Update matching for lifecycle service types to support Opensplice

[dhood]

@ironmig noticed that ros2 service list wasn't reporting Opensplice lifecycle nodes properly (e.g. ros2 run lifecycle lifecycle_talker).

The service name and type for opensplice that's reported by get_service_names_and_types is:

Topic: /lc_talker/get_state
Type: 'lifecycle_msgs/Sample_GetState'

I took a quick look at the opensplice typesupport, and it seems to be on purpose that we use samples vs msgs names so I assume that that is a requirement of opensplice's typesupport. As such this PR loosens the check to match it, rather than attempting to remove Sample_ from the type. It may very well be feasible to make get_service_names_and_types report the type without having Sample_ in it, but if so we can do that at a later date IMO.

[Karsten1987]

I can confirm that this resolves the problem.

[mikaelarguedas]

Tested this and now instead of returning "node not found" it crashes: ros2 lifecycle get lc_talker.

Failed to wait on wait set: failed to detach guard condition, at /home/mikael/work/ros2/bouncy_ws/src/ros2/rmw_opensplice/rmw_opensplice_cpp/src/rmw_wait.cpp:375, at /home/mikael/work/ros2/bouncy_ws/src/ros2/rcl/rcl/src/rcl/wait.c:647

@dhood Is this something you experienced when testing this change?

I think it's worth fixing the reported type to not include the Sample as this is messing with other tools as well. For example if I want to call the service directly with ros2 service call /lc_talker/get_state lifecycle_msgs/Sample_GetState (the types provided by the autocopletion) it will crash because the our message structures have no idea of what Sample_GetState is.

So I recommend fixing it in get_service_names_and_types directly rather than working around it with this patch.

I'll try to submit a patch later today

[dirk-thomas]

I am not sure if this is the right way to address the problem. Imo the underlying API needs to provide the same information across all rmw impl. otherwise all user level code has to handle the difference. Therefore I think this should be patched in rmw_opensplice_cpp instead.

[dhood]

Yeah, realising that what this API returns is representing ROS msg types as opposed to actual DDS type, it's reasonable to change that conversion layer without modifying the type sent on the wire. Sounds like @mikaelarguedas is on it

[mikaelarguedas]

Opened ros2/rmw_opensplice#234 with the fix I had in mind

[tfoote]

superseded by ros2/rmw_opensplice#234