Cannot print log info from another thread of a pumpkin plugin

[NriotHrreion]

I'm trying to develop a pumpkin plugin. When I use the following code for my plugin, the web server is properly started on the port, but the log info just can't be printed to the console:

#[plugin_method]
pub async fn on_load(&mut self, ctx: Arc<Context>) -> Result<(), String> {
	ctx.init_log();

	thread::spawn(|| {
		let _ = tracing_subscriber::fmt::try_init();
		let rt = Runtime::new().unwrap();
		
		rt.block_on(async {
			web::init_web_server().await.unwrap();
			tracing::info!("Web server is ready on port 3000");
		});
	});

	Ok(())
}

Anyone knows how to print a log from another thread?

[Charlie284]

There are two things happening here.

First, ctx.init_log() already installs the plugin's tracing subscriber. tracing_subscriber::fmt::try_init() tries to install another global subscriber, so it will normally return Err here; the code currently discards that result. A global tracing subscriber is used as the fallback on all threads, so a second subscriber is not needed just because this is a std::thread (tracing dispatcher docs).

For a native Pumpkin plugin, the clearest supported path is to clone the context into the thread and use Context::log. That API was added specifically with the separate subscriber used by dynamically loaded plugins (the tracing migration discussion):

ctx.init_log();

let log_ctx = Arc::clone(&ctx);
thread::spawn(move || {
    let rt = Runtime::new().expect("failed to create web runtime");

    rt.block_on(async move {
        log_ctx.log("Starting web server on port 3000");

        if let Err(error) = web::init_web_server().await {
            log_ctx.log(format!("Web server stopped: {error}"));
        }
    });
});

Second, the position of the original message matters. If init_web_server() binds the listener and then awaits an accept/serve loop, that future does not complete while the server is running. Therefore this:

web::init_web_server().await?;
tracing::info!("Web server is ready");

logs only after the server exits, not after it starts. To report ready accurately, put the log inside init_web_server() immediately after the listener has bound successfully and before awaiting the serve loop. For example, with Axum the order is TcpListener::bind(...).await?, log the listener's local_addr(), then axum::serve(listener, app).await?.

So: remove the inner try_init(), pass a cloned Arc<Context> to the thread, and move the readiness message to immediately after the successful bind.