From 4aaddecdd85adf8244b340b393e32151f1ceaddb Mon Sep 17 00:00:00 2001 From: Jouni Malinen Date: Sat, 27 Jul 2019 20:19:28 +0300 Subject: [PATCH] tests: Handle test retries through the same queue This removes the separate rerun step from the parallel-vm.py processing and instead, simply requeues the failed test cases into the same queue that is used for the initial run. This is simpler and more efficient since reruns start as soon as any VM is ready for processing them instead of having to wait for all VMs to complete the first round. Furthermore, this allows VMs to be stopped sooner when no more test cases remain and that is helpful especially with the time travel patches that make the wait-for-next-test step in the VM use all available CPU. Signed-off-by: Jouni Malinen --- tests/hwsim/vm/parallel-vm.py | 124 +++++++++++----------------------- 1 file changed, 38 insertions(+), 86 deletions(-) diff --git a/tests/hwsim/vm/parallel-vm.py b/tests/hwsim/vm/parallel-vm.py index 5f56ffac1..7eb44a92a 100755 --- a/tests/hwsim/vm/parallel-vm.py +++ b/tests/hwsim/vm/parallel-vm.py @@ -1,7 +1,7 @@ #!/usr/bin/env python3 # # Parallel VM test case executor -# Copyright (c) 2014-2018, Jouni Malinen +# Copyright (c) 2014-2019, Jouni Malinen # # This software may be distributed under the terms of the BSD license. # See README for more details. @@ -85,9 +85,10 @@ def get_failed(vm): failed += vm[i]['failed'] return failed -def vm_read_stdout(vm, i): +def vm_read_stdout(vm, i, test_queue): global total_started, total_passed, total_failed, total_skipped global rerun_failures + global first_run_failures ready = False try: @@ -125,6 +126,15 @@ def vm_read_stdout(vm, i): name = vals[1] logger.debug("VM[%d] test case failed: %s" % (i, name)) vm['failed'].append(name) + if name != vm['current_name']: + logger.info("VM[%d] test result mismatch: %s (expected %s)" % (i, name, vm['current_name'])) + else: + count = vm['current_count'] + if count == 0: + first_run_failures.append(name) + if rerun_failures and count < 1: + logger.debug("Requeue test case %s" % name) + test_queue.append((name, vm['current_count'] + 1)) elif line.startswith("NOT-FOUND"): ready = True total_failed += 1 @@ -151,9 +161,11 @@ def show_progress(scr): global tests global first_run_failures global total_started, total_passed, total_failed, total_skipped + global rerun_failures total_tests = len(tests) logger.info("Total tests: %d" % total_tests) + test_queue = [(t, 0) for t in tests] scr.leaveok(1) scr.addstr(0, 0, "Parallel test execution status", curses.A_BOLD) @@ -164,19 +176,11 @@ def show_progress(scr): scr.addstr(num_servers + 1, 20, "TOTAL={} STARTED=0 PASS=0 FAIL=0 SKIP=0".format(total_tests)) scr.refresh() - completed_first_pass = False - rerun_tests = [] - while True: running = False - first_running = False updated = False - for i in range(0, num_servers): - if completed_first_pass: - continue - if vm[i]['first_run_done']: - continue + for i in range(num_servers): if not vm[i]['proc']: continue if vm[i]['proc'].poll() is not None: @@ -188,72 +192,43 @@ def show_progress(scr): if "Kernel panic" in f.read(): scr.addstr("kernel panic") logger.info("VM[%d] kernel panic" % i) - else: + updated = True + if test_queue: scr.addstr("unexpected exit") logger.info("VM[%d] unexpected exit" % i) - updated = True + updated = True continue running = True - first_running = True try: err = vm[i]['proc'].stderr.read() if err != None: err = err.decode() vm[i]['err'] += err - logger.debug("VM[%d] stderr.read[%s]" % (i, err)) + logger.info("VM[%d] stderr.read[%s]" % (i, err)) except IOError as e: if e.errno != errno.EAGAIN: raise - if vm_read_stdout(vm[i], i): + if vm_read_stdout(vm[i], i, test_queue): scr.move(i + 1, 10) scr.clrtoeol() updated = True - if not tests: - vm[i]['first_run_done'] = True - scr.addstr("completed first round") - logger.info("VM[%d] completed first round" % i) + if not test_queue: + vm[i]['proc'].stdin.write(b'\n') + vm[i]['proc'].stdin.flush() + scr.addstr("shutting down") + logger.info("VM[%d] shutting down" % i) continue else: - name = tests.pop(0) + (name, count) = test_queue.pop(0) + vm[i]['current_name'] = name + vm[i]['current_count'] = count vm[i]['proc'].stdin.write(name.encode() + b'\n') vm[i]['proc'].stdin.flush() scr.addstr(name) logger.debug("VM[%d] start test %s" % (i, name)) - if not first_running and not completed_first_pass: - logger.info("First round of testing completed") - if tests: - logger.info("Unexpected test cases remaining from first round: " + str(tests)) - raise Exception("Unexpected test cases remaining from first round") - completed_first_pass = True - for name in get_failed(vm): - if rerun_failures: - rerun_tests.append(name) - first_run_failures.append(name) - - for i in range(num_servers): - if not completed_first_pass: - continue - if not vm[i]['proc']: - continue - if vm[i]['proc'].poll() is not None: - vm[i]['proc'] = None - scr.move(i + 1, 10) - scr.clrtoeol() - log = '{}/{}.srv.{}/console'.format(dir, timestamp, i + 1) - with open(log, 'r') as f: - if "Kernel panic" in f.read(): - scr.addstr("kernel panic") - logger.info("VM[%d] kernel panic" % i) - else: - scr.addstr("completed run") - logger.info("VM[%d] completed run" % i) - updated = True - continue - - running = True try: err = vm[i]['proc'].stderr.read() if err != None: @@ -264,28 +239,6 @@ def show_progress(scr): if e.errno != errno.EAGAIN: raise - ready = False - if vm[i]['first_run_done']: - vm[i]['first_run_done'] = False - ready = True - else: - ready = vm_read_stdout(vm[i], i) - if ready: - scr.move(i + 1, 10) - scr.clrtoeol() - updated = True - if not rerun_tests: - vm[i]['proc'].stdin.write(b'\n') - vm[i]['proc'].stdin.flush() - scr.addstr("shutting down") - logger.info("VM[%d] shutting down" % i) - else: - name = rerun_tests.pop(0) - vm[i]['proc'].stdin.write(name.encode() + b'\n') - vm[i]['proc'].stdin.flush() - scr.addstr(name + "(*)") - logger.debug("VM[%d] start test %s (*)" % (i, name)) - if not running: break @@ -309,19 +262,19 @@ def show_progress(scr): scr.addstr(f) scr.addstr(' ') - scr.move(0, 35) - scr.clrtoeol() - if rerun_tests: - scr.addstr("(RETRY FAILED %d)" % len(rerun_tests)) - elif rerun_failures: - pass - elif first_run_failures: - scr.addstr("(RETRY FAILED)") - scr.refresh() time.sleep(0.25) + for i in range(num_servers): + if not vm[i]['proc']: + continue + vm[i]['proc'] = None + scr.move(i + 1, 10) + scr.clrtoeol() + scr.addstr("still running") + logger.info("VM[%d] still running" % i) + scr.refresh() time.sleep(0.3) @@ -455,7 +408,6 @@ def main(): if args.telnet: cmd += ['--telnet', str(args.telnet + i)] vm[i] = {} - vm[i]['first_run_done'] = False vm[i]['proc'] = subprocess.Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, @@ -537,7 +489,7 @@ def main(): print("Logs: " + dir + '/' + str(timestamp)) logger.info("Logs: " + dir + '/' + str(timestamp)) - for i in range(0, num_servers): + for i in range(num_servers): if len(vm[i]['pending']) > 0: logger.info("Unprocessed stdout from VM[%d]: '%s'" % (i, vm[i]['pending']))