#include #include #include #include #include #include #include #include #include #include #include #include static void check_tests_exist(struct manifest *m, unsigned int *timeleft, struct score *score); static struct ccanlint tests_exist = { .key = "tests_exist", .name = "Module has test directory with tests in it", .check = check_tests_exist, .needs = "info_exists" }; REGISTER_TEST(tests_exist); static void handle_no_tests(struct manifest *m, struct score *score) { FILE *run; struct ccan_file *i; char *test_dir = tal_fmt(m, "%s/test", m->dir), *run_file; printf( "CCAN modules have a directory called test/ which contains tests.\n" "There are four kinds of tests: api, run, compile_ok and compile_fail:\n" "you can tell which type of test a C file is by its name, eg 'run.c'\n" "and 'run-simple.c' are both run tests.\n\n" "The simplest kind of test is a run test, which must compile with no\n" "warnings, and then run: it is expected to use ccan/tap to report its\n" "results in a simple and portable format. It should #include the C\n" "files from the module directly (so it can probe the internals): the\n" "module will not be linked in. The test will be run in a temporary\n" "directory, with the test directory symlinked under test/.\n\n" "api tests are just like a run test, except it is a guarantee of API\n" "stability: this test should pass on all future versions of the\n" "module. They *are* linked to the module, since they should only\n" "test the API, not the internal state.\n\n" "compile_ok tests are a subset of run tests: they must compile and\n" "link, but aren't run.\n\n" "compile_fail tests are tests which should fail to compile (or emit\n" "warnings) or link when FAIL is defined, but should compile and link\n" "when it's not defined: this helps ensure unrelated errors don't make\n" "compilation fail.\n\n" "Note that only API tests are linked against the files in the module!\n" ); if (!ask("Should I create a template test/run.c file for you?")) return; if (mkdir(test_dir, 0700) != 0) { if (errno != EEXIST) err(1, "Creating test/ directory"); } run_file = tal_fmt(test_dir, "%s/run.c", test_dir); run = fopen(run_file, "w"); if (!run) err(1, "Trying to create a test/run.c"); fprintf(run, "#include \n", m->modname, m->basename); if (!list_empty(&m->c_files)) { fputs("/* Include the C files directly. */\n", run); list_for_each(&m->c_files, i, list) fprintf(run, "#include \n", m->modname, i->name); } fprintf(run, "%s", "#include \n\n" "int main(void)\n" "{\n" " /* This is how many tests you plan to run */\n" " plan_tests(3);\n" "\n" " /* Simple thing we expect to succeed */\n" " ok1(some_test())\n" " /* Same, with an explicit description of the test. */\n" " ok(some_test(), \"%s with no args should return 1\", \"some_test\")\n" " /* How to print out messages for debugging. */\n" " diag(\"Address of some_test is %p\", &some_test)\n" " /* Conditional tests must be explicitly skipped. */\n" "#if HAVE_SOME_FEATURE\n" " ok1(test_some_feature())\n" "#else\n" " skip(1, \"Don\'t have SOME_FEATURE\")\n" "#endif\n" "\n" " /* This exits depending on whether all tests passed */\n" " return exit_status();\n" "}\n"); fclose(run); } static void check_tests_exist(struct manifest *m, unsigned int *timeleft, struct score *score) { struct stat st; char *test_dir = path_join(m, m->dir, "test"); if (lstat(test_dir, &st) != 0) { score->error = tal_strdup(score, "No test directory"); if (errno != ENOENT) err(1, "statting %s", test_dir); tests_exist.handle = handle_no_tests; /* We "pass" this. */ score->pass = true; return; } if (!S_ISDIR(st.st_mode)) { score->error = tal_strdup(score, "test is not a directory"); return; } if (list_empty(&m->api_tests) && list_empty(&m->run_tests) && list_empty(&m->compile_ok_tests) && list_empty(&m->compile_fail_tests)) { score->error = tal_strdup(score, "No tests in test directory"); tests_exist.handle = handle_no_tests; return; } score->pass = true; score->score = score->total; }