NetSurf Unit Testing ==================== [toc] # Overview NetSurf has unit tests integrated in the test directory. These tests use the check unit test framework for C [1]. The tests are in a logical hierachy of "suite", "case" and individual "test". Historicaly we have split suites of tests into separate test programs although the framework does not madate this and some test programs contain more than one suite. # Execution The test programs are executed by using the standard "test" target from the top level make invocation. The "coverage" target additionally generates code coverage reports allowing visibility on how much of a code module is being exercised. The check library must be installed to run the tests and the CI system automatically executes all enabled tests and generates coverage reports for each commit. # Adding tests The test/Makefile defines each indiviadual test program that should be built and executed in the TESTS variable. The test program source files are defined in a xxx_SRCS variable and the make rules will then ensure the target program is built and executed. Each individual test program requires a main function which creates one (or more) suites. The suites are added to a test runner and then executed and the results reported. int main(int argc, char **argv) { int number_failed; SRunner *sr; sr = srunner_create(foo_suite_create()); //srunner_add_suite(sr, bar_suite_create()); srunner_run_all(sr, CK_ENV); number_failed = srunner_ntests_failed(sr); srunner_free(sr); return (number_failed == 0) ? EXIT_SUCCESS : EXIT_FAILURE; } Suite creation is done with a sub function to logically split suite code into sub modules. Each suite has test cases added to it. Suite *foo_suite_create(void) { Suite *s; s = suite_create("foo"); suite_add_tcase(s, baz_case_create()); suite_add_tcase(s, qux_case_create()); return s; } Test cases include the actual tests to be performed within each case. TCase *baz_case_create(void) { TCase *tc; tc = tcase_create("Baz"); tcase_add_test(tc, xxyz_test); tcase_add_test(tc, zzyx_test); return tc; } A test case may optionally have a fixture which is code that is executed before and after each test case. Unchecked fixtures are executed once before the test process forks for each test whereas checked fixtures are executed for each and every test. static void fixture_setup(void) { } static void fixture_teardown(void) { } TCase *qux_case_create(void) { TCase *tc; /* Matching entry tests */ tc = tcase_create("Match"); tcase_add_checked_fixture(tc, fixture_setup, fixture_teardown); tcase_add_test(tc, zzz_test); return tc; } Additionally test cases can contain tests executed in a loop. The test recives a single integer as a parameter named _i which iterates between values specified in the case setup. TCase *baz_case_create(void) { TCase *tc; tc = tcase_create("Baz"); tcase_add_loop_test(tc, looping_test, 0, 5); return tc; } It is also possible to create tests which will generate a signal. The most commonly used of these is to check asserts in API calls. TCase *baz_case_create(void) { TCase *tc; tc = tcase_create("Baz"); tcase_add_test_raise_signal(tc, assert_test, 6); return tc; } Actual test code is self contained in a function which uses the ck_assert macros to test results. The check framework requires each test to use the START_TEST and END_TEST macros when definig them. /** * url access leaf test */ START_TEST(nsurl_access_leaf_test) { nserror err; nsurl *res_url; const struct test_triplets *tst = &access_tests[_i]; /* not testing create, this should always succeed */ err = nsurl_create(tst->test1, &res_url); ck_assert(err == NSERROR_OK); ck_assert_str_eq(nsurl_access_leaf(res_url), tst->res); nsurl_unref(res_url); } END_TEST [1] https://libcheck.github.io/check/