Sync method + tokio::spawn fails to locate runtime

[pronebird]

Hi,

I initialize runtime by calling a rust method that inits runtime and puts it into static OnceLock:

static TOKIO_RUNTIME: OnceLock<Runtime> = OnceLock::new();

/// Initialize tokio runtime once before interacting with nym-vpn-lib.
/// Repeat calls do nothing.
#[allow(non_snake_case)]
#[uniffi::export]
pub fn initializeTokioRuntime() {
    let _rt = TOKIO_RUNTIME.get_or_init(|| {
        tokio::runtime::Builder::new_multi_thread()
            .worker_threads(10)
            .enable_all()
            .build()
            .expect("failed to initialize tokio runtime")
    });
}

However some of the calls result in:

rustPanic("there is no reactor running, must be called from the context of a Tokio 1.x runtime")

This impacts our application. After migrating to uniffi objects and async_runtime, nothing really works anymore. How can I share the same runtime across all threads in Swift?

It seems that the issue manifests itself particularly when using Swift actors which use cooperative queue.

[pronebird]

So the bug is:

#[uniffi::export(async_runtime = "tokio")]
impl MyObject {
    // Keep this method async because spawn needs runtime!
    #[uniffi::constructor]
    pub fn new() -> Self {
        let handle = tokio::spawn(async move { .. }); // BOOM: no tokio runtime!
       Self { handle }
   }
}

Adding "async" to new() makes this issue go away. Maybe all methods marked with async_runtime should call tokio_runtime::enter() regardless whether async or not?