timeout when player never connects

This commit is contained in:
Ilion Beyst 2022-06-07 19:12:49 +02:00
parent 69421d7b25
commit ff061f2a7a

View file

@ -48,7 +48,7 @@ impl PlayerRouter {
routing_table.insert(player_id, entry);
}
fn get(&self, player_id: &str) -> Option<SyncThingData> {
fn take(&self, player_id: &str) -> Option<SyncThingData> {
// TODO: this design does not allow for reconnects. Is this desired?
let mut routing_table = self.routing_table.lock().unwrap();
routing_table.remove(player_id)
@ -75,7 +75,7 @@ impl pb::bot_api_service_server::BotApiService for BotApiServer {
let sync_data = self
.router
.get(player_id_str)
.take(player_id_str)
.ok_or_else(|| Status::not_found("player_id not found"))?;
let stream = req.into_inner();
@ -106,21 +106,35 @@ impl runner::BotSpec for RemoteBotSpec {
) -> Box<dyn PlayerHandle> {
let (tx, rx) = oneshot::channel();
let (server_msg_snd, server_msg_recv) = mpsc::unbounded_channel();
let player_key = "test_player".to_string();
self.router.put(
"test_player".to_string(),
player_key.clone(),
SyncThingData {
tx,
server_messages: server_msg_recv,
},
);
let client_messages = rx.await.unwrap();
let fut = tokio::time::timeout(Duration::from_secs(10), rx);
match fut.await {
Ok(Ok(client_messages)) => {
// let client_messages = rx.await.unwrap();
tokio::spawn(handle_bot_messages(
player_id,
event_bus.clone(),
client_messages,
));
}
_ => {
// ensure router cleanup
self.router.take(&player_key);
}
};
// If the player did not connect, the receiving half of `sender`
// will be dropped here, resulting in a time-out for every turn.
// This is fine for now, but
// TODO: provide a formal mechanism for player startup failure
Box::new(RemoteBotHandle {
sender: server_msg_snd,
player_id,