Fix inner sql execution refered to destroyed SQL bug

src/observer/ob_inner_sql_connection.cpp

@@ -61,8 +61,21 @@ public:
   virtual int execute(sql::ObSql& engine, sql::ObSqlCtx& ctx, sql::ObResultSet& res)
   {
     observer::ObReqTimeGuard req_timeinfo_guard;
-    res.get_session().store_query_string(sql_);
-    return engine.stmt_query(sql_, ctx, res);
+    int ret = OB_SUCCESS;
+    // Deep copy sql, because sql may be destroyed before result iteration.
+    const int64_t alloc_size = sizeof(ObString) + sql_.length() + 1;  // 1 for C terminate char
+    void *mem = res.get_mem_pool().alloc(alloc_size);
+    if (NULL == mem) {
+      ret = OB_ALLOCATE_MEMORY_FAILED;
+      LOG_WARN("allocate memory failed", K(ret));
+    } else {
+      ObString *dup_sql = new (mem) ObString(sql_.length(), sql_.length(), static_cast<char *>(mem) + sizeof(ObString));
+      MEMCPY(dup_sql->ptr(), sql_.ptr(), sql_.length());
+      dup_sql->ptr()[sql_.length()] = '\0';
+      res.get_session().store_query_string(*dup_sql);
+      ret = engine.stmt_query(sql_, ctx, res);
+    }
+    return ret;
   }
 
   // process result after result open