Building tests

Run make to build all the tests. The Makefile assumes that the target skiboot tree is located at ../skiboot, override SKIBOOT_PATH to change. make run will execute all the tests for feature checks. make valgrind will run all the tests through valgrind, and complain if there are any errors. Individual tests can be run with make run_test_<name>, e.g. make run_test_void will run the canary void test. Replacing s/run/valgrind will run an individual valgrind test.

The above tests will abort running if any test fails. To run all the tests regardless of pass or fail, run with make -k.

Each test will generate a <test_name>.log file containing all prlog() outputs in the log/ directory. If running in valgrind mode, each test will also generate a valgrind_<test_name>.log file, for analyzing the complaints.

If gcovr is installed, you may also run the coverage and coverage_html make targets. The former option will display the coverage information to stdout. Running coverage_html will output file-by-file coverage information in the html/ directory. Compilations are done with gcov support enabled by default, so even without the useful gcovr utility, coverage reports can still be generated manually.

make clean will remove all the garbage that will clutter the directory.

Writing tests

Tests require four things: a file name starting with test_ and ending in .c, #include "test.c" somewhere at the top, a char *test_name, and a int run_test(void) {..}.

The test_name must be set to something useful to identify the test, probably the same thing as the file name.

The main entry point to the test case is the run_test() function. Put all testing logic in there, and if the test case should pass, return a 0. Otherwise, the function should return literally any other value. Given that test.c is directly included, test cases have access to every static function or variable in keystore.c and secboot_part.c.

Wrappers to the OPAL runtime services have been provided, so that casting each parameter is not necessary. Each wrapper is same name as the original function, just without the opal_ prefix (e.g. opal_secvar_read is secvar_read).

Also included are some potentially useful functions and macros that are not originally part of libstb:

• ASSERT(condition) is a helpful shortcut to print an error and line number, and return nonzero.
• ASSERT_POST(condition, post-expr) is a more flexible version of ASSERT(..) that executes post-expr after printing the error.
  • Example: ASSERT_POST(0 == 1, goto clean); would jump to label "clean" instead of just returning 1.
• list_length counts how many items are in a keystore in-memory list.