[ROS2] increase history_depth for subscrition

[mitsudome-r]

This resolves issue : #167.

This makes the following modfiication:

• The new parameter history_depth for diagnostic_aggregator node to enable users to set history depth size(i.e. queue size) for subscription.
• Change default depth size to 1000 from 1.

Signed-off-by: mitsudome-r ryohsuke.mitsudome@tier4.jp

[Karsten1987]

This looks good to me. Please target the PR against the ros2 branch. I'll take care of a backport to existing distributions and cut a release.

[mitsudome-r]

@Karsten1987
Thank you for the review.
I changed the target to ros2-devel. I also rebased to the branch as well.

[tgreier]

@Karsten1987 Hi Karsten. We need this PR for Foxy. But, it looks like this PR has stalled. What needs to be done to progress it?

[Karsten1987]

Just a little nitpick, otherwise all almost good, except the linters.

[Karsten1987]

@tgreier The changes reflected as is in this PR can't be backported to Foxy due to ABI incompatibility.
One way I would be okay with is to "simply" hardcode the QoS to a keep_last(1000) which should solve the problem described in #167 and achieve same behavior as ROS1.
Opinions?

[Karsten1987]

actually, I've just seen that uncrustify and cpplint aren't happy. Mind addressing them?

You'll find the details in the GH actions, e.g.:

  [18.595s] 12: --- src/aggregator.cpp
  [18.595s] 12: +++ src/aggregator.cpp.uncrustify
  [18.595s] 12: @@ -103 +103,2 @@
  [18.595s] 12: -    "/diagnostics", rclcpp::SystemDefaultsQoS().keep_last(history_depth_), std::bind(&Aggregator::diagCallback, this, _1));
  [18.595s] 12: +    "/diagnostics", rclcpp::SystemDefaultsQoS().keep_last(history_depth_),
  [18.595s] 12: +    std::bind(&Aggregator::diagCallback, this, _1));