Parent: #108
Problem
crates/app/src/network_bridge.rs is peppered with let _ = event_tx.send(...) calls. Most of these are benign — event_tx is std::sync::mpsc::Sender<NetworkBridgeEvent>, which is unbounded, so send() only fails when the receiver is dropped during shutdown.
But not all of them are benign. Line 347:
NetworkEvent::ChunkRequested { channel, hash, .. } => {
let response = if let Some(data) = file_mgr.get_chunk(&hash) {
ChunkResponse::Found { hash, data: data.to_vec() }
} else {
ChunkResponse::NotFound { hash }
};
let _ = node.respond_chunk(channel, response);
}node.respond_chunk returns a real error — if it fails, the requester hangs waiting for a chunk that will never arrive. There are a handful of other non-mpsc let _ = sites that similarly swallow real errors.
Fix
- Grep
crates/app/src/network_bridge.rs for all let _ = occurrences.
- Classify each:
event_tx.send(...) — unbounded mpsc, safe to keep silent (or add a shutdown-only if let Err(...) with a debug! log).- Any other
Result-returning call (respond_chunk, identity saves, etc.) — log at warn level with context and, where possible, mark the associated state so the UI can surface it.
- For
respond_chunk specifically, at least a warn!(%hash, %err, "failed to respond to chunk request") line.
Out of scope
This issue is an audit, not a refactor. Don't change the mpsc channel type. That's a separate larger change if we ever decide we want backpressure.
Test
For respond_chunk, add a unit test that forces the channel to fail and asserts a warning is logged.
Parent: #108
Problem
crates/app/src/network_bridge.rsis peppered withlet _ = event_tx.send(...)calls. Most of these are benign —event_txisstd::sync::mpsc::Sender<NetworkBridgeEvent>, which is unbounded, sosend()only fails when the receiver is dropped during shutdown.But not all of them are benign. Line 347:
node.respond_chunkreturns a real error — if it fails, the requester hangs waiting for a chunk that will never arrive. There are a handful of other non-mpsclet _ =sites that similarly swallow real errors.Fix
crates/app/src/network_bridge.rsfor alllet _ =occurrences.event_tx.send(...)— unbounded mpsc, safe to keep silent (or add a shutdown-onlyif let Err(...)with adebug!log).Result-returning call (respond_chunk, identity saves, etc.) — log atwarnlevel with context and, where possible, mark the associated state so the UI can surface it.respond_chunkspecifically, at least awarn!(%hash, %err, "failed to respond to chunk request")line.Out of scope
This issue is an audit, not a refactor. Don't change the mpsc channel type. That's a separate larger change if we ever decide we want backpressure.
Test
For
respond_chunk, add a unit test that forces the channel to fail and asserts a warning is logged.