fastddsgen and ros2 gen msg, TypeSupport member m_typeSize id different, why?

[Eternity1987]

file Vec2Image1m.msg ：

shm_msgs/Image1m[2] images

fastddsgen:
fastddsgen -d {src_path} -de final -I . -t {temp_path} -typeros2 -cs {idls_str}

ros2 gen:

fastddsgen m_typeSize value is: 2098236 + 4
ros2 m_typeSize value is: 2098232 + 4

can you help me explain?

[ZhenshengLee]

1. the m_typeSize is decided by the definition of the struct.
2. the struct is defined by the generated headers.
3. for the reason why they are different, please check the generated header from fastddsgen and rosidl.
4. the header generated by fastddsgen can be seen easily
5. the header generated by rosidl can be checked in build/${PACKAGE_NAME}/rosidl_generator_cpp/${PACKAGE_NAME}/msg/Vec2Image1m.hpp
6. you can check more info in build/${PACKAGE_NAME}/rosidl_typesupport_fastrtps_cpp/${PACKAGE_NAME}/msg/detail

[Eternity1987]

1. the m_typeSize is decided by the definition of the struct.m_typeSize由结构体的定义决定。
2. the struct is defined by the generated headers.结构体由生成的标头定义。
3. for the reason why they are different, please check the generated header from fastddsgen and rosidl.关于它们不同的原因，请检查 FastDDSGEN 和 Rosidl 生成的 Header。
4. the header generated by fastddsgen can be seen easilyFastDDSGEN 生成的 header 可以很容易地看到
5. the header generated by rosidl can be checked in build/${PACKAGE_NAME}/rosidl_generator_cpp/${PACKAGE_NAME}/msg/Vec2Image1m.hppRosidl 生成的标头可以签入 build/${PACKAGE_NAME}/rosidl_generator_cpp/${PACKAGE_NAME}/msg/Vec2Image1m.hpp
6. you can check more info in build/${PACKAGE_NAME}/rosidl_typesupport_fastrtps_cpp/${PACKAGE_NAME}/msg/detail您可以在 build/${PACKAGE_NAME}/rosidl_typesupport_fastrtps_cpp/${PACKAGE_NAME}/msg/detail

Thank you for your response. I understand the files you are talking about. In ROS 2, the size is calculated in the code,

size_t
ROSIDL_TYPESUPPORT_FASTRTPS_CPP_PUBLIC_shm_msgs
max_serialized_size_Vec2Image1m(
  bool & full_bounded,
  bool & is_plain,
  size_t current_alignment)
{
  size_t initial_alignment = current_alignment;

  const size_t padding = 4;
  const size_t wchar_size = 4;
  size_t last_member_size = 0;
  (void)last_member_size;
  (void)padding;
  (void)wchar_size;

  full_bounded = true;
  is_plain = true;

  // Member: images
  {
    size_t array_size = 2;
    last_member_size = 0;
    for (size_t index = 0; index < array_size; ++index) {
      bool inner_full_bounded;
      bool inner_is_plain;
      size_t inner_size =
        shm_msgs::msg::typesupport_fastrtps_cpp::max_serialized_size_Image1m(
        inner_full_bounded, inner_is_plain, current_alignment);
      last_member_size += inner_size;
      current_alignment += inner_size;
      full_bounded &= inner_full_bounded;
      is_plain &= inner_is_plain;
    }
  }

  size_t ret_val = current_alignment - initial_alignment;
  if (is_plain) {
    // All members are plain, and type is not empty.
    // We still need to check that the in-memory alignment
    // is the same as the CDR mandated alignment.
    using DataType = shm_msgs::msg::Vec2Image1m;
    is_plain =
      (
      offsetof(DataType, images) +
      last_member_size
      ) == ret_val;
  }

  return ret_val;
}

similar to Fast DDS, both showing a size of 2098232 when using sizeof. However, in ROS 2, this value is determined by a function, while in Fast DDS, during template instantiation, a macro shows 2098236, which is 4 more than in ROS 2. I'm not entirely sure about this difference.

This discrepancy might cause Fast DDS to send messages that ROS 2 cannot receive, while ROS 2 can send messages that Fast DDS can receive.