When unit tests require large data files, it is better to download them from the web than include them in your source repository. This should only happen in a conditional block when testing is enabled.
Each of these files declares a target name. Running "make tests" or "make run_tests" will automatically download them all. Individual test targets do not depend on the tests target, but you can use it to download them before running one:
$ make tests run_tests_your_package
Catkin provides a convenient command, used like this:
if (CATKIN_ENABLE_TESTING)
catkin_download_test_data(
${PROJECT_NAME}_32e.pcap
http://download.ros.org/data/velodyne/32e.pcap
MD5 e41d02aac34f0967c03a5597e1d554a9)
endif()
The first parameter is the target name, which should normally include your project name to avoid conflicts with other packages.
The second parameter is the URL to read. If you release your package to the ROS build farm, make sure this URL is available reliably. Otherwise, tests will fail randomly, annoying everyone. Contact ros-release@code.ros.org if you need to host some data near the build servers.
The MD5 argument is a good way to avoid testing with corrupted data.
The default destination for the file downloaded above is within your package’s build directory. The file name is the base name of the URL. You can use the DESTINATION argument to put it somewhere else.
For example, sometimes a rostest script wants to use $(find your_package) to access the test data. Here is how to put the file in devel-space, where roslaunch can resolve it:
if (CATKIN_ENABLE_TESTING)
catkin_download_test_data(
${PROJECT_NAME}_32e.pcap
http://download.ros.org/data/velodyne/32e.pcap
DESTINATION ${CATKIN_DEVEL_PREFIX}/${CATKIN_PACKAGE_SHARE_DESTINATION}/tests
MD5 e41d02aac34f0967c03a5597e1d554a9)
endif()
Then, the test script can pass it as a parameter, like this:
<node pkg="your_package" type="your_node" name="your_node">
<param name="data" value="$(find your_package)/tests/32e.pcap"/>
</node>