Fix misleading deprecated warning when using launch arguments

[ivanpauno]

Before this patch, running:
ros2 launch any_package any_launch_file some_arg:=some_value
ended up with a warning:

[WARN] [1576175943.679492205] [rcl]: Found remap rule 'asd:=bsd'. This syntax is deprecated. Use '--ros-args --remap asd:=bsd' instead.

That's misleading, because the only way of passing launch arguments is by adding key:=value at the end of the command.

To fix this, I did the following:

• Stop using the LaunchContext arguments in the initialization of the rclpy.Context:

  launch_ros/launch_ros/launch_ros/default_launch_description.py

  Line 60 in d6dd5c0

  rclpy.init(args=context.argv)

  
  Add a custom way of passing those arguments.
• Pass no arguments in the creation of the rclpy.Context in ros2launch

  launch_ros/ros2launch/ros2launch/api/api.py

  Line 144 in 8fae305

  rclpy.init(args=[], context=context)

  
  Before, we were passing the same key values configurations that ros2launch received, which can cause problems (e.g.: being used as remapping rules).
• Finally, deleted the argv option of ros2launch arguments. IIUC, it was always empty as the previous argument (launch configurations) was taking a variable number of options.

I'm not sure if this is the correct way to go, maybe we want to do something different.
e.g.: Filtering the launch configurations from the arguments, and use the others as the arguments of the rclpy.Context. IMO, that may be confusing.

[ivanpauno]

@Karsten1987 FYI.

[wjwwood]

lgtm

I don't know if we need to pass ROS arguments to the internal launch_ros node or not, but for now it's probably ok to not do so.

[ivanpauno]

• Linux
• AArch64
• macOS
• Windows

[ivanpauno]

I don't know if we need to pass ROS arguments to the internal launch_ros node or not, but for now it's probably ok to not do so.

As far as I see, we don't need to pass remapping rules or parameters.
The other ROS arguments are logging options, for which we could provide a different way of configuring that.

[ivanpauno]

#120 reintroduced this bug.

I'm working on a fix and a regression test.