auto-sync: 2026-05-26 11:47:58

src/daemon/spawner.py

@@ -659,13 +659,38 @@ curl -X POST http://{api_host}:{api_port}/api/projects/{project_id}/tasks/{task_
         # 检查 session 状态
         state = self._check_session_state(agent_id)
 
-        # B1: 假死
+        # B1: 假死 — 先复活，连续假死 ≥2 次再 failed
         if state.get("status") == "running" and not state.get("lock_pid_alive", True):
-            logger.error("Agent %s session stuck (session=%s, lock PID dead)",
-                         agent_id, session_id)
-            self._mark_task(db_path, task_id, "failed",
-                            {"reason": "session_stuck", "diagnostics": state})
-            await self._do_on_complete_async(on_complete, agent_id, "session_stuck")
+            # 假死计数
+            stuck_count = self._stuck_counts.get(task_id, 0) + 1
+            self._stuck_counts[task_id] = stuck_count
+
+            if stuck_count >= 2:
+                # 连续假死 ≥2 次，标 failed
+                logger.error("Agent %s session stuck %d times (session=%s, lock PID dead)",
+                             agent_id, stuck_count, session_id)
+                self._mark_task(db_path, task_id, "failed",
+                                {"reason": "session_stuck", "stuck_count": stuck_count,
+                                 "diagnostics": state})
+                await self._do_on_complete_async(on_complete, agent_id, "session_stuck")
+                return
+
+            # 第 1 次假死 → 尝试复活
+            logger.warning("Agent %s session stuck (attempt %d), reviving (session=%s)",
+                           agent_id, stuck_count, session_id)
+            revived = self._revive_session(agent_id)
+            if revived:
+                logger.info("Agent %s session revived, releasing counter for ticker re-dispatch",
+                            agent_id)
+                # release counter → 任务保持 working → ticker 下次 re-dispatch
+                await self._do_on_complete_async(on_complete, agent_id, "session_revived")
+            else:
+                # 复活失败 → 标 failed
+                logger.error("Agent %s revive failed, marking failed", agent_id)
+                self._mark_task(db_path, task_id, "failed",
+                                {"reason": "revive_failed", "stuck_count": stuck_count,
+                                 "diagnostics": state})
+                await self._do_on_complete_async(on_complete, agent_id, "revive_failed")
             return
 
         # B2/B3/B4: 进程还活着