def main(): # Setup command line options argparser = argparse.ArgumentParser() output_options = argparser.add_argument_group('Options controlling output') locate_options = argparser.add_argument_group( 'Options for locating checks') select_options = argparser.add_argument_group( 'Options for selecting checks') action_options = argparser.add_argument_group( 'Options controlling actions') run_options = argparser.add_argument_group( 'Options controlling execution of checks') env_options = argparser.add_argument_group( 'Options controlling environment') misc_options = argparser.add_argument_group('Miscellaneous options') # Output directory options output_options.add_argument('--prefix', action='store', metavar='DIR', help='Set output directory prefix to DIR') output_options.add_argument('-o', '--output', action='store', metavar='DIR', help='Set output directory to DIR') output_options.add_argument('-s', '--stage', action='store', metavar='DIR', help='Set stage directory to DIR') output_options.add_argument( '--perflogdir', action='store', metavar='DIR', help='Set directory prefix for the performance logs ' '(default: ${prefix}/perflogs, ' 'relevant only if the filelog backend is used)') output_options.add_argument( '--keep-stage-files', action='store_true', help='Keep stage directory even if check is successful') output_options.add_argument( '--save-log-files', action='store_true', default=False, help='Copy the log file from the work dir to the output dir at the ' 'end of the program') # Check discovery options locate_options.add_argument('-c', '--checkpath', action='append', metavar='DIR|FILE', help='Search for checks in DIR or FILE') locate_options.add_argument('-R', '--recursive', action='store_true', help='Load checks recursively') locate_options.add_argument('--ignore-check-conflicts', action='store_true', help='Skip checks with conflicting names') # Select options select_options.add_argument('-t', '--tag', action='append', dest='tags', default=[], help='Select checks matching TAG') select_options.add_argument('-n', '--name', action='append', dest='names', default=[], metavar='NAME', help='Select checks with NAME') select_options.add_argument('-x', '--exclude', action='append', dest='exclude_names', metavar='NAME', default=[], help='Exclude checks with NAME') select_options.add_argument( '-p', '--prgenv', action='append', default=[r'.*'], help='Select tests for PRGENV programming environment only') select_options.add_argument('--gpu-only', action='store_true', help='Select only GPU tests') select_options.add_argument('--cpu-only', action='store_true', help='Select only CPU tests') # Action options action_options.add_argument('-l', '--list', action='store_true', help='List matched regression checks') action_options.add_argument( '-L', '--list-detailed', action='store_true', help='List matched regression checks with a detailed description') action_options.add_argument('-r', '--run', action='store_true', help='Run regression with the selected checks') # Run options run_options.add_argument('-A', '--account', action='store', help='Use ACCOUNT for submitting jobs') run_options.add_argument('-P', '--partition', action='store', metavar='PART', help='Use PART for submitting jobs') run_options.add_argument('--reservation', action='store', metavar='RES', help='Use RES for submitting jobs') run_options.add_argument('--nodelist', action='store', help='Run checks on the selected list of nodes') run_options.add_argument( '--exclude-nodes', action='store', metavar='NODELIST', help='Exclude the list of nodes from running checks') run_options.add_argument('--job-option', action='append', metavar='OPT', dest='job_options', default=[], help='Pass OPT to job scheduler') run_options.add_argument('--force-local', action='store_true', help='Force local execution of checks') run_options.add_argument('--skip-sanity-check', action='store_true', help='Skip sanity checking') run_options.add_argument('--skip-performance-check', action='store_true', help='Skip performance checking') run_options.add_argument('--strict', action='store_true', help='Force strict performance checking') run_options.add_argument('--skip-system-check', action='store_true', help='Skip system check') run_options.add_argument('--skip-prgenv-check', action='store_true', help='Skip prog. environment check') run_options.add_argument( '--exec-policy', metavar='POLICY', action='store', choices=['serial', 'async'], default='serial', help='Specify the execution policy for running the regression tests. ' 'Available policies: "serial" (default), "async"') run_options.add_argument('--mode', action='store', help='Execution mode to use') run_options.add_argument( '--max-retries', metavar='NUM', action='store', default=0, help='Specify the maximum number of times a failed regression test ' 'may be retried (default: 0)') run_options.add_argument( '--flex-alloc-tasks', action='store', dest='flex_alloc_tasks', metavar='{all|idle|NUM}', default='idle', help="Strategy for flexible task allocation (default: 'idle').") env_options.add_argument('-M', '--map-module', action='append', metavar='MAPPING', dest='module_mappings', default=[], help='Apply a single module mapping') env_options.add_argument( '-m', '--module', action='append', default=[], metavar='MOD', dest='user_modules', help='Load module MOD before running the regression suite') env_options.add_argument('--module-mappings', action='store', metavar='FILE', dest='module_map_file', help='Apply module mappings defined in FILE') env_options.add_argument( '-u', '--unload-module', action='append', metavar='MOD', dest='unload_modules', default=[], help='Unload module MOD before running the regression suite') env_options.add_argument( '--purge-env', action='store_true', dest='purge_env', default=False, help='Purge environment before running the regression suite') # Miscellaneous options misc_options.add_argument( '-C', '--config-file', action='store', dest='config_file', metavar='FILE', default=os.path.join(reframe.INSTALL_PREFIX, 'reframe/settings.py'), help='Specify a custom config-file for the machine. ' '(default: %s' % os.path.join(reframe.INSTALL_PREFIX, 'reframe/settings.py')) misc_options.add_argument('--nocolor', action='store_false', dest='colorize', default=True, help='Disable coloring of output') misc_options.add_argument('--performance-report', action='store_true', help='Print the performance report') # FIXME: This should move to env_options as soon as # https://github.com/eth-cscs/reframe/pull/946 is merged misc_options.add_argument('--non-default-craype', action='store_true', default=False, help='Test a non-default Cray PE') misc_options.add_argument( '--show-config', action='store_true', help='Print configuration of the current system and exit') misc_options.add_argument( '--show-config-env', action='store', metavar='ENV', help='Print configuration of environment ENV and exit') misc_options.add_argument('--system', action='store', help='Load SYSTEM configuration explicitly') misc_options.add_argument( '--timestamp', action='store', nargs='?', const='%FT%T', metavar='TIMEFMT', help='Append a timestamp component to the regression directories' '(default format "%%FT%%T")') misc_options.add_argument('-V', '--version', action='version', version=reframe.VERSION) misc_options.add_argument('-v', '--verbose', action='count', default=0, help='Increase verbosity level of output') if len(sys.argv) == 1: argparser.print_help() sys.exit(1) # Parse command line options = argparser.parse_args() # Load configuration try: settings = config.load_settings_from_file(options.config_file) except (OSError, ReframeError) as e: sys.stderr.write('%s: could not load settings: %s\n' % (sys.argv[0], e)) sys.exit(1) # Configure logging try: logging.configure_logging(settings.logging_config), except (OSError, ConfigError) as e: sys.stderr.write('could not configure logging: %s\n' % e) sys.exit(1) # Set colors in logger logging.getlogger().colorize = options.colorize # Setup printer printer = PrettyPrinter() printer.colorize = options.colorize if options.verbose: printer.inc_verbosity(options.verbose) try: runtime.init_runtime(settings.site_configuration, options.system, non_default_craype=options.non_default_craype) except SystemAutodetectionError: printer.warning( 'could not find a configuration entry for the current system; ' 'falling back to a generic system configuration; ' 'please check the online documentation on how to configure ' 'ReFrame for your system.') settings.site_configuration['systems'] = { 'generic': { 'descr': 'Generic fallback system configuration', 'hostnames': ['localhost'], 'partitions': { 'login': { 'scheduler': 'local', 'environs': ['builtin-gcc'], 'descr': 'Login nodes' } } } } settings.site_configuration['environments'] = { '*': { 'builtin-gcc': { 'type': 'ProgEnvironment', 'cc': 'gcc', 'cxx': 'g++', 'ftn': 'gfortran', } } } runtime.init_runtime(settings.site_configuration, 'generic', non_default_craype=options.non_default_craype) except Exception as e: printer.error('configuration error: %s' % e) printer.verbose(''.join(traceback.format_exception(*sys.exc_info()))) sys.exit(1) rt = runtime.runtime() try: if options.module_map_file: rt.modules_system.load_mapping_from_file(options.module_map_file) if options.module_mappings: for m in options.module_mappings: rt.modules_system.load_mapping(m) except (ConfigError, OSError) as e: printer.error('could not load module mappings: %s' % e) sys.exit(1) if options.mode: try: mode_args = rt.mode(options.mode) # Parse the mode's options and reparse the command-line options = argparser.parse_args(mode_args) options = argparser.parse_args(namespace=options) except ConfigError as e: printer.error('could not obtain execution mode: %s' % e) sys.exit(1) # Adjust system directories if options.prefix: # if prefix is set, reset all other directories rt.resources.prefix = os_ext.expandvars(options.prefix) rt.resources.outputdir = None rt.resources.stagedir = None if options.output: rt.resources.outputdir = os_ext.expandvars(options.output) if options.stage: rt.resources.stagedir = os_ext.expandvars(options.stage) if (os_ext.samefile(rt.resources.stage_prefix, rt.resources.output_prefix) and not options.keep_stage_files): printer.error('stage and output refer to the same directory; ' 'if this is on purpose, please use also the ' "`--keep-stage-files' option.") sys.exit(1) if options.timestamp: rt.resources.timefmt = options.timestamp # Configure performance logging # NOTE: we need resources to be configured in order to set the global # perf. logging prefix correctly if options.perflogdir: rt.resources.perflogdir = os_ext.expandvars(options.perflogdir) logging.LOG_CONFIG_OPTS['handlers.filelog.prefix'] = ( rt.resources.perflog_prefix) # Show configuration after everything is set up if options.show_config: printer.info(rt.show_config()) sys.exit(0) if options.show_config_env: envname = options.show_config_env for p in rt.system.partitions: environ = p.environment(envname) if environ: break if environ is None: printer.error('no such environment: ' + envname) sys.exit(1) printer.info(environ.details()) sys.exit(0) if hasattr(settings, 'perf_logging_config'): try: logging.configure_perflogging(settings.perf_logging_config) except (OSError, ConfigError) as e: printer.error('could not configure performance logging: %s\n' % e) sys.exit(1) else: printer.warning('no performance logging is configured; ' 'please check documentation') # Setup the check loader if options.checkpath: load_path = [] for d in options.checkpath: d = os_ext.expandvars(d) if not os.path.exists(d): printer.warning("%s: path `%s' does not exist. Skipping..." % (argparser.prog, d)) continue load_path.append(d) loader = RegressionCheckLoader( load_path, recurse=options.recursive, ignore_conflicts=options.ignore_check_conflicts) else: loader = RegressionCheckLoader(load_path=settings.checks_path, prefix=reframe.INSTALL_PREFIX, recurse=settings.checks_path_recurse) printer.debug(argparse.format_options(options)) # Print command line printer.info('Command line: %s' % ' '.join(sys.argv)) printer.info('Reframe version: ' + reframe.VERSION) printer.info('Launched by user: '******'<unknown>')) printer.info('Launched on host: ' + socket.gethostname()) # Print important paths printer.info('Reframe paths') printer.info('=============') printer.info(' Check prefix : %s' % loader.prefix) printer.info( '%03s Check search path : %s' % ('(R)' if loader.recurse else '', "'%s'" % ':'.join(loader.load_path))) printer.info(' Stage dir prefix : %s' % rt.resources.stage_prefix) printer.info(' Output dir prefix : %s' % rt.resources.output_prefix) printer.info( ' Perf. logging prefix : %s' % os.path.abspath(logging.LOG_CONFIG_OPTS['handlers.filelog.prefix'])) try: # Locate and load checks try: checks_found = loader.load_all() except OSError as e: raise ReframeError from e # Filter checks by name checks_matched = checks_found if options.exclude_names: for name in options.exclude_names: checks_matched = filter(filters.have_not_name(name), checks_matched) if options.names: checks_matched = filter(filters.have_name('|'.join(options.names)), checks_matched) # Filter checks by tags for tag in options.tags: checks_matched = filter(filters.have_tag(tag), checks_matched) # Filter checks by prgenv if not options.skip_prgenv_check: for prgenv in options.prgenv: checks_matched = filter(filters.have_prgenv(prgenv), checks_matched) # Filter checks by system if not options.skip_system_check: checks_matched = filter( filters.have_partition(rt.system.partitions), checks_matched) # Filter checks further if options.gpu_only and options.cpu_only: printer.error("options `--gpu-only' and `--cpu-only' " "are mutually exclusive") sys.exit(1) if options.gpu_only: checks_matched = filter(filters.have_gpu_only(), checks_matched) elif options.cpu_only: checks_matched = filter(filters.have_cpu_only(), checks_matched) # Determine the allowed programming environments allowed_environs = { e.name for env_patt in options.prgenv for p in rt.system.partitions for e in p.environs if re.match(env_patt, e.name) } # Generate the test cases, validate dependencies and sort them checks_matched = list(checks_matched) testcases = generate_testcases(checks_matched, options.skip_system_check, options.skip_prgenv_check, allowed_environs) testgraph = dependency.build_deps(testcases) dependency.validate_deps(testgraph) testcases = dependency.toposort(testgraph) # Unload regression's module and load user-specified modules if hasattr(settings, 'reframe_module'): printer.warning( "the 'reframe_module' configuration option will be ignored; " "please use the '-u' or '--unload-module' options") if options.purge_env: rt.modules_system.unload_all() else: for m in options.unload_modules: rt.modules_system.unload_module(m) # Load the environment for the current system try: env.load(rt.system.preload_environ) except EnvironError as e: printer.error("failed to load current system's environment; " "please check your configuration") printer.debug(str(e)) raise for m in options.user_modules: try: rt.modules_system.load_module(m, force=True) except EnvironError as e: printer.warning("could not load module '%s' correctly: " "Skipping..." % m) printer.debug(str(e)) # Act on checks success = True if options.list: # List matched checks list_checks(list(checks_matched), printer) elif options.list_detailed: # List matched checks with details list_checks(list(checks_matched), printer, detailed=True) elif options.run: # Setup the execution policy if options.exec_policy == 'serial': exec_policy = SerialExecutionPolicy() elif options.exec_policy == 'async': exec_policy = AsynchronousExecutionPolicy() else: # This should not happen, since choices are handled by # argparser printer.error("unknown execution policy `%s': Exiting...") sys.exit(1) exec_policy.skip_system_check = options.skip_system_check exec_policy.force_local = options.force_local exec_policy.strict_check = options.strict exec_policy.skip_sanity_check = options.skip_sanity_check exec_policy.skip_performance_check = options.skip_performance_check exec_policy.keep_stage_files = options.keep_stage_files try: errmsg = "invalid option for --flex-alloc-tasks: '{0}'" sched_flex_alloc_tasks = int(options.flex_alloc_tasks) if sched_flex_alloc_tasks <= 0: raise ConfigError(errmsg.format(options.flex_alloc_tasks)) except ValueError: if not options.flex_alloc_tasks.lower() in {'idle', 'all'}: raise ConfigError(errmsg.format( options.flex_alloc_tasks)) from None sched_flex_alloc_tasks = options.flex_alloc_tasks exec_policy.sched_flex_alloc_tasks = sched_flex_alloc_tasks exec_policy.flex_alloc_tasks = options.flex_alloc_tasks exec_policy.sched_account = options.account exec_policy.sched_partition = options.partition exec_policy.sched_reservation = options.reservation exec_policy.sched_nodelist = options.nodelist exec_policy.sched_exclude_nodelist = options.exclude_nodes exec_policy.sched_options = options.job_options try: max_retries = int(options.max_retries) except ValueError: raise ConfigError('--max-retries is not a valid integer: %s' % max_retries) from None runner = Runner(exec_policy, printer, max_retries) try: runner.runall(testcases) finally: # Print a retry report if we did any retries if runner.stats.failures(run=0): printer.info(runner.stats.retry_report()) # Print a failure report if we had failures in the last run if runner.stats.failures(): printer.info(runner.stats.failure_report()) success = False if options.performance_report: printer.info(runner.stats.performance_report()) else: printer.info('No action specified. Exiting...') printer.info("Try `%s -h' for a list of available actions." % argparser.prog) sys.exit(1) if not success: sys.exit(1) sys.exit(0) except KeyboardInterrupt: sys.exit(1) except ReframeError as e: printer.error(str(e)) sys.exit(1) except (Exception, ReframeFatalError): printer.error(format_exception(*sys.exc_info())) sys.exit(1) finally: try: if options.save_log_files: logging.save_log_files(rt.resources.output_prefix) except OSError as e: printer.error('could not save log file: %s' % e) sys.exit(1)
def main(): # Setup command line options argparser = argparse.ArgumentParser() output_options = argparser.add_argument_group('Options controlling output') locate_options = argparser.add_argument_group( 'Options for locating checks') select_options = argparser.add_argument_group( 'Options for selecting checks') action_options = argparser.add_argument_group( 'Options controlling actions') run_options = argparser.add_argument_group( 'Options controlling execution of checks') env_options = argparser.add_argument_group( 'Options controlling environment') misc_options = argparser.add_argument_group('Miscellaneous options') # Output directory options output_options.add_argument('--prefix', action='store', metavar='DIR', help='Set output directory prefix to DIR', envvar='RFM_PREFIX', configvar='systems/prefix') output_options.add_argument('-o', '--output', action='store', metavar='DIR', help='Set output directory to DIR', envvar='RFM_OUTPUT_DIR', configvar='systems/outputdir') output_options.add_argument('-s', '--stage', action='store', metavar='DIR', help='Set stage directory to DIR', envvar='RFM_STAGE_DIR', configvar='systems/stagedir') output_options.add_argument( '--perflogdir', action='store', metavar='DIR', help=('Set directory prefix for the performance logs ' '(default: ${prefix}/perflogs, ' 'relevant only if the filelog backend is used)'), envvar='RFM_PERFLOG_DIR', configvar='logging/handlers_perflog/filelog_basedir') output_options.add_argument( '--keep-stage-files', action='store_true', help='Keep stage directory even if check is successful', envvar='RFM_KEEP_STAGE_FILES', configvar='general/keep_stage_files') output_options.add_argument( '--save-log-files', action='store_true', default=False, help=('Copy the log file from the current directory to the ' 'output directory when ReFrame ends'), envvar='RFM_SAVE_LOG_FILES', configvar='general/save_log_files') # Check discovery options locate_options.add_argument( '-c', '--checkpath', action='append', metavar='DIR|FILE', help="Add DIR or FILE to the check search path", envvar='RFM_CHECK_SEARCH_PATH :', configvar='general/check_search_path') locate_options.add_argument('-R', '--recursive', action='store_true', help='Load checks recursively', envvar='RFM_CHECK_SEARCH_RECURSIVE', configvar='general/check_search_recursive') locate_options.add_argument('--ignore-check-conflicts', action='store_true', help='Skip checks with conflicting names', envvar='RFM_IGNORE_CHECK_CONFLICTS', configvar='general/ignore_check_conflicts') # Select options select_options.add_argument('-t', '--tag', action='append', dest='tags', default=[], help='Select checks matching TAG') select_options.add_argument('-n', '--name', action='append', dest='names', default=[], metavar='NAME', help='Select checks with NAME') select_options.add_argument('-x', '--exclude', action='append', dest='exclude_names', metavar='NAME', default=[], help='Exclude checks with NAME') select_options.add_argument( '-p', '--prgenv', action='append', default=[r'.*'], help='Select tests for PRGENV programming environment only') select_options.add_argument('--gpu-only', action='store_true', help='Select only GPU tests') select_options.add_argument('--cpu-only', action='store_true', help='Select only CPU tests') # Action options action_options.add_argument('-l', '--list', action='store_true', help='List matched regression checks') action_options.add_argument( '-L', '--list-detailed', action='store_true', help='List matched regression checks with a detailed description') action_options.add_argument('-r', '--run', action='store_true', help='Run regression with the selected checks') # Run options run_options.add_argument('-A', '--account', action='store', help='Use ACCOUNT for submitting jobs') run_options.add_argument('-P', '--partition', action='store', metavar='PART', help='Use PART for submitting jobs') run_options.add_argument('--reservation', action='store', metavar='RES', help='Use RES for submitting jobs') run_options.add_argument('--nodelist', action='store', help='Run checks on the selected list of nodes') run_options.add_argument( '--exclude-nodes', action='store', metavar='NODELIST', help='Exclude the list of nodes from running checks') run_options.add_argument('--job-option', action='append', metavar='OPT', dest='job_options', default=[], help='Pass OPT to job scheduler') run_options.add_argument('--force-local', action='store_true', help='Force local execution of checks') run_options.add_argument('--skip-sanity-check', action='store_true', help='Skip sanity checking') run_options.add_argument('--skip-performance-check', action='store_true', help='Skip performance checking') run_options.add_argument('--strict', action='store_true', help='Force strict performance checking') run_options.add_argument('--skip-system-check', action='store_true', help='Skip system check') run_options.add_argument('--skip-prgenv-check', action='store_true', help='Skip prog. environment check') run_options.add_argument( '--exec-policy', metavar='POLICY', action='store', choices=['async', 'serial'], default='async', help='Specify the execution policy for running the regression tests. ' 'Available policies: "async" (default), "serial"') run_options.add_argument('--mode', action='store', help='Execution mode to use') run_options.add_argument( '--max-retries', metavar='NUM', action='store', default=0, help='Specify the maximum number of times a failed regression test ' 'may be retried (default: 0)') run_options.add_argument( '--flex-alloc-tasks', action='store', dest='flex_alloc_tasks', metavar='{all|idle|NUM}', default=None, help='*deprecated*, please use --flex-alloc-nodes instead') run_options.add_argument( '--flex-alloc-nodes', action='store', dest='flex_alloc_nodes', metavar='{all|idle|NUM}', default=None, help="Strategy for flexible node allocation (default: 'idle').") env_options.add_argument('-M', '--map-module', action='append', metavar='MAPPING', dest='module_mappings', default=[], help='Apply a single module mapping', envvar='RFM_MODULE_MAPPINGS ,', configvar='general/module_mappings') env_options.add_argument( '-m', '--module', action='append', default=[], metavar='MOD', dest='user_modules', help='Load module MOD before running the regression suite', envvar='RFM_USER_MODULES', configvar='general/user_modules') env_options.add_argument('--module-mappings', action='store', metavar='FILE', dest='module_map_file', help='Apply module mappings defined in FILE', envvar='RFM_MODULE_MAP_FILE', configvar='general/module_map_file') env_options.add_argument( '-u', '--unload-module', action='append', metavar='MOD', dest='unload_modules', default=[], help='Unload module MOD before running the regression suite', envvar='RFM_UNLOAD_MODULES', configvar='general/unload_modules') env_options.add_argument( '--purge-env', action='store_true', dest='purge_env', default=False, help='Purge environment before running the regression suite', envvar='RFM_PURGE_ENVIRONMENT', configvar='general/purge_environment') env_options.add_argument('--non-default-craype', action='store_true', help='Test a non-default Cray PE', envvar='RFM_NON_DEFAULT_CRAYPE', configvar='general/non_default_craype') # Miscellaneous options misc_options.add_argument('-C', '--config-file', action='store', dest='config_file', metavar='FILE', help='ReFrame configuration file to use', envvar='RFM_CONFIG_FILE') misc_options.add_argument('--nocolor', action='store_false', dest='colorize', help='Disable coloring of output', envvar='RFM_COLORIZE', configvar='general/colorize') misc_options.add_argument('--failure-stats', action='store_true', help='Print failure statistics') misc_options.add_argument('--performance-report', action='store_true', help='Print a report for performance tests run') misc_options.add_argument('--show-config-param', action='store', nargs='?', const='all', metavar='PARAM', help=('Print how parameter PARAM is configured ' 'for the current system and exit')) misc_options.add_argument('--system', action='store', help='Load configuration for SYSTEM', envvar='RFM_SYSTEM') misc_options.add_argument( '--timestamp', action='store', nargs='?', const='', metavar='TIMEFMT', help=('Append a timestamp component to the various ' 'ReFrame directories (default format: "%%FT%%T")'), envvar='RFM_TIMESTAMP_DIRS', configvar='general/timestamp_dirs') misc_options.add_argument('-V', '--version', action='version', version=os_ext.reframe_version()) misc_options.add_argument('-v', '--verbose', action='count', help='Increase verbosity level of output', envvar='RFM_VERBOSE', configvar='general/verbose') # Options not associated with command-line arguments argparser.add_argument( dest='graylog_server', envvar='RFM_GRAYLOG_SERVER', configvar='logging/handlers_perflog/graylog_address', help='Graylog server address') if len(sys.argv) == 1: argparser.print_help() sys.exit(1) # Parse command line options = argparser.parse_args() # First configure logging with our generic configuration so as to be able # to print pretty messages; logging will be reconfigured by user's # configuration later site_config = config.load_config( os.path.join(reframe.INSTALL_PREFIX, 'reframe/core/settings.py')) site_config.select_subconfig('generic') options.update_config(site_config) logging.configure_logging(site_config) logging.getlogger().colorize = site_config.get('general/0/colorize') printer = PrettyPrinter() printer.colorize = site_config.get('general/0/colorize') if options.verbose: printer.inc_verbosity(options.verbose) # Now configure ReFrame according to the user configuration file try: try: site_config = config.load_config(options.config_file) except ReframeDeprecationWarning as e: printer.warning(e) converted = config.convert_old_config(options.config_file) printer.warning(f"configuration file has been converted " f"to the new syntax here: '{converted}'") site_config = config.load_config(converted) site_config.validate() site_config.select_subconfig(options.system) for err in options.update_config(site_config): printer.warning(str(err)) logging.configure_logging(site_config) except (OSError, ConfigError) as e: printer.error(f'failed to load configuration: {e}') sys.exit(1) logging.getlogger().colorize = site_config.get('general/0/colorize') printer.colorize = site_config.get('general/0/colorize') try: runtime.init_runtime(site_config) except ConfigError as e: printer.error(f'failed to initialize runtime: {e}') sys.exit(1) rt = runtime.runtime() try: if site_config.get('general/0/module_map_file'): rt.modules_system.load_mapping_from_file( site_config.get('general/0/module_map_file')) if site_config.get('general/0/module_mappings'): for m in site_config.get('general/0/module_mappings'): rt.modules_system.load_mapping(m) except (ConfigError, OSError) as e: printer.error('could not load module mappings: %s' % e) sys.exit(1) if options.mode: try: mode_args = rt.get_option(f'modes/@{options.mode}/options') # Parse the mode's options and reparse the command-line options = argparser.parse_args(mode_args) options = argparser.parse_args(namespace=options.cmd_options) options.update_config(rt.site_config) except ConfigError as e: printer.error('could not obtain execution mode: %s' % e) sys.exit(1) if (os_ext.samefile(rt.stage_prefix, rt.output_prefix) and not site_config.get('general/0/keep_stage_files')): printer.error("stage and output refer to the same directory; " "if this is on purpose, please use the " "'--keep-stage-files' option.") sys.exit(1) # Show configuration after everything is set up if options.show_config_param: config_param = options.show_config_param if config_param == 'all': printer.info(str(rt.site_config)) else: value = rt.get_option(config_param) if value is None: printer.error( f'no such configuration parameter found: {config_param}') else: printer.info(json.dumps(value, indent=2)) sys.exit(0) # Setup the check loader loader = RegressionCheckLoader( load_path=site_config.get('general/0/check_search_path'), recurse=site_config.get('general/0/check_search_recursive'), ignore_conflicts=site_config.get('general/0/ignore_check_conflicts')) printer.debug(argparse.format_options(options)) def print_infoline(param, value): param = param + ':' printer.info(f" {param.ljust(18)} {value}") # Print command line printer.info(f"[ReFrame Setup]") print_infoline('version', os_ext.reframe_version()) print_infoline('command', repr(' '.join(sys.argv))) print_infoline('launched by', f"{os_ext.osuser() or '<unknown>'}@{socket.gethostname()}") print_infoline('working directory', repr(os.getcwd())) print_infoline( 'check search path', f"{'(R)' if loader.recurse else ''} " f"{':'.join(loader.load_path)!r}") print_infoline('stage directory', repr(rt.stage_prefix)) print_infoline('output directory', repr(rt.output_prefix)) printer.info('') try: # Locate and load checks try: checks_found = loader.load_all() except OSError as e: raise ReframeError from e # Filter checks by name checks_matched = checks_found if options.exclude_names: for name in options.exclude_names: checks_matched = filter(filters.have_not_name(name), checks_matched) if options.names: checks_matched = filter(filters.have_name('|'.join(options.names)), checks_matched) # Filter checks by tags for tag in options.tags: checks_matched = filter(filters.have_tag(tag), checks_matched) # Filter checks by prgenv if not options.skip_prgenv_check: for prgenv in options.prgenv: checks_matched = filter(filters.have_prgenv(prgenv), checks_matched) # Filter checks by system if not options.skip_system_check: checks_matched = filter( filters.have_partition(rt.system.partitions), checks_matched) # Filter checks further if options.gpu_only and options.cpu_only: printer.error("options `--gpu-only' and `--cpu-only' " "are mutually exclusive") sys.exit(1) if options.gpu_only: checks_matched = filter(filters.have_gpu_only(), checks_matched) elif options.cpu_only: checks_matched = filter(filters.have_cpu_only(), checks_matched) # Determine the allowed programming environments allowed_environs = { e.name for env_patt in options.prgenv for p in rt.system.partitions for e in p.environs if re.match(env_patt, e.name) } # Generate the test cases, validate dependencies and sort them checks_matched = list(checks_matched) testcases = generate_testcases(checks_matched, options.skip_system_check, options.skip_prgenv_check, allowed_environs) testgraph = dependency.build_deps(testcases) dependency.validate_deps(testgraph) testcases = dependency.toposort(testgraph) # Manipulate ReFrame's environment if site_config.get('general/0/purge_environment'): rt.modules_system.unload_all() else: for m in site_config.get('general/0/unload_modules'): rt.modules_system.unload_module(m) # Load the environment for the current system try: runtime.loadenv(rt.system.preload_environ) except EnvironError as e: printer.error("failed to load current system's environment; " "please check your configuration") printer.debug(str(e)) raise for m in site_config.get('general/0/user_modules'): try: rt.modules_system.load_module(m, force=True) except EnvironError as e: printer.warning("could not load module '%s' correctly: " "Skipping..." % m) printer.debug(str(e)) if options.flex_alloc_tasks: printer.warning("`--flex-alloc-tasks' is deprecated and " "will be removed in the future; " "you should use --flex-alloc-nodes instead") options.flex_alloc_nodes = (options.flex_alloc_nodes or options.flex_alloc_tasks) options.flex_alloc_nodes = options.flex_alloc_nodes or 'idle' # Act on checks success = True if options.list: # List matched checks list_checks(list(checks_matched), printer) elif options.list_detailed: # List matched checks with details list_checks(list(checks_matched), printer, detailed=True) elif options.run: # Setup the execution policy if options.exec_policy == 'serial': exec_policy = SerialExecutionPolicy() elif options.exec_policy == 'async': exec_policy = AsynchronousExecutionPolicy() else: # This should not happen, since choices are handled by # argparser printer.error("unknown execution policy `%s': Exiting...") sys.exit(1) exec_policy.skip_system_check = options.skip_system_check exec_policy.force_local = options.force_local exec_policy.strict_check = options.strict exec_policy.skip_sanity_check = options.skip_sanity_check exec_policy.skip_performance_check = options.skip_performance_check exec_policy.keep_stage_files = site_config.get( 'general/0/keep_stage_files') try: errmsg = "invalid option for --flex-alloc-nodes: '{0}'" sched_flex_alloc_nodes = int(options.flex_alloc_nodes) if sched_flex_alloc_nodes <= 0: raise ConfigError(errmsg.format(options.flex_alloc_nodes)) except ValueError: if not options.flex_alloc_nodes.casefold() in {'idle', 'all'}: raise ConfigError(errmsg.format( options.flex_alloc_nodes)) from None sched_flex_alloc_nodes = options.flex_alloc_nodes exec_policy.sched_flex_alloc_nodes = sched_flex_alloc_nodes exec_policy.flex_alloc_nodes = options.flex_alloc_nodes exec_policy.sched_account = options.account exec_policy.sched_partition = options.partition exec_policy.sched_reservation = options.reservation exec_policy.sched_nodelist = options.nodelist exec_policy.sched_exclude_nodelist = options.exclude_nodes exec_policy.sched_options = options.job_options try: max_retries = int(options.max_retries) except ValueError: raise ConfigError('--max-retries is not a valid integer: %s' % max_retries) from None runner = Runner(exec_policy, printer, max_retries) try: runner.runall(testcases) finally: # Print a retry report if we did any retries if runner.stats.failures(run=0): printer.info(runner.stats.retry_report()) # Print a failure report if we had failures in the last run if runner.stats.failures(): printer.info(runner.stats.failure_report()) success = False if options.failure_stats: printer.info(runner.stats.failure_stats()) if options.performance_report: printer.info(runner.stats.performance_report()) else: printer.error("No action specified. Please specify `-l'/`-L' for " "listing or `-r' for running. " "Try `%s -h' for more options." % argparser.prog) sys.exit(1) if not success: sys.exit(1) sys.exit(0) except KeyboardInterrupt: sys.exit(1) except ReframeError as e: printer.error(str(e)) sys.exit(1) except (Exception, ReframeFatalError): printer.error(format_exception(*sys.exc_info())) sys.exit(1) finally: try: if site_config.get('general/0/save_log_files'): logging.save_log_files(rt.output_prefix) except OSError as e: printer.error('could not save log file: %s' % e) sys.exit(1)
def main(): # Setup command line options argparser = argparse.ArgumentParser() output_options = argparser.add_argument_group( 'Options controlling ReFrame output') locate_options = argparser.add_argument_group( 'Options for discovering checks') select_options = argparser.add_argument_group( 'Options for selecting checks') action_options = argparser.add_argument_group( 'Options controlling actions') run_options = argparser.add_argument_group( 'Options controlling the execution of checks') env_options = argparser.add_argument_group( 'Options controlling the ReFrame environment') misc_options = argparser.add_argument_group('Miscellaneous options') # Output directory options output_options.add_argument('--dont-restage', action='store_false', dest='clean_stagedir', help='Reuse the test stage directory', envvar='RFM_CLEAN_STAGEDIR', configvar='general/clean_stagedir') output_options.add_argument( '--keep-stage-files', action='store_true', help='Keep stage directories even for successful checks', envvar='RFM_KEEP_STAGE_FILES', configvar='general/keep_stage_files') output_options.add_argument('-o', '--output', action='store', metavar='DIR', help='Set output directory prefix to DIR', envvar='RFM_OUTPUT_DIR', configvar='systems/outputdir') output_options.add_argument( '--perflogdir', action='store', metavar='DIR', help=('Set performance log data directory prefix ' '(relevant only to the filelog log handler)'), envvar='RFM_PERFLOG_DIR', configvar='logging/handlers_perflog/filelog_basedir') output_options.add_argument('--prefix', action='store', metavar='DIR', help='Set general directory prefix to DIR', envvar='RFM_PREFIX', configvar='systems/prefix') output_options.add_argument('--report-file', action='store', metavar='FILE', help="Store JSON run report in FILE", envvar='RFM_REPORT_FILE', configvar='general/report_file') output_options.add_argument('--report-junit', action='store', metavar='FILE', help="Store a JUnit report in FILE", envvar='RFM_REPORT_JUNIT', configvar='general/report_junit') output_options.add_argument('-s', '--stage', action='store', metavar='DIR', help='Set stage directory prefix to DIR', envvar='RFM_STAGE_DIR', configvar='systems/stagedir') output_options.add_argument( '--save-log-files', action='store_true', default=False, help='Save ReFrame log files to the output directory', envvar='RFM_SAVE_LOG_FILES', configvar='general/save_log_files') output_options.add_argument( '--timestamp', action='store', nargs='?', const='%FT%T', metavar='TIMEFMT', help=('Append a timestamp to the output and stage directory prefixes ' '(default: "%%FT%%T")'), envvar='RFM_TIMESTAMP_DIRS', configvar='general/timestamp_dirs') # Check discovery options locate_options.add_argument('-c', '--checkpath', action='append', metavar='PATH', help="Add PATH to the check search path list", envvar='RFM_CHECK_SEARCH_PATH :', configvar='general/check_search_path') locate_options.add_argument( '--ignore-check-conflicts', action='store_true', help=('Skip checks with conflicting names ' '(this option is deprecated and has no effect)'), envvar='RFM_IGNORE_CHECK_CONFLICTS', configvar='general/ignore_check_conflicts') locate_options.add_argument( '-R', '--recursive', action='store_true', help='Search for checks in the search path recursively', envvar='RFM_CHECK_SEARCH_RECURSIVE', configvar='general/check_search_recursive') # Select options select_options.add_argument('--cpu-only', action='store_true', help='Select only CPU checks') select_options.add_argument( '--failed', action='store_true', help="Select failed test cases (only when '--restore-session' is used)" ) select_options.add_argument('--gpu-only', action='store_true', help='Select only GPU checks') select_options.add_argument( '--maintainer', action='append', dest='maintainers', default=[], metavar='PATTERN', help='Select checks with at least one maintainer matching PATTERN') select_options.add_argument( '-n', '--name', action='append', dest='names', default=[], metavar='PATTERN', help='Select checks whose name matches PATTERN') # FIXME: The following is the only selection option that has an associated # (undocumented) configuration variable. This is to support pruning of the # partition environments as the runtime is created, similarly to how the # system partitions are treated. Currently, this facilitates the # implementation of fixtures, but we should reconsider it: see discussion # in https://github.com/eth-cscs/reframe/issues/2245 select_options.add_argument( '-p', '--prgenv', action='append', default=[r'.*'], metavar='PATTERN', configvar='general/valid_env_names', help=('Select checks with at least one ' 'programming environment matching PATTERN')) select_options.add_argument( '-T', '--exclude-tag', action='append', dest='exclude_tags', metavar='PATTERN', default=[], help='Exclude checks whose tag matches PATTERN') select_options.add_argument( '-t', '--tag', action='append', dest='tags', metavar='PATTERN', default=[], help='Select checks with at least one tag matching PATTERN') select_options.add_argument( '-x', '--exclude', action='append', dest='exclude_names', metavar='PATTERN', default=[], help='Exclude checks whose name matches PATTERN') # Action options action_options.add_argument( '--ci-generate', action='store', metavar='FILE', help=('Generate into FILE a Gitlab CI pipeline ' 'for the selected tests and exit'), ) action_options.add_argument( '-L', '--list-detailed', action='store_true', help='List the selected checks providing details for each test') action_options.add_argument('-l', '--list', action='store_true', help='List the selected checks') action_options.add_argument( '--list-tags', action='store_true', help='List the unique tags found in the selected tests and exit') action_options.add_argument('-r', '--run', action='store_true', help='Run the selected checks') # Run options run_options.add_argument('--disable-hook', action='append', metavar='NAME', dest='hooks', default=[], help='Disable a pipeline hook for this run') run_options.add_argument( '--exec-policy', metavar='POLICY', action='store', choices=['async', 'serial'], default='async', help='Set the execution policy of ReFrame (default: "async")') run_options.add_argument( '--flex-alloc-nodes', action='store', dest='flex_alloc_nodes', metavar='{all|STATE|NUM}', default=None, help='Set strategy for the flexible node allocation (default: "idle").' ) run_options.add_argument('--force-local', action='store_true', help='Force local execution of checks') run_options.add_argument('-J', '--job-option', action='append', metavar='OPT', dest='job_options', default=[], help='Pass option OPT to job scheduler') run_options.add_argument( '--max-retries', metavar='NUM', action='store', default=0, help='Set the maximum number of times a failed regression test ' 'may be retried (default: 0)') run_options.add_argument('--maxfail', metavar='NUM', action='store', default=sys.maxsize, help='Exit after first NUM failures') run_options.add_argument('--mode', action='store', help='Execution mode to use') run_options.add_argument('--restore-session', action='store', nargs='?', const='', metavar='REPORT', help='Restore a testing session from REPORT file') run_options.add_argument('-S', '--setvar', action='append', metavar='[TEST.]VAR=VAL', dest='vars', default=[], help=('Set test variable VAR to VAL in all tests ' 'or optionally in TEST only')) run_options.add_argument('--skip-performance-check', action='store_true', help='Skip performance checking') run_options.add_argument('--skip-prgenv-check', action='store_true', help='Skip programming environment check') run_options.add_argument('--skip-sanity-check', action='store_true', help='Skip sanity checking') run_options.add_argument('--skip-system-check', action='store_true', help='Skip system check') run_options.add_argument('--strict', action='store_true', help='Enforce strict performance checking') # Environment options env_options.add_argument('-M', '--map-module', action='append', metavar='MAPPING', dest='module_mappings', default=[], help='Add a module mapping', envvar='RFM_MODULE_MAPPINGS ,', configvar='general/module_mappings') env_options.add_argument( '-m', '--module', action='append', default=[], metavar='MOD', dest='user_modules', help='Load module MOD before running any regression check', envvar='RFM_USER_MODULES ,', configvar='general/user_modules') env_options.add_argument('--module-mappings', action='store', metavar='FILE', dest='module_map_file', help='Load module mappings from FILE', envvar='RFM_MODULE_MAP_FILE', configvar='general/module_map_file') env_options.add_argument( '--module-path', action='append', metavar='PATH', dest='module_paths', default=[], help='(Un)use module path PATH before running any regression check', ) env_options.add_argument( '--non-default-craype', action='store_true', help='Test a non-default Cray Programming Environment', envvar='RFM_NON_DEFAULT_CRAYPE', configvar='general/non_default_craype') env_options.add_argument( '--purge-env', action='store_true', dest='purge_env', default=False, help='Unload all modules before running any regression check', envvar='RFM_PURGE_ENVIRONMENT', configvar='general/purge_environment') env_options.add_argument( '-u', '--unload-module', action='append', metavar='MOD', dest='unload_modules', default=[], help='Unload module MOD before running any regression check', envvar='RFM_UNLOAD_MODULES ,', configvar='general/unload_modules') # Miscellaneous options misc_options.add_argument('-C', '--config-file', action='store', dest='config_file', metavar='FILE', help='Set configuration file', envvar='RFM_CONFIG_FILE') misc_options.add_argument('--detect-host-topology', action='store', nargs='?', const='-', help='Detect the local host topology and exit') misc_options.add_argument('--failure-stats', action='store_true', help='Print failure statistics') misc_options.add_argument('--nocolor', action='store_false', dest='colorize', help='Disable coloring of output', envvar='RFM_COLORIZE', configvar='general/colorize') misc_options.add_argument('--performance-report', action='store_true', help='Print a report for performance tests') misc_options.add_argument( '--show-config', action='store', nargs='?', const='all', metavar='PARAM', help='Print the value of configuration parameter PARAM and exit') misc_options.add_argument('--system', action='store', help='Load configuration for SYSTEM', envvar='RFM_SYSTEM') misc_options.add_argument( '--upgrade-config-file', action='store', metavar='OLD[:NEW]', help='Upgrade ReFrame 2.x configuration file to ReFrame 3.x syntax') misc_options.add_argument('-V', '--version', action='version', version=osext.reframe_version()) misc_options.add_argument('-v', '--verbose', action='count', help='Increase verbosity level of output', envvar='RFM_VERBOSE', configvar='general/verbose') misc_options.add_argument( '-q', '--quiet', action='count', default=0, help='Decrease verbosity level of output', ) # Options not associated with command-line arguments argparser.add_argument( dest='git_timeout', envvar='RFM_GIT_TIMEOUT', configvar='general/git_timeout', help=('Timeout in seconds when checking if the url is a ' 'valid repository.')) argparser.add_argument( dest='graylog_server', envvar='RFM_GRAYLOG_ADDRESS', configvar='logging/handlers_perflog/graylog_address', help='Graylog server address') argparser.add_argument(dest='httpjson_url', envvar='RFM_HTTPJSON_URL', configvar='logging/handlers_perflog/httpjson_url', help='URL of HTTP server accepting JSON logs') argparser.add_argument(dest='ignore_reqnodenotavail', envvar='RFM_IGNORE_REQNODENOTAVAIL', configvar='schedulers/ignore_reqnodenotavail', action='store_true', help='Graylog server address') argparser.add_argument(dest='compact_test_names', envvar='RFM_COMPACT_TEST_NAMES', configvar='general/compact_test_names', action='store_true', help='Use a compact test naming scheme') argparser.add_argument(dest='remote_detect', envvar='RFM_REMOTE_DETECT', configvar='general/remote_detect', action='store_true', help='Detect remote system topology') argparser.add_argument( dest='remote_workdir', envvar='RFM_REMOTE_WORKDIR', configvar='general/remote_workdir', action='store', help='Working directory for launching ReFrame remotely') argparser.add_argument(dest='resolve_module_conflicts', envvar='RFM_RESOLVE_MODULE_CONFLICTS', configvar='general/resolve_module_conflicts', action='store_true', help='Resolve module conflicts automatically') argparser.add_argument(dest='syslog_address', envvar='RFM_SYSLOG_ADDRESS', configvar='logging/handlers_perflog/syslog_address', help='Syslog server address') argparser.add_argument( dest='trap_job_errors', envvar='RFM_TRAP_JOB_ERRORS', configvar='general/trap_job_errors', action='store_true', help='Trap job errors in job scripts and fail tests automatically') argparser.add_argument(dest='use_login_shell', envvar='RFM_USE_LOGIN_SHELL', configvar='general/use_login_shell', action='store_true', help='Use a login shell for job scripts') # Parse command line options = argparser.parse_args() if len(sys.argv) == 1: argparser.print_help() sys.exit(1) # First configure logging with our generic configuration so as to be able # to print pretty messages; logging will be reconfigured by user's # configuration later site_config = config.load_config( os.path.join(reframe.INSTALL_PREFIX, 'reframe/core/settings.py')) site_config.select_subconfig('generic') options.update_config(site_config) logging.configure_logging(site_config) logging.getlogger().colorize = site_config.get('general/0/colorize') printer = PrettyPrinter() printer.colorize = site_config.get('general/0/colorize') printer.adjust_verbosity(calc_verbosity(site_config, options.quiet)) if os.getenv('RFM_GRAYLOG_SERVER'): printer.warning( 'RFM_GRAYLOG_SERVER environment variable is deprecated; ' 'please use RFM_GRAYLOG_ADDRESS instead') os.environ['RFM_GRAYLOG_ADDRESS'] = os.getenv('RFM_GRAYLOG_SERVER') if options.upgrade_config_file is not None: old_config, *new_config = options.upgrade_config_file.split(':', maxsplit=1) new_config = new_config[0] if new_config else None try: new_config = config.convert_old_config(old_config, new_config) except Exception as e: printer.error(f'could not convert file: {e}') sys.exit(1) printer.info(f'Conversion successful! ' f'The converted file can be found at {new_config!r}.') sys.exit(0) # Now configure ReFrame according to the user configuration file try: try: printer.debug('Loading user configuration') site_config = config.load_config(options.config_file) except warnings.ReframeDeprecationWarning as e: printer.warning(e) converted = config.convert_old_config(options.config_file) printer.warning(f"configuration file has been converted " f"to the new syntax here: '{converted}'") site_config = config.load_config(converted) site_config.validate() # We ignore errors about unresolved sections or configuration # parameters here, because they might be defined at the individual # partition level and will be caught when we will instantiating # internally the system and partitions later on. site_config.select_subconfig(options.system, ignore_resolve_errors=True) for err in options.update_config(site_config): printer.warning(str(err)) # Update options from the selected execution mode if options.mode: mode_args = site_config.get(f'modes/@{options.mode}/options') # We lexically split the mode options, because otherwise spaces # will be treated as part of the option argument; see GH bug #1554 mode_args = list( itertools.chain.from_iterable( shlex.split(m) for m in mode_args)) # Parse the mode's options and reparse the command-line options = argparser.parse_args(mode_args) options = argparser.parse_args(namespace=options.cmd_options) options.update_config(site_config) logging.configure_logging(site_config) except (OSError, errors.ConfigError) as e: printer.error(f'failed to load configuration: {e}') printer.error(logfiles_message()) sys.exit(1) logging.getlogger().colorize = site_config.get('general/0/colorize') printer.colorize = site_config.get('general/0/colorize') printer.adjust_verbosity(calc_verbosity(site_config, options.quiet)) try: printer.debug('Initializing runtime') runtime.init_runtime(site_config) except errors.ConfigError as e: printer.error(f'failed to initialize runtime: {e}') printer.error(logfiles_message()) sys.exit(1) if site_config.get('general/0/ignore_check_conflicts'): logging.getlogger().warning( "the 'ignore_check_conflicts' option is deprecated " "and will be removed in the future") rt = runtime.runtime() try: if site_config.get('general/0/module_map_file'): rt.modules_system.load_mapping_from_file( site_config.get('general/0/module_map_file')) if site_config.get('general/0/module_mappings'): for m in site_config.get('general/0/module_mappings'): rt.modules_system.load_mapping(m) except (errors.ConfigError, OSError) as e: printer.error('could not load module mappings: %s' % e) sys.exit(1) if (osext.samefile(rt.stage_prefix, rt.output_prefix) and not site_config.get('general/0/keep_stage_files')): printer.error("stage and output refer to the same directory; " "if this is on purpose, please use the " "'--keep-stage-files' option.") printer.error(logfiles_message()) sys.exit(1) # Show configuration after everything is set up if options.show_config: config_param = options.show_config if config_param == 'all': printer.info(str(rt.site_config)) else: value = rt.get_option(config_param) if value is None: printer.error( f'no such configuration parameter found: {config_param}') else: printer.info(json.dumps(value, indent=2)) sys.exit(0) if options.detect_host_topology: from reframe.utility.cpuinfo import cpuinfo topofile = options.detect_host_topology if topofile == '-': json.dump(cpuinfo(), sys.stdout, indent=2) sys.stdout.write('\n') else: try: with open(topofile, 'w') as fp: json.dump(cpuinfo(), fp, indent=2) fp.write('\n') except OSError as e: getlogger().error( f'could not write topology file: {topofile!r}') sys.exit(1) sys.exit(0) autodetect.detect_topology() printer.debug(format_env(options.env_vars)) # Setup the check loader if options.restore_session is not None: # We need to load the failed checks only from a list of reports if options.restore_session: filenames = options.restore_session.split(',') else: filenames = [ runreport.next_report_filename(osext.expandvars( site_config.get('general/0/report_file')), new=False) ] report = runreport.load_report(*filenames) check_search_path = list(report.slice('filename', unique=True)) check_search_recursive = False # If `-c` or `-R` are passed explicitly outside the configuration # file, override the values set from the report file if site_config.is_sticky_option('general/check_search_path'): printer.warning( 'Ignoring check search path set in the report file: ' 'search path set explicitly in the command-line or ' 'the environment') check_search_path = site_config.get('general/0/check_search_path') if site_config.is_sticky_option('general/check_search_recursive'): printer.warning( 'Ignoring check search recursive option from the report file: ' 'option set explicitly in the command-line or the environment') check_search_recursive = site_config.get( 'general/0/check_search_recursive') else: check_search_recursive = site_config.get( 'general/0/check_search_recursive') check_search_path = site_config.get('general/0/check_search_path') # Collect any variables set from the command line external_vars = {} for expr in options.vars: try: lhs, rhs = expr.split('=', maxsplit=1) except ValueError: printer.warning( f'invalid test variable assignment: {expr!r}; skipping') else: external_vars[lhs] = rhs loader = RegressionCheckLoader(check_search_path, check_search_recursive, external_vars) def print_infoline(param, value): param = param + ':' printer.info(f" {param.ljust(18)} {value}") session_info = { 'cmdline': ' '.join(sys.argv), 'config_file': rt.site_config.filename, 'data_version': runreport.DATA_VERSION, 'hostname': socket.getfqdn(), 'prefix_output': rt.output_prefix, 'prefix_stage': rt.stage_prefix, 'user': osext.osuser(), 'version': osext.reframe_version(), 'workdir': os.getcwd(), } # Print command line printer.info(f"[ReFrame Setup]") print_infoline('version', session_info['version']) print_infoline('command', repr(session_info['cmdline'])) print_infoline( f"launched by", f"{session_info['user'] or '<unknown>'}@{session_info['hostname']}") print_infoline('working directory', repr(session_info['workdir'])) print_infoline('settings file', f"{session_info['config_file']!r}") print_infoline( 'check search path', f"{'(R) ' if loader.recurse else ''}" f"{':'.join(loader.load_path)!r}") print_infoline('stage directory', repr(session_info['prefix_stage'])) print_infoline('output directory', repr(session_info['prefix_output'])) printer.info('') try: # Locate and load checks checks_found = loader.load_all() printer.verbose(f'Loaded {len(checks_found)} test(s)') # Generate all possible test cases first; we will need them for # resolving dependencies after filtering testcases_all = generate_testcases(checks_found, options.skip_system_check, options.skip_prgenv_check) testcases = testcases_all printer.verbose(f'Generated {len(testcases)} test case(s)') # Filter test cases by name if options.exclude_names: for name in options.exclude_names: testcases = filter(filters.have_not_name(name), testcases) if options.names: testcases = filter(filters.have_name('|'.join(options.names)), testcases) testcases = list(testcases) printer.verbose( f'Filtering test cases(s) by name: {len(testcases)} remaining') # Filter test cases by tags for tag in options.exclude_tags: testcases = filter(filters.have_not_tag(tag), testcases) for tag in options.tags: testcases = filter(filters.have_tag(tag), testcases) testcases = list(testcases) printer.verbose( f'Filtering test cases(s) by tags: {len(testcases)} remaining') # Filter test cases by maintainers for maint in options.maintainers: testcases = filter(filters.have_maintainer(maint), testcases) # Filter test cases further if options.gpu_only and options.cpu_only: printer.error("options `--gpu-only' and `--cpu-only' " "are mutually exclusive") sys.exit(1) if options.gpu_only: testcases = filter(filters.have_gpu_only(), testcases) elif options.cpu_only: testcases = filter(filters.have_cpu_only(), testcases) testcases = list(testcases) printer.verbose(f'Filtering test cases(s) by other attributes: ' f'{len(testcases)} remaining') # Filter in failed cases if options.failed: if options.restore_session is None: printer.error( "the option '--failed' can only be used " "in combination with the '--restore-session' option") sys.exit(1) def _case_failed(t): rec = report.case(*t) if not rec: return False return (rec['result'] == 'failure' or rec['result'] == 'aborted') testcases = list(filter(_case_failed, testcases)) printer.verbose(f'Filtering successful test case(s): ' f'{len(testcases)} remaining') # Prepare for running printer.debug('Building and validating the full test DAG') testgraph, skipped_cases = dependencies.build_deps(testcases_all) if skipped_cases: # Some cases were skipped, so adjust testcases testcases = list(set(testcases) - set(skipped_cases)) printer.verbose( f'Filtering test case(s) due to unresolved dependencies: ' f'{len(testcases)} remaining') dependencies.validate_deps(testgraph) printer.debug('Full test DAG:') printer.debug(dependencies.format_deps(testgraph)) restored_cases = [] if len(testcases) != len(testcases_all): testgraph = dependencies.prune_deps( testgraph, testcases, max_depth=1 if options.restore_session is not None else None) printer.debug('Pruned test DAG') printer.debug(dependencies.format_deps(testgraph)) if options.restore_session is not None: testgraph, restored_cases = report.restore_dangling(testgraph) testcases = dependencies.toposort(testgraph, is_subgraph=options.restore_session is not None) printer.verbose(f'Final number of test cases: {len(testcases)}') # Disable hooks for tc in testcases: for h in options.hooks: tc.check.disable_hook(h) # Act on checks if options.list or options.list_detailed: list_checks(testcases, printer, options.list_detailed) sys.exit(0) if options.list_tags: list_tags(testcases, printer) sys.exit(0) if options.ci_generate: list_checks(testcases, printer) printer.info('[Generate CI]') with open(options.ci_generate, 'wt') as fp: ci.emit_pipeline(fp, testcases) printer.info(f' Gitlab pipeline generated successfully ' f'in {options.ci_generate!r}.\n') sys.exit(0) if not options.run: printer.error("No action option specified. Available options:\n" " - `-l'/`-L' for listing\n" " - `-r' for running\n" " - `--list-tags' for listing unique test tags\n" " - `--ci-generate' for generating a CI pipeline\n" f"Try `{argparser.prog} -h' for more options.") sys.exit(1) # Manipulate ReFrame's environment if site_config.get('general/0/purge_environment'): rt.modules_system.unload_all() else: for m in site_config.get('general/0/unload_modules'): rt.modules_system.unload_module(**m) # Load the environment for the current system try: printer.debug(f'Loading environment for current system') runtime.loadenv(rt.system.preload_environ) except errors.EnvironError as e: printer.error("failed to load current system's environment; " "please check your configuration") printer.debug(str(e)) raise def module_use(*paths): try: rt.modules_system.searchpath_add(*paths) except errors.EnvironError as e: printer.warning(f'could not add module paths correctly') printer.debug(str(e)) def module_unuse(*paths): try: rt.modules_system.searchpath_remove(*paths) except errors.EnvironError as e: printer.warning(f'could not remove module paths correctly') printer.debug(str(e)) printer.debug('(Un)using module paths from command line') module_paths = {} for d in options.module_paths: if d.startswith('-'): module_paths.setdefault('-', []) module_paths['-'].append(d[1:]) elif d.startswith('+'): module_paths.setdefault('+', []) module_paths['+'].append(d[1:]) else: module_paths.setdefault('x', []) module_paths['x'].append(d) for op, paths in module_paths.items(): if op == '+': module_use(*paths) elif op == '-': module_unuse(*paths) else: # First empty the current module path in a portable way searchpath = [p for p in rt.modules_system.searchpath if p] if searchpath: rt.modules_system.searchpath_remove(*searchpath) # Treat `A:B` syntax as well in this case paths = itertools.chain(*(p.split(':') for p in paths)) module_use(*paths) printer.debug('Loading user modules from command line') for m in site_config.get('general/0/user_modules'): try: rt.modules_system.load_module(**m, force=True) except errors.EnvironError as e: printer.warning( f'could not load module {m["name"]!r} correctly; ' f'skipping...') printer.debug(str(e)) options.flex_alloc_nodes = options.flex_alloc_nodes or 'idle' # Run the tests # Setup the execution policy if options.exec_policy == 'serial': exec_policy = SerialExecutionPolicy() elif options.exec_policy == 'async': exec_policy = AsynchronousExecutionPolicy() else: # This should not happen, since choices are handled by # argparser printer.error("unknown execution policy `%s': Exiting...") sys.exit(1) exec_policy.skip_system_check = options.skip_system_check exec_policy.force_local = options.force_local exec_policy.strict_check = options.strict exec_policy.skip_sanity_check = options.skip_sanity_check exec_policy.skip_performance_check = options.skip_performance_check exec_policy.keep_stage_files = site_config.get( 'general/0/keep_stage_files') try: errmsg = "invalid option for --flex-alloc-nodes: '{0}'" sched_flex_alloc_nodes = int(options.flex_alloc_nodes) if sched_flex_alloc_nodes <= 0: raise errors.ConfigError( errmsg.format(options.flex_alloc_nodes)) except ValueError: sched_flex_alloc_nodes = options.flex_alloc_nodes exec_policy.sched_flex_alloc_nodes = sched_flex_alloc_nodes parsed_job_options = [] for opt in options.job_options: opt_split = opt.split('=', maxsplit=1) optstr = opt_split[0] valstr = opt_split[1] if len(opt_split) > 1 else '' if opt.startswith('-') or opt.startswith('#'): parsed_job_options.append(opt) elif len(optstr) == 1: parsed_job_options.append(f'-{optstr} {valstr}') else: parsed_job_options.append(f'--{optstr} {valstr}') exec_policy.sched_options = parsed_job_options try: max_retries = int(options.max_retries) except ValueError: raise errors.ConfigError( f'--max-retries is not a valid integer: {max_retries}' ) from None try: max_failures = int(options.maxfail) if max_failures < 0: raise errors.ConfigError( f'--maxfail should be a non-negative integer: ' f'{options.maxfail!r}') except ValueError: raise errors.ConfigError( f'--maxfail is not a valid integer: {options.maxfail!r}' ) from None runner = Runner(exec_policy, printer, max_retries, max_failures) try: time_start = time.time() session_info['time_start'] = time.strftime( '%FT%T%z', time.localtime(time_start), ) runner.runall(testcases, restored_cases) finally: time_end = time.time() session_info['time_end'] = time.strftime('%FT%T%z', time.localtime(time_end)) session_info['time_elapsed'] = time_end - time_start # Print a retry report if we did any retries if runner.stats.failed(run=0): printer.info(runner.stats.retry_report()) # Print a failure report if we had failures in the last run success = True if runner.stats.failed(): success = False runner.stats.print_failure_report(printer) if options.failure_stats: runner.stats.print_failure_stats(printer) if options.performance_report: printer.info(runner.stats.performance_report()) # Generate the report for this session report_file = os.path.normpath( osext.expandvars(rt.get_option('general/0/report_file'))) basedir = os.path.dirname(report_file) if basedir: os.makedirs(basedir, exist_ok=True) # Build final JSON report run_stats = runner.stats.json() session_info.update({ 'num_cases': run_stats[0]['num_cases'], 'num_failures': run_stats[-1]['num_failures'] }) json_report = { 'session_info': session_info, 'runs': run_stats, 'restored_cases': [] } if options.restore_session is not None: for c in restored_cases: json_report['restored_cases'].append(report.case(*c)) report_file = runreport.next_report_filename(report_file) try: with open(report_file, 'w') as fp: jsonext.dump(json_report, fp, indent=2) fp.write('\n') printer.info(f'Run report saved in {report_file!r}') except OSError as e: printer.warning( f'failed to generate report in {report_file!r}: {e}') # Generate the junit xml report for this session junit_report_file = rt.get_option('general/0/report_junit') if junit_report_file: # Expand variables in filename junit_report_file = osext.expandvars(junit_report_file) junit_xml = runreport.junit_xml_report(json_report) try: with open(junit_report_file, 'w') as fp: runreport.junit_dump(junit_xml, fp) except OSError as e: printer.warning( f'failed to generate report in {junit_report_file!r}: ' f'{e}') if not success: sys.exit(1) sys.exit(0) except (Exception, KeyboardInterrupt, errors.ReframeFatalError): exc_info = sys.exc_info() tb = ''.join(traceback.format_exception(*exc_info)) printer.error(f'run session stopped: {errors.what(*exc_info)}') if errors.is_exit_request(*exc_info): # Print stack traces for exit requests only when TOO verbose printer.debug2(tb) elif errors.is_severe(*exc_info): printer.error(tb) else: printer.verbose(tb) sys.exit(1) finally: try: log_files = logging.log_files() if site_config.get('general/0/save_log_files'): log_files = logging.save_log_files(rt.output_prefix) except OSError as e: printer.error(f'could not save log file: {e}') sys.exit(1) finally: printer.info(logfiles_message())
def main(): # Setup command line options argparser = argparse.ArgumentParser() output_options = argparser.add_argument_group( 'Options controlling ReFrame output') locate_options = argparser.add_argument_group( 'Options for discovering checks') select_options = argparser.add_argument_group( 'Options for selecting checks') action_options = argparser.add_argument_group( 'Options controlling actions') run_options = argparser.add_argument_group( 'Options controlling the execution of checks') env_options = argparser.add_argument_group( 'Options controlling the ReFrame environment') misc_options = argparser.add_argument_group('Miscellaneous options') # Output directory options output_options.add_argument('--prefix', action='store', metavar='DIR', help='Set general directory prefix to DIR', envvar='RFM_PREFIX', configvar='systems/prefix') output_options.add_argument('-o', '--output', action='store', metavar='DIR', help='Set output directory prefix to DIR', envvar='RFM_OUTPUT_DIR', configvar='systems/outputdir') output_options.add_argument('-s', '--stage', action='store', metavar='DIR', help='Set stage directory prefix to DIR', envvar='RFM_STAGE_DIR', configvar='systems/stagedir') output_options.add_argument( '--timestamp', action='store', nargs='?', const='', metavar='TIMEFMT', help=('Append a timestamp to the output and stage directory prefixes ' '(default: "%%FT%%T")'), envvar='RFM_TIMESTAMP_DIRS', configvar='general/timestamp_dirs') output_options.add_argument( '--perflogdir', action='store', metavar='DIR', help=('Set performance log data directory prefix ' '(relevant only to the filelog log handler)'), envvar='RFM_PERFLOG_DIR', configvar='logging/handlers_perflog/filelog_basedir') output_options.add_argument( '--keep-stage-files', action='store_true', help='Keep stage directories even for successful checks', envvar='RFM_KEEP_STAGE_FILES', configvar='general/keep_stage_files') output_options.add_argument('--dont-restage', action='store_false', dest='clean_stagedir', help='Reuse the test stage directory', envvar='RFM_CLEAN_STAGEDIR', configvar='general/clean_stagedir') output_options.add_argument( '--save-log-files', action='store_true', default=False, help='Save ReFrame log files to the output directory', envvar='RFM_SAVE_LOG_FILES', configvar='general/save_log_files') output_options.add_argument('--report-file', action='store', metavar='FILE', help="Store JSON run report in FILE", envvar='RFM_REPORT_FILE', configvar='general/report_file') # Check discovery options locate_options.add_argument('-c', '--checkpath', action='append', metavar='PATH', help="Add PATH to the check search path list", envvar='RFM_CHECK_SEARCH_PATH :', configvar='general/check_search_path') locate_options.add_argument( '-R', '--recursive', action='store_true', help='Search for checks in the search path recursively', envvar='RFM_CHECK_SEARCH_RECURSIVE', configvar='general/check_search_recursive') locate_options.add_argument('--ignore-check-conflicts', action='store_true', help='Skip checks with conflicting names', envvar='RFM_IGNORE_CHECK_CONFLICTS', configvar='general/ignore_check_conflicts') # Select options select_options.add_argument( '-t', '--tag', action='append', dest='tags', metavar='PATTERN', default=[], help='Select checks with at least one tag matching PATTERN') select_options.add_argument( '-n', '--name', action='append', dest='names', default=[], metavar='PATTERN', help='Select checks whose name matches PATTERN') select_options.add_argument( '-x', '--exclude', action='append', dest='exclude_names', metavar='PATTERN', default=[], help='Exclude checks whose name matches PATTERN') select_options.add_argument( '-p', '--prgenv', action='append', default=[r'.*'], metavar='PATTERN', help=('Select checks with at least one ' 'programming environment matching PATTERN')) select_options.add_argument('--gpu-only', action='store_true', help='Select only GPU checks') select_options.add_argument('--cpu-only', action='store_true', help='Select only CPU checks') # Action options action_options.add_argument('-l', '--list', action='store_true', help='List the selected checks') action_options.add_argument( '-L', '--list-detailed', action='store_true', help='List the selected checks providing details for each test') action_options.add_argument('-r', '--run', action='store_true', help='Run the selected checks') # Run options run_options.add_argument('-J', '--job-option', action='append', metavar='OPT', dest='job_options', default=[], help='Pass option OPT to job scheduler') run_options.add_argument('--force-local', action='store_true', help='Force local execution of checks') run_options.add_argument('--skip-sanity-check', action='store_true', help='Skip sanity checking') run_options.add_argument('--skip-performance-check', action='store_true', help='Skip performance checking') run_options.add_argument('--strict', action='store_true', help='Enforce strict performance checking') run_options.add_argument('--skip-system-check', action='store_true', help='Skip system check') run_options.add_argument('--skip-prgenv-check', action='store_true', help='Skip programming environment check') run_options.add_argument( '--exec-policy', metavar='POLICY', action='store', choices=['async', 'serial'], default='async', help='Set the execution policy of ReFrame (default: "async")') run_options.add_argument('--mode', action='store', help='Execution mode to use') run_options.add_argument( '--max-retries', metavar='NUM', action='store', default=0, help='Set the maximum number of times a failed regression test ' 'may be retried (default: 0)') run_options.add_argument( '--flex-alloc-nodes', action='store', dest='flex_alloc_nodes', metavar='{all|STATE|NUM}', default=None, help='Set strategy for the flexible node allocation (default: "idle").' ) run_options.add_argument('--disable-hook', action='append', metavar='NAME', dest='hooks', default=[], help='Disable a pipeline hook for this run') env_options.add_argument('-M', '--map-module', action='append', metavar='MAPPING', dest='module_mappings', default=[], help='Add a module mapping', envvar='RFM_MODULE_MAPPINGS ,', configvar='general/module_mappings') env_options.add_argument( '-m', '--module', action='append', default=[], metavar='MOD', dest='user_modules', help='Load module MOD before running any regression check', envvar='RFM_USER_MODULES ,', configvar='general/user_modules') env_options.add_argument('--module-mappings', action='store', metavar='FILE', dest='module_map_file', help='Load module mappings from FILE', envvar='RFM_MODULE_MAP_FILE', configvar='general/module_map_file') env_options.add_argument( '-u', '--unload-module', action='append', metavar='MOD', dest='unload_modules', default=[], help='Unload module MOD before running any regression check', envvar='RFM_UNLOAD_MODULES ,', configvar='general/unload_modules') env_options.add_argument( '--purge-env', action='store_true', dest='purge_env', default=False, help='Unload all modules before running any regression check', envvar='RFM_PURGE_ENVIRONMENT', configvar='general/purge_environment') env_options.add_argument( '--non-default-craype', action='store_true', help='Test a non-default Cray Programming Environment', envvar='RFM_NON_DEFAULT_CRAYPE', configvar='general/non_default_craype') # Miscellaneous options misc_options.add_argument('-C', '--config-file', action='store', dest='config_file', metavar='FILE', help='Set configuration file', envvar='RFM_CONFIG_FILE') misc_options.add_argument('--nocolor', action='store_false', dest='colorize', help='Disable coloring of output', envvar='RFM_COLORIZE', configvar='general/colorize') misc_options.add_argument('--failure-stats', action='store_true', help='Print failure statistics') misc_options.add_argument('--performance-report', action='store_true', help='Print a report for performance tests') misc_options.add_argument( '--show-config', action='store', nargs='?', const='all', metavar='PARAM', help='Print the value of configuration parameter PARAM and exit') misc_options.add_argument('--system', action='store', help='Load configuration for SYSTEM', envvar='RFM_SYSTEM') misc_options.add_argument( '--upgrade-config-file', action='store', metavar='OLD[:NEW]', help='Upgrade ReFrame 2.x configuration file to ReFrame 3.x syntax') misc_options.add_argument('-V', '--version', action='version', version=os_ext.reframe_version()) misc_options.add_argument('-v', '--verbose', action='count', help='Increase verbosity level of output', envvar='RFM_VERBOSE', configvar='general/verbose') # Options not associated with command-line arguments argparser.add_argument( dest='graylog_server', envvar='RFM_GRAYLOG_ADDRESS', configvar='logging/handlers_perflog/graylog_address', help='Graylog server address') argparser.add_argument(dest='syslog_address', envvar='RFM_SYSLOG_ADDRESS', configvar='logging/handlers_perflog/syslog_address', help='Syslog server address') argparser.add_argument(dest='ignore_reqnodenotavail', envvar='RFM_IGNORE_REQNODENOTAVAIL', configvar='schedulers/ignore_reqnodenotavail', action='store_true', help='Graylog server address') argparser.add_argument(dest='use_login_shell', envvar='RFM_USE_LOGIN_SHELL', configvar='general/use_login_shell', action='store_true', help='Use a login shell for job scripts') if len(sys.argv) == 1: argparser.print_help() sys.exit(1) # Parse command line options = argparser.parse_args() # First configure logging with our generic configuration so as to be able # to print pretty messages; logging will be reconfigured by user's # configuration later site_config = config.load_config( os.path.join(reframe.INSTALL_PREFIX, 'reframe/core/settings.py')) site_config.select_subconfig('generic') options.update_config(site_config) logging.configure_logging(site_config) logging.getlogger().colorize = site_config.get('general/0/colorize') printer = PrettyPrinter() printer.colorize = site_config.get('general/0/colorize') printer.inc_verbosity(site_config.get('general/0/verbose')) if os.getenv('RFM_GRAYLOG_SERVER'): printer.warning( 'RFM_GRAYLOG_SERVER environment variable is deprecated; ' 'please use RFM_GRAYLOG_ADDRESS instead') os.environ['RFM_GRAYLOG_ADDRESS'] = os.getenv('RFM_GRAYLOG_SERVER') if options.upgrade_config_file is not None: old_config, *new_config = options.upgrade_config_file.split(':', maxsplit=1) new_config = new_config[0] if new_config else None try: new_config = config.convert_old_config(old_config, new_config) except Exception as e: printer.error(f'could not convert file: {e}') sys.exit(1) printer.info(f'Conversion successful! ' f'The converted file can be found at {new_config!r}.') sys.exit(0) # Now configure ReFrame according to the user configuration file try: try: site_config = config.load_config(options.config_file) except ReframeDeprecationWarning as e: printer.warning(e) converted = config.convert_old_config(options.config_file) printer.warning(f"configuration file has been converted " f"to the new syntax here: '{converted}'") site_config = config.load_config(converted) site_config.validate() # We ignore errors about unresolved sections or configuration # parameters here, because they might be defined at the individual # partition level and will be caught when we will instantiating # internally the system and partitions later on. site_config.select_subconfig(options.system, ignore_resolve_errors=True) for err in options.update_config(site_config): printer.warning(str(err)) # Update options from the selected execution mode if options.mode: mode_args = site_config.get(f'modes/@{options.mode}/options') # Parse the mode's options and reparse the command-line options = argparser.parse_args(mode_args) options = argparser.parse_args(namespace=options.cmd_options) options.update_config(site_config) logging.configure_logging(site_config) except (OSError, ConfigError) as e: printer.error(f'failed to load configuration: {e}') sys.exit(1) logging.getlogger().colorize = site_config.get('general/0/colorize') printer.colorize = site_config.get('general/0/colorize') printer.inc_verbosity(site_config.get('general/0/verbose')) try: runtime.init_runtime(site_config) except ConfigError as e: printer.error(f'failed to initialize runtime: {e}') sys.exit(1) rt = runtime.runtime() try: if site_config.get('general/0/module_map_file'): rt.modules_system.load_mapping_from_file( site_config.get('general/0/module_map_file')) if site_config.get('general/0/module_mappings'): for m in site_config.get('general/0/module_mappings'): rt.modules_system.load_mapping(m) except (ConfigError, OSError) as e: printer.error('could not load module mappings: %s' % e) sys.exit(1) if (os_ext.samefile(rt.stage_prefix, rt.output_prefix) and not site_config.get('general/0/keep_stage_files')): printer.error("stage and output refer to the same directory; " "if this is on purpose, please use the " "'--keep-stage-files' option.") sys.exit(1) # Show configuration after everything is set up if options.show_config: config_param = options.show_config if config_param == 'all': printer.info(str(rt.site_config)) else: value = rt.get_option(config_param) if value is None: printer.error( f'no such configuration parameter found: {config_param}') else: printer.info(json.dumps(value, indent=2)) sys.exit(0) printer.debug(format_env(options.env_vars)) # Setup the check loader loader = RegressionCheckLoader( load_path=site_config.get('general/0/check_search_path'), recurse=site_config.get('general/0/check_search_recursive'), ignore_conflicts=site_config.get('general/0/ignore_check_conflicts')) def print_infoline(param, value): param = param + ':' printer.info(f" {param.ljust(18)} {value}") session_info = { 'cmdline': ' '.join(sys.argv), 'config_file': rt.site_config.filename, 'data_version': '1.0', 'hostname': socket.gethostname(), 'prefix_output': rt.output_prefix, 'prefix_stage': rt.stage_prefix, 'user': os_ext.osuser(), 'version': os_ext.reframe_version(), 'workdir': os.getcwd(), } # Print command line printer.info(f"[ReFrame Setup]") print_infoline('version', session_info['version']) print_infoline('command', repr(session_info['cmdline'])) print_infoline( f"launched by", f"{session_info['user'] or '<unknown>'}@{session_info['hostname']}") print_infoline('working directory', repr(session_info['workdir'])) print_infoline('settings file', f"{session_info['config_file']!r}") print_infoline( 'check search path', f"{'(R) ' if loader.recurse else ''}" f"{':'.join(loader.load_path)!r}") print_infoline('stage directory', repr(session_info['prefix_stage'])) print_infoline('output directory', repr(session_info['prefix_output'])) printer.info('') try: # Locate and load checks try: checks_found = loader.load_all() except OSError as e: raise ReframeError from e # Filter checks by name checks_matched = checks_found if options.exclude_names: for name in options.exclude_names: checks_matched = filter(filters.have_not_name(name), checks_matched) if options.names: checks_matched = filter(filters.have_name('|'.join(options.names)), checks_matched) # Filter checks by tags for tag in options.tags: checks_matched = filter(filters.have_tag(tag), checks_matched) # Filter checks by prgenv if not options.skip_prgenv_check: for prgenv in options.prgenv: checks_matched = filter(filters.have_prgenv(prgenv), checks_matched) # Filter checks by system if not options.skip_system_check: checks_matched = filter( filters.have_partition(rt.system.partitions), checks_matched) # Filter checks further if options.gpu_only and options.cpu_only: printer.error("options `--gpu-only' and `--cpu-only' " "are mutually exclusive") sys.exit(1) if options.gpu_only: checks_matched = filter(filters.have_gpu_only(), checks_matched) elif options.cpu_only: checks_matched = filter(filters.have_cpu_only(), checks_matched) # Determine the allowed programming environments allowed_environs = { e.name for env_patt in options.prgenv for p in rt.system.partitions for e in p.environs if re.match(env_patt, e.name) } # Generate the test cases, validate dependencies and sort them checks_matched = list(checks_matched) # Disable hooks for c in checks_matched: for h in options.hooks: type(c).disable_hook(h) testcases = generate_testcases(checks_matched, options.skip_system_check, options.skip_prgenv_check, allowed_environs) testgraph = dependency.build_deps(testcases) dependency.validate_deps(testgraph) testcases = dependency.toposort(testgraph) # Manipulate ReFrame's environment if site_config.get('general/0/purge_environment'): rt.modules_system.unload_all() else: for m in site_config.get('general/0/unload_modules'): rt.modules_system.unload_module(m) # Load the environment for the current system try: runtime.loadenv(rt.system.preload_environ) except EnvironError as e: printer.error("failed to load current system's environment; " "please check your configuration") printer.debug(str(e)) raise for m in site_config.get('general/0/user_modules'): try: rt.modules_system.load_module(m, force=True) except EnvironError as e: printer.warning("could not load module '%s' correctly: " "Skipping..." % m) printer.debug(str(e)) options.flex_alloc_nodes = options.flex_alloc_nodes or 'idle' # Act on checks success = True if options.list or options.list_detailed: list_checks(list(checks_matched), printer, options.list_detailed) elif options.run: # Setup the execution policy if options.exec_policy == 'serial': exec_policy = SerialExecutionPolicy() elif options.exec_policy == 'async': exec_policy = AsynchronousExecutionPolicy() else: # This should not happen, since choices are handled by # argparser printer.error("unknown execution policy `%s': Exiting...") sys.exit(1) exec_policy.skip_system_check = options.skip_system_check exec_policy.force_local = options.force_local exec_policy.strict_check = options.strict exec_policy.skip_sanity_check = options.skip_sanity_check exec_policy.skip_performance_check = options.skip_performance_check exec_policy.keep_stage_files = site_config.get( 'general/0/keep_stage_files') try: errmsg = "invalid option for --flex-alloc-nodes: '{0}'" sched_flex_alloc_nodes = int(options.flex_alloc_nodes) if sched_flex_alloc_nodes <= 0: raise ConfigError(errmsg.format(options.flex_alloc_nodes)) except ValueError: sched_flex_alloc_nodes = options.flex_alloc_nodes exec_policy.sched_flex_alloc_nodes = sched_flex_alloc_nodes parsed_job_options = [] for opt in options.job_options: if opt.startswith('-') or opt.startswith('#'): parsed_job_options.append(opt) elif len(opt) == 1: parsed_job_options.append(f'-{opt}') else: parsed_job_options.append(f'--{opt}') exec_policy.sched_options = parsed_job_options try: max_retries = int(options.max_retries) except ValueError: raise ConfigError('--max-retries is not a valid integer: %s' % max_retries) from None runner = Runner(exec_policy, printer, max_retries) try: time_start = time.time() session_info['time_start'] = time.strftime( '%FT%T%z', time.localtime(time_start), ) runner.runall(testcases) finally: time_end = time.time() session_info['time_end'] = time.strftime( '%FT%T%z', time.localtime(time_end)) session_info['time_elapsed'] = time_end - time_start # Print a retry report if we did any retries if runner.stats.failures(run=0): printer.info(runner.stats.retry_report()) # Print a failure report if we had failures in the last run if runner.stats.failures(): printer.info(runner.stats.failure_report()) success = False if options.failure_stats: printer.info(runner.stats.failure_stats()) if options.performance_report: printer.info(runner.stats.performance_report()) # Generate the report for this session report_file = os.path.normpath( os_ext.expandvars(rt.get_option('general/0/report_file'))) basedir = os.path.dirname(report_file) if basedir: os.makedirs(basedir, exist_ok=True) # Build final JSON report run_stats = runner.stats.json() session_info.update({ 'num_cases': run_stats[0]['num_cases'], 'num_failures': run_stats[-1]['num_failures'] }) json_report = {'session_info': session_info, 'runs': run_stats} report_file = generate_report_filename(report_file) try: with open(report_file, 'w') as fp: jsonext.dump(json_report, fp, indent=2) except OSError as e: printer.warning( f'failed to generate report in {report_file!r}: {e}') else: printer.error("No action specified. Please specify `-l'/`-L' for " "listing or `-r' for running. " "Try `%s -h' for more options." % argparser.prog) sys.exit(1) if not success: sys.exit(1) sys.exit(0) except KeyboardInterrupt: sys.exit(1) except ReframeError as e: printer.error(str(e)) sys.exit(1) except (Exception, ReframeFatalError): printer.error(format_exception(*sys.exc_info())) sys.exit(1) finally: try: log_files = logging.log_files() if site_config.get('general/0/save_log_files'): log_files = logging.save_log_files(rt.output_prefix) except OSError as e: printer.error('could not save log file: %s' % e) sys.exit(1) finally: if not log_files: msg = '<no log file was generated>' else: msg = f'{", ".join(repr(f) for f in log_files)}' printer.info(f'Log file(s) saved in: {msg}')
def main(): # Setup command line options argparser = ArgumentParser() output_options = argparser.add_argument_group( 'Options controlling regression directories') locate_options = argparser.add_argument_group( 'Options for locating checks') select_options = argparser.add_argument_group( 'Options for selecting checks') action_options = argparser.add_argument_group( 'Options controlling actions') run_options = argparser.add_argument_group( 'Options controlling execution of checks') misc_options = argparser.add_argument_group('Miscellaneous options') # Output directory options output_options.add_argument( '--prefix', action='store', metavar='DIR', help='Set regression prefix directory to DIR') output_options.add_argument( '-o', '--output', action='store', metavar='DIR', help='Set regression output directory to DIR') output_options.add_argument( '-s', '--stage', action='store', metavar='DIR', help='Set regression stage directory to DIR') output_options.add_argument( '--logdir', action='store', metavar='DIR', help='Set regression log directory to DIR') output_options.add_argument( '--keep-stage-files', action='store_true', help='Keep stage directory even if check is successful') output_options.add_argument( '--save-log-files', action='store_true', default=False, help='Copy the log file from the work dir to the output dir at the ' 'end of the program') # Check discovery options locate_options.add_argument( '-c', '--checkpath', action='append', metavar='DIR|FILE', help='Search for checks in DIR or FILE') locate_options.add_argument( '-R', '--recursive', action='store_true', help='Load checks recursively') # Select options select_options.add_argument( '-t', '--tag', action='append', dest='tags', default=[], help='Select checks matching TAG') select_options.add_argument( '-n', '--name', action='append', dest='names', default=[], metavar='NAME', help='Select checks with NAME') select_options.add_argument( '-x', '--exclude', action='append', dest='exclude_names', metavar='NAME', default=[], help='Exclude checks with NAME') select_options.add_argument( '-p', '--prgenv', action='append', default=[], help='Select tests for PRGENV programming environment only') select_options.add_argument( '--gpu-only', action='store_true', help='Select only GPU tests') select_options.add_argument( '--cpu-only', action='store_true', help='Select only CPU tests') # Action options action_options.add_argument( '-l', '--list', action='store_true', help='list matched regression checks') action_options.add_argument( '-r', '--run', action='store_true', help='Run regression with the selected checks') # Run options run_options.add_argument( '-A', '--account', action='store', help='Use ACCOUNT for submitting jobs') run_options.add_argument( '-P', '--partition', action='store', metavar='PART', help='Use PART for submitting jobs') run_options.add_argument( '--reservation', action='store', metavar='RES', help='Use RES for submitting jobs') run_options.add_argument( '--nodelist', action='store', help='Run checks on the selected list of nodes') run_options.add_argument( '--exclude-nodes', action='store', metavar='NODELIST', help='Exclude the list of nodes from runnning checks') run_options.add_argument( '--job-option', action='append', metavar='OPT', dest='job_options', default=[], help='Pass OPT to job scheduler') run_options.add_argument( '--force-local', action='store_true', help='Force local execution of checks') run_options.add_argument( '--skip-sanity-check', action='store_true', help='Skip sanity checking') run_options.add_argument( '--skip-performance-check', action='store_true', help='Skip performance checking') run_options.add_argument( '--strict', action='store_true', help='Force strict performance checking') run_options.add_argument( '--skip-system-check', action='store_true', help='Skip system check') run_options.add_argument( '--skip-prgenv-check', action='store_true', help='Skip prog. environment check') run_options.add_argument( '--exec-policy', metavar='POLICY', action='store', choices=['serial', 'async'], default='serial', help='Specify the execution policy for running the regression tests. ' 'Available policies: "serial" (default), "async"') run_options.add_argument( '--mode', action='store', help='Execution mode to use' ) misc_options.add_argument( '-m', '--module', action='append', default=[], metavar='MOD', dest='user_modules', help='Load module MOD before running the regression') misc_options.add_argument( '--nocolor', action='store_false', dest='colorize', default=True, help='Disable coloring of output') misc_options.add_argument( '--timestamp', action='store', nargs='?', const='%FT%T', metavar='TIMEFMT', help='Append a timestamp component to the regression directories' '(default format "%%FT%%T")' ) misc_options.add_argument( '--system', action='store', help='Load SYSTEM configuration explicitly') misc_options.add_argument('-V', '--version', action='version', version=settings.version) if len(sys.argv) == 1: argparser.print_help() sys.exit(1) # Parse command line options = argparser.parse_args() # Configure logging logging.configure_logging(settings.logging_config) # Setup printer printer = PrettyPrinter() printer.colorize = options.colorize # Load site configuration site_config = SiteConfiguration() site_config.load_from_dict(settings.site_configuration) if options.system: try: sysname, sep, partname = options.system.partition(':') system = site_config.systems[sysname] if partname: # Disable all partitions except partname for p in system.partitions: if p.name != partname: p.disable() if not system.partitions: raise KeyError(options.system) except KeyError: printer.error("unknown system specified: `%s'" % options.system) list_supported_systems(site_config.systems.values(), printer) sys.exit(1) else: # Try to autodetect system system = autodetect_system(site_config) if not system: printer.error("could not auto-detect system. Please specify " "it manually using the `--system' option.") list_supported_systems(site_config.systems.values(), printer) sys.exit(1) # Init modules system init_modules_system(system.modules_system) if options.mode: try: mode_args = site_config.modes[options.mode] # Parse the mode's options and reparse the command-line options = argparser.parse_args(mode_args) options = argparser.parse_args(namespace=options) except KeyError: printer.error("no such execution mode: `%s'" % (options.mode)) sys.exit(1) # Setup the check loader if options.checkpath: load_path = [] for d in options.checkpath: if not os.path.exists(d): printer.info("%s: path `%s' does not exist. Skipping...\n" % (argparser.prog, d)) continue load_path.append(d) loader = RegressionCheckLoader(load_path, recurse=options.recursive) else: loader = RegressionCheckLoader( load_path=settings.checks_path, prefix=os.path.abspath( os.path.join(os.path.dirname(__file__), '..', '..') ), recurse=settings.checks_path_recurse, ) # Adjust system directories if options.prefix: # if prefix is set, reset all other directories system.prefix = os.path.expandvars(options.prefix) system.outputdir = None system.stagedir = None system.logdir = None if options.output: system.outputdir = os.path.expandvars(options.output) if options.stage: system.stagedir = os.path.expandvars(options.stage) if options.logdir: system.logdir = os.path.expandvars(options.logdir) resources = ResourcesManager(prefix=system.prefix, output_prefix=system.outputdir, stage_prefix=system.stagedir, log_prefix=system.logdir, timestamp=options.timestamp) if (os_ext.samefile(resources.stage_prefix, resources.output_prefix) and not options.keep_stage_files): printer.error('stage and output refer to the same directory. ' 'If this is on purpose, please use also the ' "`--keep-stage-files' option.") sys.exit(1) printer.log_config(options) # Print command line printer.info('Command line: %s' % ' '.join(sys.argv)) printer.info('Reframe version: ' + settings.version) printer.info('Launched by user: '******'USER']) printer.info('Launched on host: ' + socket.gethostname()) # Print important paths printer.info('Reframe paths') printer.info('=============') printer.info(' Check prefix : %s' % loader.prefix) printer.info('%03s Check search path : %s' % ('(R)' if loader.recurse else '', "'%s'" % ':'.join(loader.load_path))) printer.info(' Stage dir prefix : %s' % resources.stage_prefix) printer.info(' Output dir prefix : %s' % resources.output_prefix) printer.info(' Logging dir : %s' % resources.log_prefix) try: # Locate and load checks checks_found = loader.load_all(system=system, resources=resources) # Filter checks by name checks_matched = filter( lambda c: c if c.name not in options.exclude_names else None, checks_found ) if options.names: checks_matched = filter( lambda c: c if c.name in options.names else None, checks_matched ) # Filter checks by tags user_tags = set(options.tags) checks_matched = filter( lambda c: c if user_tags.issubset(c.tags) else None, checks_matched ) # Filter checks by prgenv if not options.skip_prgenv_check: checks_matched = filter( lambda c: c if sum([c.supports_progenv(p) for p in options.prgenv]) == len(options.prgenv) else None, checks_matched ) # Filter checks further if options.gpu_only and options.cpu_only: printer.error("options `--gpu-only' and `--cpu-only' " "are mutually exclusive") sys.exit(1) if options.gpu_only: checks_matched = filter( lambda c: c if c.num_gpus_per_node > 0 else None, checks_matched ) elif options.cpu_only: checks_matched = filter( lambda c: c if c.num_gpus_per_node == 0 else None, checks_matched ) checks_matched = [c for c in checks_matched] # Act on checks # Unload regression's module and load user-specified modules get_modules_system().unload_module(settings.module_name) for m in options.user_modules: try: get_modules_system().load_module(m, force=True) except ModuleError: printer.info("Could not load module `%s': Skipping..." % m) success = True if options.list: # List matched checks list_checks(list(checks_matched), printer) elif options.run: # Setup the execution policy if options.exec_policy == 'serial': exec_policy = SerialExecutionPolicy() elif options.exec_policy == 'async': exec_policy = AsynchronousExecutionPolicy() else: # This should not happen, since choices are handled by # argparser printer.error("unknown execution policy `%s': Exiting...") sys.exit(1) exec_policy.skip_system_check = options.skip_system_check exec_policy.force_local = options.force_local exec_policy.strict_check = options.strict exec_policy.skip_environ_check = options.skip_prgenv_check exec_policy.skip_sanity_check = options.skip_sanity_check exec_policy.skip_performance_check = options.skip_performance_check exec_policy.only_environs = options.prgenv exec_policy.keep_stage_files = options.keep_stage_files exec_policy.sched_account = options.account exec_policy.sched_partition = options.partition exec_policy.sched_reservation = options.reservation exec_policy.sched_nodelist = options.nodelist exec_policy.sched_exclude_nodelist = options.exclude_nodes exec_policy.sched_options = options.job_options runner = Runner(exec_policy, printer) try: runner.runall(checks_matched, system) finally: # always print a report if runner.stats.num_failures(): printer.info(runner.stats.failure_report()) success = False else: printer.info('No action specified. Exiting...') printer.info("Try `%s -h' for a list of available actions." % argparser.prog) if not success: sys.exit(1) sys.exit(0) except KeyboardInterrupt: sys.exit(1) except OSError as e: printer.error("`%s': %s" % (e.filename, e.strerror)) sys.exit(1) except Exception as e: printer.error('fatal error: %s\n' % e) traceback.print_exc() sys.exit(1) finally: try: if options.save_log_files: logging.save_log_files(resources.output_prefix) except OSError as e: printer.error("`%s': %s" % (e.filename, e.strerror))