move child management into supervisor

colinrozzi · colinrozzi · commit 962cf4e5ab00 · 2026-03-14T13:28:03.000-04:00
diff --git a/2026-03-14-the-sky-is-blue.txt b/2026-03-14-the-sky-is-blue.txt
diff --git a/crates/deprecated/theater-handler-io/tests/integration_test.rs b/crates/deprecated/theater-handler-io/tests/integration_test.rs
@@ -197,7 +197,6 @@ async fn test_io_handler_with_test_actor() -> Result<()> {
             manifest_path: manifest_content,
             wasm_bytes: None,
             init_bytes: None,
-            parent_id: None,
             response_tx,
             supervisor_tx: None,
             subscription_tx: Some(event_tx),
diff --git a/crates/deprecated/theater-handler-random/tests/integration_test.rs b/crates/deprecated/theater-handler-random/tests/integration_test.rs
@@ -203,7 +203,6 @@ async fn test_random_handler_with_test_actor() -> Result<()> {
             manifest_path: manifest_content,
             wasm_bytes: None,
             init_bytes: None,
-            parent_id: None,
             response_tx,
             supervisor_tx: None,
             subscription_tx: Some(event_tx),
diff --git a/crates/deprecated/theater-handler-timing/tests/integration_test.rs b/crates/deprecated/theater-handler-timing/tests/integration_test.rs
@@ -202,7 +202,6 @@ async fn test_timing_handler_with_test_actor() -> Result<()> {
             manifest_path: manifest_content,
             wasm_bytes: None,
             init_bytes: None,
-            parent_id: None,
             response_tx,
             supervisor_tx: None,
             subscription_tx: Some(event_tx),
diff --git a/crates/theater-cli/src/commands/run.rs b/crates/theater-cli/src/commands/run.rs
@@ -276,7 +276,6 @@ pub async fn execute_async(args: &RunArgs, _ctx: &CommandContext) -> Result<(),
             name: Some(manifest.name.clone()),
             manifest: Some(manifest),
             response_tx,
-            parent_id: None,
             supervisor_tx: Some(supervisor_tx),
             subscription_tx: None,
         })
diff --git a/crates/theater-cli/src/commands/start.rs b/crates/theater-cli/src/commands/start.rs
@@ -286,7 +286,6 @@ pub async fn execute_async(args: &StartArgs, ctx: &CommandContext) -> Result<(),
             name: Some(manifest.name.clone()),
             manifest: Some(manifest),
             response_tx,
-            parent_id: None,
             supervisor_tx: Some(supervisor_tx),
             subscription_tx: None, // Using global subscription instead
         })
diff --git a/crates/theater-handler-rpc/tests/rpc_integration.rs b/crates/theater-handler-rpc/tests/rpc_integration.rs
@@ -136,7 +136,6 @@ async fn test_rpc_calculator_demo() {
             name: Some(calc_manifest.name.clone()),
             manifest: Some(calc_manifest),
             response_tx: calc_response_tx,
-            parent_id: None,
             supervisor_tx: None,
             subscription_tx: Some(calc_subscription_tx),
         })
@@ -170,7 +169,6 @@ async fn test_rpc_calculator_demo() {
             name: Some(caller_manifest.name.clone()),
             manifest: Some(caller_manifest),
             response_tx: caller_response_tx,
-            parent_id: None,
             supervisor_tx: None,
             subscription_tx: Some(caller_subscription_tx),
         })
diff --git a/crates/theater-handler-supervisor/src/lib.rs b/crates/theater-handler-supervisor/src/lib.rs
diff --git a/crates/theater-handler-tcp/tests/tcp_echo_integration.rs b/crates/theater-handler-tcp/tests/tcp_echo_integration.rs
@@ -112,7 +112,6 @@ async fn test_tcp_echo_and_chain() {
             name: Some(manifest.name.clone()),
             manifest: Some(manifest),
             response_tx,
-            parent_id: None,
             supervisor_tx: None,
             subscription_tx: Some(subscription_tx),
         })
diff --git a/crates/theater-replay-experimenting/src/main.rs b/crates/theater-replay-experimenting/src/main.rs
@@ -327,7 +327,6 @@ pub async fn run_replay_verification(
             wasm_bytes,
             name: Some(recording_manifest.name.clone()),
             manifest: Some(recording_manifest),
-            parent_id: None,
             response_tx,
             supervisor_tx: None,
             subscription_tx: Some(event_tx),
@@ -455,7 +454,6 @@ pub async fn run_replay_verification(
             wasm_bytes: replay_wasm_bytes,
             name: Some(replay_manifest.name.clone()),
             manifest: Some(replay_manifest),
-            parent_id: None,
             response_tx: response_tx2,
             supervisor_tx: None,
             subscription_tx: Some(replay_event_tx),
@@ -568,7 +566,6 @@ pub async fn run_request_replay_verification(
             wasm_bytes,
             name: Some(recording_manifest.name.clone()),
             manifest: Some(recording_manifest),
-            parent_id: None,
             response_tx,
             supervisor_tx: None,
             subscription_tx: Some(event_tx),
@@ -736,7 +733,6 @@ pub async fn run_request_replay_verification(
             wasm_bytes: replay_wasm_bytes,
             name: Some(replay_manifest.name.clone()),
             manifest: Some(replay_manifest),
-            parent_id: None,
             response_tx: response_tx2,
             supervisor_tx: None,
             subscription_tx: Some(replay_event_tx),
@@ -869,7 +865,6 @@ pub async fn run_supervisor_replay_verification(
             wasm_bytes,
             name: Some(recording_manifest.name.clone()),
             manifest: Some(recording_manifest),
-            parent_id: None,
             response_tx,
             supervisor_tx: None,
             subscription_tx: Some(event_tx),
@@ -1051,7 +1046,6 @@ pub async fn run_supervisor_replay_verification(
             wasm_bytes: replay_wasm_bytes,
             name: Some(replay_manifest.name.clone()),
             manifest: Some(replay_manifest),
-            parent_id: None,
             response_tx: response_tx2,
             supervisor_tx: None,
             subscription_tx: Some(replay_event_tx),
@@ -1238,7 +1232,6 @@ pub async fn run_tcp_replay_verification(
             wasm_bytes,
             name: Some(recording_manifest.name.clone()),
             manifest: Some(recording_manifest),
-            parent_id: None,
             response_tx,
             supervisor_tx: None,
             subscription_tx: Some(event_tx),
@@ -1401,7 +1394,6 @@ pub async fn run_tcp_replay_verification(
             wasm_bytes: replay_wasm_bytes,
             name: Some(replay_manifest.name.clone()),
             manifest: Some(replay_manifest),
-            parent_id: None,
             response_tx: response_tx2,
             supervisor_tx: None,
             subscription_tx: Some(replay_event_tx),
diff --git a/crates/theater-server/src/server.rs b/crates/theater-server/src/server.rs
@@ -694,7 +694,6 @@ impl TheaterServer {
                             name: Some(manifest_config.name.clone()),
                             manifest: Some(manifest_config),
                             response_tx: cmd_tx,
-                            parent_id: None,
                             supervisor_tx,
                             subscription_tx,
                         })
diff --git a/crates/theater/examples/run_actor.rs b/crates/theater/examples/run_actor.rs
@@ -92,7 +92,6 @@ async fn main() -> Result<()> {
             name: Some(manifest.name.clone()),
             manifest: Some(manifest),
             response_tx,
-            parent_id: None,
             supervisor_tx: None,
             subscription_tx: None,
         })
diff --git a/crates/theater/src/messages.rs b/crates/theater/src/messages.rs
@@ -128,7 +128,8 @@ pub enum TheaterCommand {
     /// * `name` - Optional actor name for debugging/logging
     /// * `manifest` - Optional manifest for handler configs, replay settings, etc.
     /// * `response_tx` - Channel to receive the result (actor ID or error)
-    /// * `parent_id` - Optional parent actor ID for supervision hierarchy
+    /// * `supervisor_tx` - Optional channel for supervisor to receive lifecycle events
+    /// * `subscription_tx` - Optional channel to subscribe to all actor events
     ///
     /// ## Notes
     ///
@@ -141,7 +142,6 @@ pub enum TheaterCommand {
         name: Option<String>,
         manifest: Option<ManifestConfig>,
         response_tx: oneshot::Sender<Result<TheaterId>>,
-        parent_id: Option<TheaterId>,
         supervisor_tx: Option<Sender<ActorResult>>,
         subscription_tx: Option<Sender<Result<ChainEvent, ActorError>>>,
     },
@@ -156,12 +156,12 @@ pub enum TheaterCommand {
     /// * `manifest_path` - Path to the actor's manifest file
     /// * `wasm_bytes` - Optional pre-loaded WASM bytes. If None, bytes are resolved from manifest.package
     /// * `response_tx` - Channel to receive the result (actor ID or error)
-    /// * `parent_id` - Optional parent actor ID for supervision hierarchy
+    /// * `supervisor_tx` - Optional channel for supervisor to receive lifecycle events
+    /// * `subscription_tx` - Optional channel to subscribe to all actor events
     ResumeActor {
         manifest_path: String,
         wasm_bytes: Option<Vec<u8>>,
         response_tx: oneshot::Sender<Result<TheaterId>>,
-        parent_id: Option<TheaterId>,
         supervisor_tx: Option<Sender<ActorResult>>,
         subscription_tx: Option<Sender<Result<ChainEvent, ActorError>>>,
     },
diff --git a/crates/theater/src/theater_runtime.rs b/crates/theater/src/theater_runtime.rs
@@ -143,8 +143,6 @@ pub struct ActorProcess {
     pub info_tx: mpsc::Sender<ActorInfo>,
     /// Channel for sending control commands to the actor
     pub control_tx: mpsc::Sender<ActorControl>,
-    /// Set of child actor IDs
-    pub children: HashSet<TheaterId>,
     /// Current status of the actor
     pub status: ActorStatus,
     /// Optional actor manifest (for handler configs, replay, etc.)
@@ -244,13 +242,10 @@ impl TheaterRuntime {
                     parent_id,
                     response_tx,
                 } => {
-                    debug!("Getting children for actor: {:?}", parent_id);
-                    if let Some(proc) = self.actors.get(&parent_id) {
-                        let children = proc.children.iter().cloned().collect();
-                        let _ = response_tx.send(children);
-                    } else {
-                        let _ = response_tx.send(Vec::new());
-                    }
+                    // Note: Child tracking is now handled by the supervisor handler.
+                    // This command returns empty - use supervisor handler's internal tracking instead.
+                    debug!("ListChildren called for {:?} (deprecated - supervisor handles child tracking)", parent_id);
+                    let _ = response_tx.send(Vec::new());
                 }
                 TheaterCommand::RestartActor {
                     actor_id,
@@ -298,7 +293,6 @@ impl TheaterRuntime {
                     wasm_bytes,
                     name,
                     manifest,
-                    parent_id,
                     response_tx,
                     supervisor_tx,
                     subscription_tx,
@@ -310,7 +304,6 @@ impl TheaterRuntime {
                             wasm_bytes,
                             name,
                             manifest,
-                            parent_id,
                             supervisor_tx,
                             subscription_tx,
                         )
@@ -337,7 +330,6 @@ impl TheaterRuntime {
                     manifest_path,
                     wasm_bytes,
                     response_tx,
-                    parent_id,
                     supervisor_tx,
                     subscription_tx,
                 } => {
@@ -404,7 +396,6 @@ impl TheaterRuntime {
                             wasm_bytes,
                             name,
                             Some(manifest),
-                            parent_id,
                             supervisor_tx,
                             subscription_tx,
                         )
@@ -622,7 +613,8 @@ impl TheaterRuntime {
     /// * `wasm_bytes` - The WASM module bytes to instantiate
     /// * `name` - Optional actor name for debugging/logging
     /// * `manifest` - Optional manifest for handler configs, replay settings, etc.
-    /// * `parent_id` - Optional ID of the parent actor
+    /// * `supervisor_tx` - Optional channel for supervisor to receive lifecycle events
+    /// * `subscription_tx` - Optional channel to subscribe to all actor events
     ///
     /// ## Returns
     ///
@@ -635,15 +627,13 @@ impl TheaterRuntime {
     /// - Creating communication channels
     /// - Spawning the actor runtime in a new task
     /// - Registering the actor with the runtime
-    /// - Setting up parent-child relationships
     ///
     /// If no manifest is provided, the actor uses global handler defaults.
     async fn spawn_actor(
         &mut self,
         wasm_bytes: Vec<u8>,
         name: Option<String>,
         manifest: Option<ManifestConfig>,
-        parent_id: Option<TheaterId>,
         supervisor_tx: Option<Sender<ActorResult>>,
         subscription_tx: Option<Sender<Result<ChainEvent, ActorError>>>,
     ) -> Result<TheaterId> {
@@ -761,26 +751,12 @@ impl TheaterRuntime {
             operation_tx,
             info_tx,
             control_tx,
-            children: HashSet::new(),
             status: ActorStatus::Running,
             manifest,
             shutdown_controller,
             supervisor_tx,
         };
 
-        if let Some(parent_id) = parent_id {
-            debug!("Adding actor {:?} as child of {:?}", actor_id, parent_id);
-            if let Some(parent) = self.actors.get_mut(&parent_id) {
-                parent.children.insert(actor_id.clone());
-                debug!("Added actor {:?} as child of {:?}", actor_id, parent_id);
-            } else {
-                warn!(
-                    "Parent actor {:?} not found for new actor {:?}",
-                    parent_id, actor_id
-                );
-            }
-        }
-
         self.actors.insert(actor_id.clone(), process);
         debug!("Actor process registered with runtime");
 
@@ -989,34 +965,9 @@ impl TheaterRuntime {
 
         self.chains.remove(&actor_id);
 
-        // Find the actor's children to stop them first
-        let children = if let Some(proc) = self.actors.get(&actor_id) {
-            debug!(
-                "Actor {:?} has {} children to stop first",
-                actor_id,
-                proc.children.len()
-            );
-            proc.children.clone()
-        } else {
-            debug!("Actor {:?} not found", actor_id);
-            return Ok(());
-        };
-
-        // First, stop all children recursively
-        for (index, child_id) in children.iter().enumerate() {
-            debug!(
-                "Stopping child {}/{} with ID {:?} of parent {:?}",
-                index + 1,
-                children.len(),
-                child_id,
-                actor_id
-            );
-            Box::pin(self.stop_actor(child_id.clone(), shutdown_type)).await?;
-            debug!("Successfully stopped child {:?}", child_id);
-        }
-
         // Get the actor process - but DON'T remove it from the map yet
         // We need to keep it in the map while shutdown is in progress
+        // Note: Child stopping is now handled by the supervisor handler
         let proc = match self.actors.get(&actor_id) {
             Some(proc) => proc,
             None => {
diff --git a/crates/theater/tests/unit/messages_tests.rs b/crates/theater/tests/unit/messages_tests.rs
@@ -61,27 +61,23 @@ async fn test_theater_command_stop_actor() {
 #[tokio::test]
 async fn test_theater_command_spawn_actor() {
     let actor_name = "test-actor".to_string();
-    let parent_id = TheaterId::generate();
     let (tx, _rx) = oneshot::channel();
 
     let command = TheaterCommand::SpawnActor {
         wasm_bytes: vec![0, 1, 2, 3],
         name: Some(actor_name.clone()),
         manifest: None,
         response_tx: tx,
-        parent_id: Some(parent_id.clone()),
         supervisor_tx: None,
         subscription_tx: None,
     };
 
     match command {
         TheaterCommand::SpawnActor {
             name: ref n,
-            parent_id: ref parent,
             ..
         } => {
             assert_eq!(*n, Some(actor_name.clone()));
-            assert_eq!(*parent, Some(parent_id));
         }
         _ => panic!("Wrong command type"),
     }
diff --git a/pact/supervisor.pact b/pact/supervisor.pact
@@ -39,3 +39,45 @@ interface supervisor {
         get-child-events: func(child-id: string) -> result<list<list<u8>>, string>
     }
 }
+
+// Supervisor Handler Callbacks
+//
+// Exports that actors should implement to receive notifications about child lifecycle events.
+// These are called by the supervisor handler when children report events.
+interface supervisor-handlers {
+    @package: string = "theater:simple"
+
+    // Simplified error type for WIT compatibility
+    enum wit-error-type {
+        operation-timeout,
+        channel-closed,
+        shutting-down,
+        function-not-found,
+        type-mismatch,
+        internal,
+        serialization-error,
+        update-component-error,
+        paused
+    }
+
+    record wit-actor-error {
+        error-type: wit-error-type,
+        data: option<list<u8>>
+    }
+
+    exports {
+        // Called when a child actor encounters an error
+        handle-child-error: func(child-id: string, error: wit-actor-error) -> result<_, string>
+
+        // Called when a child actor exits successfully
+        handle-child-exit: func(child-id: string, result: option<list<u8>>) -> result<_, string>
+
+        // Called when a child actor is stopped externally (e.g., by stop-child or system shutdown)
+        handle-child-external-stop: func(child-id: string) -> result<_, string>
+
+        // Called for every event a child records (optional - may not be implemented)
+        // event-type: the type string of the event
+        // event-data: the serialized event data
+        handle-child-event: func(event-type: string, event-data: list<u8>) -> result<_, string>
+    }
+}
