`tokio::select` running wrong arm

[compact shard]

Hey, i've got the following function which is responsible for calling self.on_start, self.on_refresh, and self.on_stop, which are all async. There's a tokio::sync::mpsc::Receiver for when to refresh and when to stop, but for whatever reason, when i'm emitting to the receiver for stopping, it's sometimes matching on the refresh arm instead (seemingly 50/50 randomly). is a biased select what I'm looking for? is a tokio::select the wrong tool to use in the first place?

async fn start(
  &mut self,
  mut refresh_rx: Receiver<()>,
  mut stop_rx: Receiver<()>,
) {
  tokio::select! {
    output = self.on_start() => output,
    _ = refresh_rx.recv() => {
      println!("refresh called");
      _ = self.on_refresh();
    },
    _ = stop_rx.recv() => {
      println!("stop called");
      _ = self.on_stop();
    },
  }

[merry perch]

are you perhaps dropping the channel senders right after sending on the stop channel? If yes, you are running into a race condition because recv completes when the sender is dropped as well

try Some(_) = refresh_rx.recv() => to only consider actual sending events

[compact shard]

wow i can't believe you figured that out so quickly, that's exactly what I was doing

btw would you happen to know whether i'd run into problems with the refresh arm here if the expectation is that it can be called multiple times?

if i understand it correctly, it'd exit the select once any of the arms complete

[merry perch]

yes, usually select is placed in a loop

[compact shard]

roger that. you're a godsend 🙌

[compact shard]

hmmmm but if i run it in a loop, it'll call on_start again after stopping 🤔 rather than only looping if its receives refresh_rx

just realized i can break within the stop arm, my bad

[merry perch]

either that or you use the "precondition" functionality of select to disable certain branches

i.e. the if ... part in v1 = (&mut rx1), if a.is_none() => a = Some(v1.unwrap()),