When a JavaScript Error with a .cause property is thrown inside a callee-handled ThreadsafeFunction callback, the resulting napi::Error on the Rust side always has cause: None, discarding the error chain.

Root cause

In call_js_cb in threadsafe_function.rs, when a pending JavaScript exception is detected in the WithCallback branch, Error is constructed directly with cause: None hardcoded:

get_error_message_and_stack_trace(raw_env, exception).and_then(|reason| {
    Err(Error {
        maybe_raw: error_reference,
        maybe_env: raw_env,
        cause: None,  // never extracted
        status: Status::from(raw_status),
        reason,
    })
})

The From<Unknown> impl in error.rs already knows how to extract .cause from a JavaScript error object:

let maybe_cause: Option<Box<Error>> = value
    .coerce_to_object()
    .and_then(|obj| obj.get_named_property::<Unknown>("cause"))
    .map(|cause| Box::new(cause.into()))
    .ok();

The exception path in call_js_cb bypasses this entirely.

Reproduction

JavaScript — throw an error with a cause:

const cause = new Error("root cause");
const err = new Error("wrapper");
err.cause = cause;
throw err;

Rust — cause is always None:

Err(err) => {
    println!("{:?}", err.cause); // None — cause is discarded
}

Expected

napi::Error::cause is populated from the thrown JavaScript error's .cause property, consistent with the behavior of From<Unknown>.

Actual

napi::Error::cause is always None, even when the thrown JavaScript exception has a populated .cause property.

Suggested fix

In the WithCallback branch of call_js_cb, extract .cause from the raw exception value using the same approach as From<Unknown> — read get_named_property::<Unknown>("cause") from the exception object and populate the cause field rather than hardcoding None.

Version

napi 3.8.3