#!/usr/bin/python # ex:ts=4:sw=4:sts=4:et # -*- tab-width: 4; c-basic-offset: 4; indent-tabs-mode: nil -*- # # Copyright (C) 2015 Alexandru Damian for Intel Corp. # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License version 2 as # published by the Free Software Foundation. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License along # with this program; if not, write to the Free Software Foundation, Inc., # 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. # This is the main test execution controller. It is designed to be run # manually from the command line, or to be called from a different program # that schedules test execution. # # Execute runner.py -h for help. from __future__ import print_function import sys, os import unittest, importlib import logging, pprint, json import re from shellutils import ShellCmdException, mkdirhier, run_shell_cmd import config # we also log to a file, in addition to console, because our output is important __log_file_name__ = os.path.join(os.path.dirname(__file__), "log/tts_%d.log" % config.OWN_PID) mkdirhier(os.path.dirname(__log_file_name__)) __log_file__ = open(__log_file_name__, "w") __file_handler__ = logging.StreamHandler(__log_file__) __file_handler__.setFormatter(logging.Formatter("%(asctime)s %(levelname)s: %(message)s")) config.logger.addHandler(__file_handler__) # set up log directory try: if not os.path.exists(config.LOGDIR): os.mkdir(config.LOGDIR) else: if not os.path.isdir(config.LOGDIR): raise Exception("Expected log dir '%s' is not actually a directory." % config.LOGDIR) except OSError as exc: raise exc # creates the under-test-branch as a separate directory def set_up_test_branch(settings, branch_name): testdir = "%s/%s.%d" % (settings['workdir'], config.TEST_DIR_NAME, config.OWN_PID) # creates the host dir if os.path.exists(testdir): raise Exception("Test dir '%s'is already there, aborting" % testdir) # may raise OSError, is to be handled by the caller os.makedirs(testdir) # copies over the .git from the localclone run_shell_cmd("cp -a '%s'/.git '%s'" % (settings['localclone'], testdir)) # add the remote if it doesn't exist crt_remotes = run_shell_cmd("git remote -v", cwd=testdir) remotes = [word for line in crt_remotes.split("\n") for word in line.split()] if not config.CONTRIB_REPO in remotes: remote_name = "tts_contrib" run_shell_cmd("git remote add %s %s" % (remote_name, config.CONTRIB_REPO), cwd=testdir) else: remote_name = remotes[remotes.index(config.CONTRIB_REPO) - 1] # do the fetch run_shell_cmd("git fetch %s -p" % remote_name, cwd=testdir) # do the checkout run_shell_cmd("git checkout origin/master && git branch -D %s; git checkout %s/%s -b %s && git reset --hard" % (branch_name, remote_name, branch_name, branch_name), cwd=testdir) return testdir def __search_for_tests(): # we find all classes that can run, and run them tests = [] for _, _, files_list in os.walk(os.path.dirname(os.path.abspath(__file__))): for module_file in [f[:-3] for f in files_list if f.endswith(".py") and not f.startswith("__init__")]: config.logger.debug("Inspecting module %s", module_file) current_module = importlib.import_module(module_file) crtclass_names = vars(current_module) for name in crtclass_names: tested_value = crtclass_names[name] if isinstance(tested_value, type(unittest.TestCase)) and issubclass(tested_value, unittest.TestCase): tests.append((module_file, name)) break return tests # boilerplate to self discover tests and run them def execute_tests(dir_under_test, testname): if testname is not None and "." in testname: tests = [] tests.append(tuple(testname.split(".", 2))) else: tests = __search_for_tests() # let's move to the directory under test crt_dir = os.getcwd() os.chdir(dir_under_test) # execute each module # pylint: disable=broad-except # we disable the broad-except because we want to actually catch all possible exceptions try: # sorting the tests by the numeric order in the class name tests = sorted(tests, key=lambda x: int(re.search(r"[0-9]+", x[1]).group(0))) config.logger.debug("Discovered test clases: %s", pprint.pformat(tests)) unittest.installHandler() suite = unittest.TestSuite() loader = unittest.TestLoader() result = unittest.TestResult() result.failfast = True for module_file, test_name in tests: suite.addTest(loader.loadTestsFromName("%s.%s" % (module_file, test_name))) config.logger.info("Running %d test(s)", suite.countTestCases()) suite.run(result) for error in result.errors: config.logger.error("Exception on test: %s\n%s", error[0], "\n".join(["-- %s" % x for x in error[1].split("\n")])) for failure in result.failures: config.logger.error("Failed test: %s:\n%s\n", failure[0], "\n".join(["-- %s" % x for x in failure[1].split("\n")])) config.logger.info("Test results: %d ran, %d errors, %d failures", result.testsRun, len(result.errors), len(result.failures)) except Exception as exc: import traceback config.logger.error("Exception while running test. Tracedump: \n%s", traceback.format_exc(exc)) finally: os.chdir(crt_dir) return len(result.failures) # verify that we had a branch-under-test name as parameter def validate_args(): from optparse import OptionParser parser = OptionParser(usage="usage: %prog [options] branch_under_test") parser.add_option("-t", "--test-dir", dest="testdir", default=None, help="Use specified directory to run tests, inhibits the checkout.") parser.add_option("-s", "--single", dest="singletest", default=None, help="Run only the specified test") (options, args) = parser.parse_args() if len(args) < 1: raise Exception("Please specify the branch to run on. Use option '-h' when in doubt.") return (options, args) # load the configuration options def read_settings(): if not os.path.exists(config.SETTINGS_FILE) or not os.path.isfile(config.SETTINGS_FILE): raise Exception("Config file '%s' cannot be openend" % config.SETTINGS_FILE) return json.loads(open(config.SETTINGS_FILE, "r").read()) # cleanup ! def clean_up(testdir): run_shell_cmd("rm -rf -- '%s'" % testdir) def dump_info(settings, options, args): """ detailed information about current run configuration, for debugging purposes. """ config.logger.debug("Settings:\n%s\nOptions:\n%s\nArguments:\n%s\n", settings, options, args) def main(): (options, args) = validate_args() settings = read_settings() need_cleanup = False # dump debug info dump_info(settings, options, args) testdir = None no_failures = 1 try: if options.testdir is not None and os.path.exists(options.testdir): testdir = os.path.abspath(options.testdir) config.logger.info("No checkout, using %s", testdir) else: need_cleanup = True testdir = set_up_test_branch(settings, args[0]) # we expect a branch name as first argument config.TESTDIR = testdir # we let tests know where to run # ensure that the test dir only contains no *.pyc leftovers run_shell_cmd("find '%s' -type f -name *.pyc -exec rm {} \\;" % testdir) no_failures = execute_tests(testdir, options.singletest) except ShellCmdException as exc: import traceback config.logger.error("Error while setting up testing. Traceback: \n%s", traceback.format_exc(exc)) finally: if need_cleanup and testdir is not None: clean_up(testdir) sys.exit(no_failures) if __name__ == "__main__": main()