blob: 32e71920be2c52a67f2c277db987541551806138 [file] [log] [blame]
#!/usr/bin/env python
#-------------------------------------------------------------------------------------------------------
# Copyright (C) Microsoft. All rights reserved.
# Licensed under the MIT license. See LICENSE.txt file in the project root for full license information.
#-------------------------------------------------------------------------------------------------------
from __future__ import print_function
from datetime import datetime
from multiprocessing import Pool, Manager
from threading import Timer
import sys
import os
import subprocess as SP
import traceback
import argparse
import xml.etree.ElementTree as ET
import re
import time
# handle command line args
parser = argparse.ArgumentParser(
description='ChakraCore *nix Test Script',
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog='''\
Samples:
test all folders:
runtests.py
test only Array:
runtests.py Array
test a single file:
runtests.py Basics/hello.js
''')
parser.add_argument('folders', metavar='folder', nargs='*',
help='folder subset to run tests')
parser.add_argument('-b', '--binary', metavar='bin', help='ch full path')
parser.add_argument('-d', '--debug', action='store_true',
help='use debug build');
parser.add_argument('-t', '--test', action='store_true', help='use test build')
parser.add_argument('--include-slow', action='store_true',
help='include slow tests')
parser.add_argument('--only-slow', action='store_true',
help='run only slow tests')
parser.add_argument('--nightly', action='store_true',
help='run as nightly tests')
parser.add_argument('--tag', nargs='*',
help='select tests with given tags')
parser.add_argument('--not-tag', nargs='*',
help='exclude tests with given tags')
parser.add_argument('--timeout', type=int, default=60,
help='test timeout (default 60 seconds)')
parser.add_argument('-l', '--logfile', metavar='logfile', help='file to log results to', default=None)
parser.add_argument('--x86', action='store_true', help='use x86 build')
parser.add_argument('--x64', action='store_true', help='use x64 build')
args = parser.parse_args()
test_root = os.path.dirname(os.path.realpath(__file__))
repo_root = os.path.dirname(test_root)
# arch: x86, x64
arch = 'x86' if args.x86 else ('x64' if args.x64 else None)
if arch == None:
arch = os.environ.get('_BuildArch', 'x86')
if sys.platform != 'win32':
arch = 'x64' # xplat: hard code arch == x64
arch_alias = 'amd64' if arch == 'x64' else None
# flavor: debug, test, release
type_flavor = {'chk':'Debug', 'test':'Test', 'fre':'Release'}
flavor = 'Debug' if args.debug else ('Test' if args.test else None)
if flavor == None:
print("ERROR: Test build target wasn't defined.")
print("Try '-t' (test build) or '-d' (debug build).")
sys.exit(1)
flavor_alias = 'chk' if flavor == 'Debug' else 'fre'
# binary: full ch path
binary = args.binary
if binary == None:
if sys.platform == 'win32':
binary = 'Build/VcBuild/bin/{}_{}/ch.exe'.format(arch, flavor)
else:
binary = 'BuildLinux/{0}/ch'.format(flavor)
binary = os.path.join(repo_root, binary)
if not os.path.isfile(binary):
print('{} not found. Did you run ./build.sh already?'.format(binary))
sys.exit(1)
# global tags/not_tags
tags = set(args.tag or [])
not_tags = set(args.not_tag or []).union(['fail', 'exclude_' + arch])
if arch_alias:
not_tags.add('exclude_' + arch_alias)
if flavor_alias:
not_tags.add('exclude_' + flavor_alias)
if args.only_slow:
tags.add('Slow')
elif not args.include_slow:
not_tags.add('Slow')
not_tags.add('exclude_nightly' if args.nightly else 'nightly')
# xplat: temp hard coded to exclude unsupported tests
if sys.platform != 'win32':
not_tags.add('exclude_xplat')
not_tags.add('Intl')
not_tags.add('require_backend')
not_tags.add('require_debugger')
if sys.platform == 'darwin':
not_tags.add('exclude_mac')
not_compile_flags = set(['-simdjs']) \
if sys.platform != 'win32' else None
# use tags/not_tags/not_compile_flags as case-insensitive
def lower_set(s):
return set([x.lower() for x in s] if s else [])
tags = lower_set(tags)
not_tags = lower_set(not_tags)
not_compile_flags = lower_set(not_compile_flags)
# split tags text into tags set
_empty_set = set()
def split_tags(text):
return set(x.strip() for x in text.lower().split(',')) if text \
else _empty_set
class LogFile(object):
def __init__(self, log_file_path = None):
self.file = None
if log_file_path is None:
# Set up the log file paths
# Make sure the right directory exists and the log file doesn't
log_file_name = "testrun.{0}{1}.log".format(arch, flavor)
log_file_directory = os.path.join(repo_root, "test", "logs")
if not os.path.exists(log_file_directory):
os.mkdir(log_file_directory)
self.log_file_path = os.path.join(log_file_directory, log_file_name)
if os.path.exists(self.log_file_path):
os.remove(self.log_file_path)
else:
self.log_file_path = log_file_path
self.file = open(self.log_file_path, "w")
def log(self, args):
self.file.write(args)
def __del__(self):
if not (self.file is None):
self.file.close()
if __name__ == '__main__':
log_file = LogFile(args.logfile)
def log_message(msg = ""):
log_file.log(msg + "\n")
def print_and_log(msg = ""):
print(msg)
log_message(msg)
# remove carriage returns at end of line to avoid platform difference
def normalize_new_line(text):
return re.sub(b'[\r]+\n', b'\n', text)
# A test simply contains a collection of test attributes.
# Misc attributes added by test run:
# filename full path of test file
# elapsed_time elapsed time when running the test
#
class Test(dict):
__setattr__ = dict.__setitem__
__delattr__ = dict.__delitem__
# support dot syntax for normal attribute access
def __getattr__(self, key):
return super(Test, self).__getattr__(key) if key.startswith('__') \
else self.get(key)
# mark start of this test run, to compute elapsed_time
def start(self):
self.start_time = datetime.now()
# mark end of this test run, compute elapsed_time
def done(self):
if not self.elapsed_time:
self.elapsed_time = (datetime.now() - self.start_time)\
.total_seconds()
# records pass_count/fail_count
class PassFailCount(object):
def __init__(self):
self.pass_count = 0
self.fail_count = 0
def __str__(self):
return 'passed {}, failed {}'.format(self.pass_count, self.fail_count)
def total_count(self):
return self.pass_count + self.fail_count
# records total and individual folder's pass_count/fail_count
class TestResult(PassFailCount):
def __init__(self):
super(self.__class__, self).__init__()
self.folders = {}
def _get_folder_result(self, folder):
r = self.folders.get(folder)
if not r:
r = PassFailCount()
self.folders[folder] = r
return r
def log(self, filename, fail=False):
folder = os.path.basename(os.path.dirname(filename))
r = self._get_folder_result(folder)
if fail:
r.fail_count += 1
self.fail_count += 1
else:
r.pass_count += 1
self.pass_count += 1
# test variants:
# interpreted: -maxInterpretCount:1 -maxSimpleJitRunCount:1 -bgjit-
# dynapogo: -forceNative -off:simpleJit -bgJitDelay:0
class TestVariant(object):
def __init__(self, name, compile_flags=[]):
self.name = name
self.compile_flags = \
['-WERExceptionSupport', '-ExtendedErrorStackForTestHost',
'-BaselineMode'] + compile_flags
self.tags = tags.copy()
self.not_tags = not_tags.union(
['{}_{}'.format(x, name) for x in ('fails','exclude')])
self.msg_queue = Manager().Queue() # messages from multi processes
self.test_result = TestResult()
self._print_lines = [] # _print lines buffer
self._last_len = 0
# check if this test variant should run a given test
def _should_test(self, test):
tags = split_tags(test.get('tags'))
if not tags.isdisjoint(self.not_tags):
return False
if self.tags and not self.tags.issubset(tags):
return False
if not_compile_flags: # exclude unsupported compile-flags if any
flags = test.get('compile-flags')
if flags and \
not not_compile_flags.isdisjoint(flags.lower().split()):
return False
return True
# print output from multi-process run, to be sent with result message
def _print(self, line):
self._print_lines.append(str(line))
# queue a test result from multi-process runs
def _log_result(self, test, fail):
output = '\n'.join(self._print_lines) # collect buffered _print output
self._print_lines = []
self.msg_queue.put((test.filename, fail, test.elapsed_time, output))
# (on main process) process one queued message
def _process_msg(self, msg):
filename, fail, elapsed_time, output = msg
self.test_result.log(filename, fail=fail)
line = '[{}/{} {:4.2f}] {} -> {}'.format(
self.test_result.total_count(),
self.test_count,
elapsed_time,
'Failed' if fail else 'Passed',
self._short_name(filename))
padding = self._last_len - len(line)
print(line + ' ' * padding, end='\n' if fail else '\r')
log_message(line)
self._last_len = len(line) if not fail else 0
if len(output) > 0:
print_and_log(output)
# get a shorter test file path for display only
def _short_name(self, filename):
folder = os.path.basename(os.path.dirname(filename))
return os.path.join(folder, os.path.basename(filename))
# (on main process) wait and process one queued message
def _process_one_msg(self):
self._process_msg(self.msg_queue.get())
# log a failed test with details
def _show_failed(self, test, flags, exit_code, output,
expected_output=None, timedout=False):
if timedout:
self._print('ERROR: Test timed out!')
self._print('{} {} {}'.format(binary, ' '.join(flags), test.filename))
if expected_output == None or timedout:
self._print("\nOutput:")
self._print("----------------------------")
self._print(output)
self._print("----------------------------")
else:
lst_output = output.split(b'\n')
lst_expected = expected_output.split(b'\n')
ln = min(len(lst_output), len(lst_expected))
for i in range(0, ln):
if lst_output[i] != lst_expected[i]:
self._print("Output: (at line " + str(i) + ")")
self._print("----------------------------")
self._print(lst_output[i])
self._print("----------------------------")
self._print("Expected Output:")
self._print("----------------------------")
self._print(lst_expected[i])
self._print("----------------------------")
break
self._print("exit code: {}".format(exit_code))
self._log_result(test, fail=True)
# temp: try find real file name on hard drive if case mismatch
def _check_file(self, folder, filename):
path = os.path.join(folder, filename)
if os.path.isfile(path):
return path # file exists on disk
filename_lower = filename.lower()
files = os.listdir(folder)
for i in range(len(files)):
if files[i].lower() == filename_lower:
self._print('\nWARNING: {} should be {}\n'.format(
path, files[i]))
return os.path.join(folder, files[i])
# cann't find the file, just return the path and let it error out
return path
# run one test under this variant
def test_one(self, test):
try:
test.start()
self._run_one_test(test)
except Exception:
test.done()
self._print(traceback.format_exc())
self._log_result(test, fail=True)
# internally perform one test run
def _run_one_test(self, test):
folder = test.folder
js_file = test.filename = self._check_file(folder, test.files)
js_output = b''
working_path = os.path.dirname(js_file)
flags = test.get('compile-flags')
flags = self.compile_flags + (flags.split() if flags else [])
cmd = [binary] + flags + [os.path.basename(js_file)]
test.start()
proc = SP.Popen(cmd, stdout=SP.PIPE, stderr=SP.STDOUT, cwd=working_path)
timeout_data = [proc, False]
def timeout_func(timeout_data):
timeout_data[0].kill()
timeout_data[1] = True
timeout = test.get('timeout', args.timeout) # test override or default
timer = Timer(timeout, timeout_func, [timeout_data])
try:
timer.start()
js_output = normalize_new_line(proc.communicate()[0])
exit_code = proc.wait()
finally:
timer.cancel()
test.done()
# shared _show_failed args
fail_args = { 'test': test, 'flags': flags,
'exit_code': exit_code, 'output': js_output };
# check timed out
if (timeout_data[1]):
return self._show_failed(timedout=True, **fail_args)
# check ch failed
if exit_code != 0:
return self._show_failed(**fail_args)
# check output
if 'baseline' not in test:
# output lines must be 'pass' or 'passed' or empty
lines = (line.lower() for line in js_output.split(b'\n'))
if any(line != b'' and line != b'pass' and line != b'passed'
for line in lines):
return self._show_failed(**fail_args)
else:
baseline = test.get('baseline')
if baseline:
# perform baseline comparison
baseline = self._check_file(working_path, baseline)
with open(baseline, 'rb') as bs_file:
baseline_output = bs_file.read()
# Cleanup carriage return
# todo: remove carriage return at the end of the line
# or better fix ch to output same on all platforms
expected_output = normalize_new_line(baseline_output)
if expected_output != js_output:
return self._show_failed(
expected_output=expected_output, **fail_args)
# passed
self._log_result(test, fail=False)
# run tests under this variant, using given multiprocessing Pool
def run(self, tests, pool):
print_and_log('\n############# Starting {} variant #############'\
.format(self.name))
if self.tags:
print_and_log(' tags: {}'.format(self.tags))
for x in self.not_tags:
print_and_log(' exclude: {}'.format(x))
print_and_log()
# filter tests to run
tests = [x for x in tests if self._should_test(x)]
self.test_count = len(tests)
# run tests in parallel
result = pool.map_async(run_one, [(self,test) for test in tests])
while self.test_result.total_count() != self.test_count:
self._process_one_msg()
# print test result summary
def print_summary(self):
print_and_log('\n######## Logs for {} variant ########'\
.format(self.name))
for folder, result in sorted(self.test_result.folders.items()):
print_and_log('{}: {}'.format(folder, result))
print_and_log("----------------------------")
print_and_log('Total: {}'.format(self.test_result))
# global run one test function for multiprocessing, used by TestVariant
def run_one(data):
try:
variant, test = data
variant.test_one(test)
except Exception:
print('ERROR: Unhandled exception!!!')
traceback.print_exc()
# record folder/tags info from test_root/rlexedirs.xml
class FolderTags(object):
def __init__(self):
xmlpath = os.path.join(test_root, 'rlexedirs.xml')
try:
xml = ET.parse(xmlpath).getroot()
except IOError:
print_and_log('ERROR: failed to read {}'.format(xmlpath))
exit(-1)
self._folder_tags = {}
for x in xml:
d = x.find('default')
key = d.find('files').text.lower() # avoid case mismatch
tags = d.find('tags')
self._folder_tags[key] = \
split_tags(tags.text) if tags != None else _empty_set
# check if should test a given folder
def should_test(self, folder):
key = os.path.basename(os.path.normpath(folder)).lower()
ftags = self._folder_tags.get(key)
# folder listed in rlexedirs.xml and not exlucded by global not_tags
return ftags != None and ftags.isdisjoint(not_tags)
# load all tests in folder using rlexe.xml file
def load_tests(folder, file):
try:
xmlpath = os.path.join(folder, 'rlexe.xml')
xml = ET.parse(xmlpath).getroot()
except IOError:
return []
def test_override(condition, check_tag, check_value, test):
target = condition.find(check_tag)
if target != None and target.text == check_value:
for override in condition.find('override'):
test[override.tag] = override.text
def load_test(testXml):
test = Test(folder=folder)
for c in testXml.find('default'):
if c.tag == 'timeout': # timeout seconds
test[c.tag] = int(c.text)
elif c.tag == 'tags' and c.tag in test: # merge multiple <tags>
test[c.tag] = test[c.tag] + ',' + c.text
else:
test[c.tag] = c.text
condition = testXml.find('condition')
if condition != None:
test_override(condition, 'target', arch_alias, test)
return test
tests = [load_test(x) for x in xml]
if file != None:
tests = [x for x in tests if x.files == file]
if len(tests) == 0 and is_jsfile(file):
tests = [Test(folder=folder, files=file, baseline='')]
return tests
def is_jsfile(path):
return os.path.splitext(path)[1] == '.js'
def main():
# By default run all tests
if len(args.folders) == 0:
files = (os.path.join(test_root, x) for x in os.listdir(test_root))
args.folders = [f for f in sorted(files) if not os.path.isfile(f)]
# Set the right timezone, the tests need Pacific Standard Time
os.environ['TZ'] = 'US/Pacific'
time.tzset()
# load all tests
tests = []
folder_tags = FolderTags()
for path in args.folders:
if os.path.isfile(path):
folder, file = os.path.dirname(path), os.path.basename(path)
else:
folder, file = path, None
if folder_tags.should_test(folder):
tests += load_tests(folder, file)
# test variants
variants = [
TestVariant('interpreted', [
'-maxInterpretCount:1', '-maxSimpleJitRunCount:1', '-bgjit-'])
]
# run each variant
pool = Pool(1) # Use a multiprocessing process Pool -- TODO: forcing this to single threaded until we have support for sequential tag
start_time = datetime.now()
for variant in variants:
variant.run(tests, pool)
elapsed_time = datetime.now() - start_time
# print summary
for variant in variants:
variant.print_summary()
print()
failed = any(variant.test_result.fail_count > 0 for variant in variants)
print('[{}] {}'.format(
str(elapsed_time), 'Success!' if not failed else 'Failed!'))
return 1 if failed else 0
if __name__ == '__main__':
sys.exit(main())