Unexpected hang behavior when using webtransport

[FallingSnow]

When refreshing my browser the webtransport connection is not terminated properly. Salvo's webtransport recv half starts reading 0 bytes, which is acceptable. I use this 0 byte read to find that the connection is terminated.

However sometimes I try to .write_all() on the send half right when the connection has been terminated. This results in .write_all() hanging indefinitely.

Is this expected behavior?

For the time being I've solved this by placing a timeout on write_all.

use tokio::time::timeout;

timeout(Duration::from_secs(1), send.write_all(&packet)).await

---

Edit:

I'm also running into this issue when sending large amounts of data... After sending a lot of data write_all() will hang.

[FallingSnow]

This seems to have been caused by the WebTransport on the client end not being able to read quick enough. This shouldn't be a problem as backpressure should automatically be applied and the write_all() should just slow down, not hang. No whether this is caused by the client not sending the proper signals or salvo is unknown.

As for the solution, I ended up putting the WebTransport inside it's own web worker thread to ensure it could read fast enough.