QOS history depth

SecretaryBirds · November 11, 2016, 5:10pm

Playing around with QOS depths, I noticed you can set the depth to 0.
Not sure if that’s allowed? Will it really set it to 0, or does it just force it to 1 in the DDS layer?
I couldn’t tell a difference between a depth of 1 and 0, but maybe I just didn’t test the right combo of Pub/Sub to see a difference.

Playing with the OpenSplice demo_ishapes executable, they don’t even allow a depth size of 0.

wjwwood · November 11, 2016, 10:17pm

I believe that is a special (“reserved”) value that indicates that it should use the system default depth:

github.com

ros2/rmw/blob/847784e6a0b714630c12a67eb753439ae234121a/rmw/include/rmw/types.h#L204


uint8_t data[RMW_GID_STORAGE_SIZE];
} rmw_gid_t;


typedef struct RMW_PUBLIC_TYPE rmw_message_info_t
{
// const rmw_time_t received_timestamp;
rmw_gid_t publisher_gid;
bool from_intra_process;
} rmw_message_info_t;


enum {RMW_QOS_POLICY_DEPTH_SYSTEM_DEFAULT = 0};


#if __cplusplus
}
#endif


#endif  // RMW__TYPES_H_

Where the system default may be influenced by some global configuration file or something. In fact you can see that we have a condition in each of our implementations to handle this, for example in our fast rtps code:

github.com

ros2/rmw_fastrtps/blob/11d990d00872261c6b9fce6778180a9f0d761fd3/rmw_fastrtps_cpp/src/functions.cpp#L528


  case RMW_QOS_POLICY_VOLATILE_DURABILITY:
    sattr.qos.m_durability.kind = VOLATILE_DURABILITY_QOS;
    break;
  case RMW_QOS_POLICY_DURABILITY_SYSTEM_DEFAULT:
    break;
  default:
    RMW_SET_ERROR_MSG("Unknown QoS durability policy");
    return false;
}


if (qos_policies.depth != RMW_QOS_POLICY_DEPTH_SYSTEM_DEFAULT) {
  sattr.topic.historyQos.depth = static_cast<int32_t>(qos_policies.depth);
}


// ensure the history depth is at least the requested queue size
assert(sattr.topic.historyQos.depth >= 0);
if (
  sattr.topic.historyQos.kind == KEEP_LAST_HISTORY_QOS &&
  static_cast<size_t>(sattr.topic.historyQos.depth) < qos_policies.depth
)
{

Topic		Replies	Views
What should be the correct behavior of a "reliable" transmission? (throughput problem) Next Generation ROS	3	2001	January 12, 2019
DDS vendor interop -- anybody tried it lately? Next Generation ROS	2	1008	December 5, 2017
A more unified and standard way of configuring the DDS layer Next Generation ROS	18	1890	February 24, 2020
Plan to port more DDS QoS to ROS2 Next Generation ROS ros2 , dds	6	1114	March 23, 2021
Configuring QoS profiles from an external file [Proposal] Next Generation ROS ros2	0	531	August 19, 2020

QOS history depth

Related topics