinx.coroutines/-coroutine-dispatcher/is-dispatch-needed.html https://kotlin.github.io/kotlinx.coroutines/kotlinx-coroutines-core/kotlinx.coroutines/-coroutine-dispatcher/dispatch.html

Alternatively I wonder if you ran on a dispatcher with only a single thread if you wouldn't see corruption...

EDIT:

That doesn't make a ton of sense because the compiler is called synchronous without a coroutine context, It could also be that the default dispatchers are being reused in the kotlin compiler, leading to state being (reused) potentially when it shouldn't?