feat: add timeout handling for agent session closure

Added timeout mechanisms with logging for graceful session closure across multiple agent types (claudecode, codex, cursor, gemini, opencode, qoder). Modified engine cleanup to handle timeouts and improve logging. Sessions now timeout after 8-10 seconds with appropriate warnings/errors.

generated by llmgit

Co-Authored-By: Claude <noreply@anthropic.com>

agent/claudecode/session.go

@@ -466,8 +466,18 @@ func (cs *claudeSession) Alive() bool {
 
 func (cs *claudeSession) Close() error {
 	cs.cancel()
-	<-cs.done
-	return nil
+
+	select {
+	case <-cs.done:
+		return nil
+	case <-time.After(8 * time.Second):
+		slog.Warn("claudeSession: graceful close timed out, killing process")
+		if cs.cmd != nil && cs.cmd.Process != nil {
+			_ = cs.cmd.Process.Kill()
+		}
+		<-cs.done
+		return nil
+	}
 }
 
 // filterEnv returns a copy of env with entries matching the given key removed.

agent/codex/session.go

@@ -13,6 +13,7 @@ import (
 	"strings"
 	"sync"
 	"sync/atomic"
+	"time"
 	"unicode/utf8"
 
 	"github.com/chenhg5/cc-connect/core"
@@ -291,7 +292,16 @@ func (cs *codexSession) Alive() bool {
 func (cs *codexSession) Close() error {
 	cs.alive.Store(false)
 	cs.cancel()
-	cs.wg.Wait()
+	done := make(chan struct{})
+	go func() {
+		cs.wg.Wait()
+		close(done)
+	}()
+	select {
+	case <-done:
+	case <-time.After(8 * time.Second):
+		slog.Warn("codexSession: close timed out, abandoning wg.Wait")
+	}
 	close(cs.events)
 	return nil
 }

agent/cursor/session.go

@@ -13,6 +13,7 @@ import (
 	"strings"
 	"sync"
 	"sync/atomic"
+	"time"
 	"unicode/utf8"
 
 	"github.com/chenhg5/cc-connect/core"
@@ -382,7 +383,16 @@ func (cs *cursorSession) Alive() bool {
 func (cs *cursorSession) Close() error {
 	cs.alive.Store(false)
 	cs.cancel()
-	cs.wg.Wait()
+	done := make(chan struct{})
+	go func() {
+		cs.wg.Wait()
+		close(done)
+	}()
+	select {
+	case <-done:
+	case <-time.After(8 * time.Second):
+		slog.Warn("cursorSession: close timed out, abandoning wg.Wait")
+	}
 	close(cs.events)
 	return nil
 }

agent/gemini/session.go

@@ -13,6 +13,7 @@ import (
 	"strings"
 	"sync"
 	"sync/atomic"
+	"time"
 	"unicode/utf8"
 
 	"github.com/chenhg5/cc-connect/core"
@@ -370,7 +371,16 @@ func (gs *geminiSession) Alive() bool {
 func (gs *geminiSession) Close() error {
 	gs.alive.Store(false)
 	gs.cancel()
-	gs.wg.Wait()
+	done := make(chan struct{})
+	go func() {
+		gs.wg.Wait()
+		close(done)
+	}()
+	select {
+	case <-done:
+	case <-time.After(8 * time.Second):
+		slog.Warn("geminiSession: close timed out, abandoning wg.Wait")
+	}
 	close(gs.events)
 	return nil
 }

agent/opencode/session.go

@@ -12,6 +12,7 @@ import (
 	"os/exec"
 	"sync"
 	"sync/atomic"
+	"time"
 	"unicode/utf8"
 
 	"github.com/chenhg5/cc-connect/core"
@@ -305,7 +306,16 @@ func (s *opencodeSession) Alive() bool {
 func (s *opencodeSession) Close() error {
 	s.alive.Store(false)
 	s.cancel()
-	s.wg.Wait()
+	done := make(chan struct{})
+	go func() {
+		s.wg.Wait()
+		close(done)
+	}()
+	select {
+	case <-done:
+	case <-time.After(8 * time.Second):
+		slog.Warn("opencodeSession: close timed out, abandoning wg.Wait")
+	}
 	close(s.events)
 	return nil
 }

agent/qoder/session.go

@@ -13,6 +13,7 @@ import (
 	"strings"
 	"sync"
 	"sync/atomic"
+	"time"
 	"unicode/utf8"
 
 	"github.com/chenhg5/cc-connect/core"
@@ -276,7 +277,16 @@ func (qs *qoderSession) Alive() bool {
 func (qs *qoderSession) Close() error {
 	qs.alive.Store(false)
 	qs.cancel()
-	qs.wg.Wait()
+	done := make(chan struct{})
+	go func() {
+		qs.wg.Wait()
+		close(done)
+	}()
+	select {
+	case <-done:
+	case <-time.After(8 * time.Second):
+		slog.Warn("qoderSession: close timed out, abandoning wg.Wait")
+	}
 	close(qs.events)
 	return nil
 }

core/engine.go

@@ -253,14 +253,20 @@ func (e *Engine) Stop() error {
 	e.cancel()
 
 	e.interactiveMu.Lock()
-	for key, state := range e.interactiveStates {
-		if state.agentSession != nil {
-			state.agentSession.Close()
-		}
-		delete(e.interactiveStates, key)
+	states := make(map[string]*interactiveState, len(e.interactiveStates))
+	for k, v := range e.interactiveStates {
+		states[k] = v
+		delete(e.interactiveStates, k)
 	}
 	e.interactiveMu.Unlock()
 
+	for key, state := range states {
+		if state.agentSession != nil {
+			slog.Debug("engine.Stop: closing agent session", "session", key)
+			state.agentSession.Close()
+		}
+	}
+
 	var errs []error
 	for _, p := range e.platforms {
 		if err := p.Stop(); err != nil {
@@ -561,17 +567,29 @@ func (e *Engine) getOrCreateInteractiveState(sessionKey string, p Platform, repl
 
 func (e *Engine) cleanupInteractiveState(sessionKey string) {
 	e.interactiveMu.Lock()
-	defer e.interactiveMu.Unlock()
-
 	state, ok := e.interactiveStates[sessionKey]
-	if ok && state.agentSession != nil {
+	delete(e.interactiveStates, sessionKey)
+	e.interactiveMu.Unlock()
+
+	if ok && state != nil && state.agentSession != nil {
+		slog.Debug("cleanupInteractiveState: closing agent session", "session", sessionKey)
 		closeStart := time.Now()
-		state.agentSession.Close()
-		if elapsed := time.Since(closeStart); elapsed >= slowAgentClose {
-			slog.Warn("slow agent session close", "elapsed", elapsed, "session", sessionKey)
+
+		done := make(chan struct{})
+		go func() {
+			state.agentSession.Close()
+			close(done)
+		}()
+
+		select {
+		case <-done:
+			if elapsed := time.Since(closeStart); elapsed >= slowAgentClose {
+				slog.Warn("slow agent session close", "elapsed", elapsed, "session", sessionKey)
+			}
+		case <-time.After(10 * time.Second):
+			slog.Error("agent session close timed out (10s), abandoning", "session", sessionKey)
 		}
 	}
-	delete(e.interactiveStates, sessionKey)
 }
 
 func (e *Engine) processInteractiveEvents(state *interactiveState, session *Session, sessionKey string, turnStart time.Time) {
@@ -803,7 +821,9 @@ func (e *Engine) handleCommand(p Platform, msg *Message, raw string) bool {
 }
 
 func (e *Engine) cmdNew(p Platform, msg *Message, args []string) {
+	slog.Info("cmdNew: cleaning up old session", "session_key", msg.SessionKey)
 	e.cleanupInteractiveState(msg.SessionKey)
+	slog.Info("cmdNew: cleanup done, creating new session", "session_key", msg.SessionKey)
 	name := "session"
 	if len(args) > 0 {
 		name = strings.Join(args, " ")
@@ -866,6 +886,7 @@ func (e *Engine) cmdSwitch(p Platform, msg *Message, args []string) {
 	}
 	prefix := strings.TrimSpace(args[0])
 
+	slog.Info("cmdSwitch: listing agent sessions", "session_key", msg.SessionKey)
 	agentSessions, err := e.agent.ListSessions(e.ctx)
 	if err != nil {
 		e.reply(p, msg.ReplyCtx, fmt.Sprintf("❌ %v", err))
@@ -884,7 +905,9 @@ func (e *Engine) cmdSwitch(p Platform, msg *Message, args []string) {
 		return
 	}
 
+	slog.Info("cmdSwitch: cleaning up old session", "session_key", msg.SessionKey)
 	e.cleanupInteractiveState(msg.SessionKey)
+	slog.Info("cmdSwitch: cleanup done", "session_key", msg.SessionKey)
 
 	session := e.sessions.GetOrCreateActive(msg.SessionKey)
 	session.AgentSessionID = matched.ID