huskies: merge 918

server/src/service/gateway/io.rs

@@ -431,6 +431,10 @@ pub type ActiveProject = std::sync::Arc<tokio::sync::RwLock<String>>;
 /// `gateway_event_tx` — when `Some`, the bot will subscribe to the gateway
 /// status broadcaster and forward [`super::GatewayStatusEvent`]s to its
 /// configured Matrix rooms with a `[project-name]` prefix.
+///
+/// Returns `(abort_handle, shutdown_tx)`.  The caller **must** hold
+/// `shutdown_tx` for the bot's lifetime and send `Some(ShutdownReason)` on it
+/// before process exit so the bot can announce "going offline" to its rooms.
 pub fn spawn_gateway_bot(
     config_dir: &Path,
     active_project: ActiveProject,
@@ -438,7 +442,10 @@ pub fn spawn_gateway_bot(
     gateway_project_urls: BTreeMap<String, String>,
     port: u16,
     gateway_event_tx: Option<tokio::sync::broadcast::Sender<super::GatewayStatusEvent>>,
-) -> Option<tokio::task::AbortHandle> {
+) -> (
+    Option<tokio::task::AbortHandle>,
+    tokio::sync::watch::Sender<Option<crate::rebuild::ShutdownReason>>,
+) {
     use crate::agents::AgentPool;
     use crate::services::Services;
     use tokio::sync::{broadcast, mpsc};
@@ -458,7 +465,8 @@ pub fn spawn_gateway_bot(
 
     let (shutdown_tx, shutdown_rx) =
         tokio::sync::watch::channel::<Option<crate::rebuild::ShutdownReason>>(None);
-    std::mem::forget(shutdown_tx);
+    // shutdown_tx is intentionally NOT forgotten — the caller holds it and
+    // sends Some(reason) on gateway shutdown so the bot announces "going offline".
 
     let agents = std::sync::Arc::new(AgentPool::new(port, watcher_tx.clone()));
 
@@ -498,7 +506,7 @@ pub fn spawn_gateway_bot(
         config_dir.join(".huskies").join("timers.json"),
     ));
     let gateway_event_rx = gateway_event_tx.map(|tx| tx.subscribe());
-    crate::chat::transport::matrix::spawn_bot(
+    let handle = crate::chat::transport::matrix::spawn_bot(
         config_dir,
         watcher_tx,
         services,
@@ -508,5 +516,48 @@ pub fn spawn_gateway_bot(
         gateway_project_urls,
         timer_store,
         gateway_event_rx,
-    )
+    );
+    (handle, shutdown_tx)
+}
+
+// ── Tests ────────────────────────────────────────────────────────────────────
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Regression test for story 918: `spawn_gateway_bot` must return a live
+    /// `shutdown_tx` (not one dropped via `std::mem::forget`) so that callers
+    /// can signal `Some(reason)` and the bot's shutdown watcher receives it.
+    #[tokio::test]
+    async fn spawn_gateway_bot_shutdown_tx_not_forgotten() {
+        let tmp = tempfile::tempdir().unwrap();
+        let active = std::sync::Arc::new(tokio::sync::RwLock::new("proj".to_string()));
+        let (event_tx, _) = tokio::sync::broadcast::channel(4);
+
+        let (handle, shutdown_tx) = spawn_gateway_bot(
+            tmp.path(),
+            active,
+            vec!["proj".to_string()],
+            std::collections::BTreeMap::new(),
+            3001,
+            Some(event_tx),
+        );
+
+        // No bot.toml in tmp → no abort handle spawned.
+        assert!(handle.is_none());
+
+        // The shutdown_tx must be live: subscribe a receiver and verify that
+        // sending Some(reason) is observed — this would fail if the sender
+        // had been forgotten (channel closed, send returns Err).
+        let rx = shutdown_tx.subscribe();
+        shutdown_tx
+            .send(Some(crate::rebuild::ShutdownReason::Manual))
+            .expect("shutdown_tx must not be closed (was previously std::mem::forget'd)");
+        assert_eq!(
+            *rx.borrow(),
+            Some(crate::rebuild::ShutdownReason::Manual),
+            "shutdown receiver must see the Manual reason"
+        );
+    }
 }

server/src/service/gateway/mod.rs

@@ -22,6 +22,7 @@ pub use io::{
     spawn_gateway_notification_poller,
 };
 
+use crate::rebuild::ShutdownReason;
 use io::Client;
 use std::collections::{BTreeMap, HashMap};
 use std::path::PathBuf;
@@ -111,6 +112,12 @@ pub struct GatewayState {
     pub port: u16,
     /// Abort handle for the running Matrix bot task (if any).
     pub bot_handle: Arc<TokioMutex<Option<tokio::task::AbortHandle>>>,
+    /// Watch sender used to signal the Matrix bot to shut down gracefully.
+    ///
+    /// Send `Some(ShutdownReason)` on this channel before process exit so the
+    /// bot task receives the signal and posts an "going offline" announcement.
+    pub bot_shutdown_tx:
+        Arc<TokioMutex<Option<tokio::sync::watch::Sender<Option<ShutdownReason>>>>>,
     /// Broadcast sender for [`GatewayStatusEvent`]s pushed by project nodes.
     ///
     /// Call `event_tx.subscribe()` to obtain a receiver for outbound fan-out.
@@ -142,6 +149,7 @@ impl GatewayState {
             config_dir,
             port,
             bot_handle: Arc::new(TokioMutex::new(None)),
+            bot_shutdown_tx: Arc::new(TokioMutex::new(None)),
             event_tx,
         })
     }
@@ -462,7 +470,7 @@ pub async fn save_bot_config_and_restart(state: &GatewayState, content: &str) ->
             .map(|(name, entry)| (name.clone(), entry.url.clone()))
             .collect();
 
-        let new_handle = io::spawn_gateway_bot(
+        let (new_handle, new_shutdown_tx) = io::spawn_gateway_bot(
             &state.config_dir,
             Arc::clone(&state.active_project),
             gateway_projects,
@@ -471,6 +479,7 @@ pub async fn save_bot_config_and_restart(state: &GatewayState, content: &str) ->
             Some(state.event_tx.clone()),
         );
         *handle = new_handle;
+        *state.bot_shutdown_tx.lock().await = Some(new_shutdown_tx);
     }
 
     crate::slog!("[gateway] Bot configuration saved; bot restarted");