Bug in referencing ROS2 package paths in xml

[Serafadam]

Hi, I think I've found a bug, in ROS2 version of package, for xml referencing ros1 version of package searching is used, in ROS2 it's ament_index_cpp::get_package_share_directory instead of ros::package::getPath as per https://answers.ros.org/question/288501/ros2-equivalent-of-rospackagegetpath/

BehaviorTree.CPP/src/xml_parsing.cpp

Line 143 in 47d770c

auto ros_pkg_path = ros::package::getPath( include_node->Attribute("ros_pkg") );

Edit: Also, I have some problems when including files with relative paths, for example to the previous directory ../tree_that_i_want_to_include.xml

On a side note, is there a plan to release newest version of the code in ROS Dashing?

[facontidavide]

thanks. I will have a look at it

[asasine]

I'm not sure this bug would be hit since USING_ROS isn't defined when building with ament

BehaviorTree.CPP/CMakeLists.txt

Lines 74 to 76 in 725eba7

	if ( ament_cmake_FOUND )
	# Not adding -DUSING_ROS since xml_parsing.cpp hasn't been ported to ROS2

Using the ros_pkg attribute would cause a runtime error in that case since #ifdef USING_ROS doesn't pass and #else is called

BehaviorTree.CPP/src/xml_parsing.cpp

Lines 136 to 149 in 725eba7

	#ifdef USING_ROS
	if( file_path.is_absolute() )
	{
	std::cout << "WARNING: <include path=\"...\"> containes an absolute path.\n"
	<< "Attribute [ros_pkg] will be ignored."<< std::endl;
	}
	else {
	auto ros_pkg_path = ros::package::getPath( include_node->Attribute("ros_pkg") );
	file_path = filesystem::path( ros_pkg_path ) / file_path;
	}
	#else
	throw RuntimeError("Using attribute [ros_pkg] in <include>, but this library was compiled "
	"without ROS support. Recompile the BehaviorTree.CPP using catkin");
	#endif

[asasine]

Edit: Also, I have some problems when including files with relative paths, for example to the previous directory ../tree_that_i_want_to_include.xml

@Serafadam I've encountered the same issue and started #324 to track it. I think I found a resolution and created #358, a review from a maintainer would be good since it does change the existing behavior of tree inclusion.