D-Bus: Get peer-to-peer connection via unix sockets to work

[sophie-h]

I have the following example code. It spawns a D-Bus server and client in the same binary. At the end there is a function call from the client. Instead of triggering the debug print in the server, the function call is timing out.

I have no clue what I'm doing wrong. I also tried to create a proxy, but that's also timing out.

I also tried the new server code from 17db771 but that also didn't made a difference. The function call still timed out. I also wrote a complete _sync based version of everything. The function call still timed out.

Help would be much appreciated.

use gio::{
    glib::{self, ExitCode, VariantTy},
    prelude::SocketExt,
    DBusConnectionFlags, DBusNodeInfo, Socket,
};
use std::os::{fd::IntoRawFd, unix::net::UnixStream};

use gio::prelude::*;

const EXAMPLE_XML: &str = r#"
  <node>
    <interface name='org.gnome.glycin'>
      <method name='Hello'>
        <arg type='s' name='name' direction='in'/>
        <arg type='s' name='greet' direction='out'/>
      </method>
    </interface>
  </node>
"#;

pub fn main() -> ExitCode {
    let app = gio::Application::builder()
        .application_id("com.example.app")
        .build();
    let _guard = app.hold();

    app.connect_startup(move |_app| {
        let (x, y) = std::os::unix::net::UnixStream::pair().unwrap();

        x.set_nonblocking(true).unwrap();
        y.set_nonblocking(true).unwrap();

        glib::spawn_future_local(spawn_server(x));
        glib::spawn_future_local(spawn_client(y));
    });

    app.connect_activate(move |_| {
        println!("Waiting for DBus Hello method to be called.");
    });

    app.run()
}

pub async fn spawn_server(fd: UnixStream) {
    let socket = unsafe { Socket::from_fd(fd.into_raw_fd()) }.unwrap();
    let socket_connection = socket.connection_factory_create_connection();

    let guid = gio::dbus_generate_guid();

    dbg!("server connecting");

    let connection = gio::DBusConnection::new_future(
        &socket_connection,
        Some(&guid),
        DBusConnectionFlags::AUTHENTICATION_SERVER
            .union(DBusConnectionFlags::DELAY_MESSAGE_PROCESSING),
        None,
    )
    .await
    .unwrap();

    dbg!("server connected");

    let interface_info = DBusNodeInfo::for_xml(EXAMPLE_XML)
        .unwrap()
        .lookup_interface("org.gnome.glycin")
        .unwrap();

    let _id = connection
        .register_object("/org/gnome/glycin", &interface_info)
        .method_call(|_connection, y, z, a, b, c, d| {
            dbg!(y, z, a, b, c, d);
        })
        .build()
        .unwrap();

    dbg!("server awaiting calls");

    std::mem::forget(connection);
}

pub async fn spawn_client(fd: UnixStream) {
    let socket_client = unsafe { Socket::from_fd(fd.into_raw_fd()) }.unwrap();
    let socket_connection_client = socket_client.connection_factory_create_connection();

    dbg!("client connecting");

    let connection = gio::DBusConnection::new_future(
        &socket_connection_client,
        None,
        DBusConnectionFlags::AUTHENTICATION_CLIENT
            .union(DBusConnectionFlags::DELAY_MESSAGE_PROCESSING),
        None,
    )
    .await
    .unwrap();

    dbg!("client connected");

    let result = connection
        .call_future(
            None,
            "/org/gnome/glycin",
            "org.gnome.glycin",
            "Hello",
            Some(&("x",).into()),
            Some(VariantTy::STRING),
            gio::DBusCallFlags::NONE,
            2000,
        )
        .await
        .unwrap();

    dbg!(result);
}

[sdroege]

There are a couple of bugs there. In your code and in the bindings too apparently. I'll get you a runnable version in a bit.

[sdroege]

Part of it is https://gitlab.gnome.org/GNOME/glib/-/merge_requests/4414 btw. Other PRs coming soon, now just fixing some memory leaks...

[sdroege]

https://gitlab.gnome.org/GNOME/glib/-/merge_requests/4415 is the memory leak. I think that one is problematic for you because you'd leak every single method call with GLib versions without that fix.

[sdroege]

See #1584 for the bindings part of the whole story. I've added a test that is basically your code in correct

[sdroege]

Also CC @swsnr . The memory leak in GLib might also be relevant for you.

[swsnr]

Thanks, but I think I'm fine 🙏. I'm not using Gdbus anywhere where leaking a few bytes per call causes trouble 😅