ROS_Test - RicoJia/notes GitHub Wiki

========================================================================

========================================================================

1. testing: you have three levels of testing:

   1. rosunit (unit test, same functionality as gtest)
   2. google gtest
   3. rostest (launch file that launches multiple nodes)

2. gtest for ROS

   1. Basic procedure:
   • In test.cpp

   TEST(TestSuite, testCase2)
   {
     const tf2::Stamped<Eigen::Vector3d> v(Eigen::Vector3d(1,2,3), ros::Time(5), "test");
     tf2::Stamped<Eigen::Vector3d> v1;
     geometry_msgs::PointStamped p1;
     tf2::convert(v, p1);
     tf2::convert(p1, v1);
     EXPECT_EQ(v, v1);
   }
   int main(int argc, char **argv){
    testing::InitGoogleTest(&argc, argv);
    return RUN_ALL_TESTS();
   }

   • In CMakelist.txt

   if(CATKIN_ENABLE_TESTING)
    catkin_add_gtest(tf2_eigen-test test/tf2_eigen-test.cpp)
    target_link_libraries(tf2_eigen-test ${catkin_LIBRARIES} ${GTEST_LIBRARIES})
   endif()
   

   • In package.xml

    <test_depend>rosunit</test_depend>
   

   • Run test with
     1. catkin_make
        • ./bin/test/utest
        • or ```catkin_make run_tests``, or
        • or catkin_make run_tests<TAB><TAB> to find the package to run.
     2. catkin build

     catkin build -v -i --make-args tests -- PKG_NAME
     catkin build -v -i --make-args test -- PKG_NAME
     catkin_test_results --verbose build  #(need this to view results)
     

   2. TestSuite is CamelCased testFunction is camelCased.

   3. std::cout should work!

   4. Test expressions Use ASSERT when we stop the test when a test fail. Use EXPECT so we can still get values out and continue the test.

   • add custom message ASSERT_EQ(ans.x, vec.x)<< "Talker add x's failed"<<.ans.x;
   • string comparison EXPECT_STREQ(str1,str2);
   • near EXPECT_NEAR(val1, val2, abs_error);

   Internal Working: Print from rostest node: rostest is running on a randomly assigned port, and you don't know which port they're running. A: so after catkin_make run_all_tests, which compiles and launches the rostest. test, you can do $rostest test_file -r, then use rosconsole

========================================================================

======================================================================== https://web.eecs.umich.edu/~sugih/pointers/gdbQS.html

Basic procedure:

0. catkin_make -DCMAKE_BUILD_TYPE=Debug

   • if there's no extra compile mode, use this to debug. (in gdb, it should be g++ -g something.cc -o something)

1. rosrun --prefix 'gdb --args' nuslam landmarks

2. rosparam load dump.yaml namespace for loading the yaml file.

3. if your hit run, the program will stop at where it crashes. or hit ctrl+c, where it will stop somewhere.

4. if you see some system function, turn on tui: tui enable, hit up

   • Debugging: GDB https://u.osu.edu/cstutorials/2018/09/28/how-to-debug-c-program-using-gdb-in-6-simple-steps/
     1. compile your code, get a.out
     2. gdb

5. Keep Gazebo and Rviz running, and run the launch file separately.

6. Use a launchfile. add launch-prefix="gdb --args" output="screen"

7. GDB commands

   1. bt: backtrace.

   2. p is print a. print eigen vectors and values. print *X.data()@Length_X where X is the eigen variable and Length_X is the product of its rows and columns

   3. you can do p mu.operator(something) you can run an operation.

   4. n is next line, it will skip an entire function!!

   5. b linenum: sets break point. in ros, do b namespace::class::function:linenum

   6. info b: see breakpoints

   7. delete breakpoint_num: delete a breakpoint

   8. s is step

   9. c is continue

   10. 牛逼: f is frame, which is the level of the function stack you're in. so you can go back to the previous frame and print the variables there!!!!

========================================================================

========================================================================

1. rostest is 100% compatible with roslaunch
   • One difference is all nodes are shutdown automatically, once tests are done
   • are tests launched one at a time?