Unable to write bytes to IOStream using futures about gio HOT 4 CLOSED

Can you provide a small testcase that can be run to reproduce this?

from gio.

Sure thing - I've created a new repo for it, which can be found here. It works very similarly to what I posted above; it'll connect to one of GitHub's servers, try to send "test" as an array of bytes to it, and receive an error when trying to perform that write operation.

from gio.

diff --git a/src/main.rs b/src/main.rs
index 6e988d9..44454cc 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -41,6 +41,7 @@ fn main() {
                         })
                         .and_then(|(_strm, (_buf, count, _err))| {
                             dbg!(count);
+                            dbg!(stream);
                             Ok(())
                         })
                 })

This works around the problem. The problem is that stream, the gio::SocketConnection, is getting out of scope after your connect_to_host_async_future() continuation and is then destroyed. While you still have a reference to the gio::OutputStream, not having any reference to the connection anymore is causing it to be closed.

Quite unintuitive behaviour in gio, nothing to do with how the bindings work. You can easily reproduce it also with the following synchronous code:

use gio::prelude::*;

fn main() {
    let client = gio::SocketClient::new();
    client.set_timeout(10);

    let req = String::from("test");

    let conn = client.connect_to_host("192.30.253.112", 80, None::<&gio::Cancellable>).unwrap();

    println!("Is connected: {}", &conn.is_connected());
    println!("Stream is closed: {}", &conn.is_closed());
    let writer = conn.get_output_stream().expect("Failed to obtain output stream");

    // comment this out to make it work
    drop(conn);
    let res = writer.write_all(req.as_bytes(), None::<&gio::Cancellable>);
    println!("write finished: {:?}", res);
}

PS: Don't run a main context/main loop inside your startup signal handler. The gio::Application is already doing that for you and startup is called from an existing main loop.

from gio.

Reported to GLib: https://gitlab.gnome.org/GNOME/glib/issues/1690

Let's close this on our side then.

from gio.