dart-sdk/tools/utils.py
Clement Skau 0ce83987d6 Reland "[SDK] Adds an SDK hash to kernels and the VM."
Note: This is a reland of https://dart-review.googlesource.com/c/sdk/+/150343

Adds a new SDK hash to kernels and the VM which is optionally checked
to verify kernels are built for the same SDK as the VM.
This helps catch incompatibilities that are currently causing
subtle bugs and (not so subtle) crashes.

The SDK hash is encoded in kernels as a new field in components.
The hash is derived from the 10 byte git short hash.

This new check can be disabled via:
  tools/gn.py ... --no-verify-sdk-hash

This CL bumps the min. (and max.) supported kernel format version,
making the VM backwards incompatible from this point back.

This also bumps the min. and current ABI version.

Bug: https://github.com/dart-lang/sdk/issues/41802
Change-Id: I2f85945045a603eb9dcfd1f2c0d0d024bd84a956
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/152802
Commit-Queue: Clement Skau <cskau@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
2020-07-07 06:08:20 +00:00

1264 lines
41 KiB
Python

# Copyright (c) 2012, the Dart project authors. Please see the AUTHORS file
# for details. All rights reserved. Use of this source code is governed by a
# BSD-style license that can be found in the LICENSE file.
# This file contains a set of utilities functions used by other Python-based
# scripts.
from __future__ import print_function
import contextlib
import datetime
import glob
import imp
import json
import os
import platform
import re
import shutil
import subprocess
import sys
import tarfile
import tempfile
import uuid
try:
# Not available on Windows.
import resource
except:
pass
# To eliminate clashing with older archived builds on bleeding edge we add
# a base number bigger the largest svn revision (this also gives us an easy
# way of seeing if an archive comes from git based or svn based commits).
GIT_NUMBER_BASE = 100000
# Mapping table between build mode and build configuration.
BUILD_MODES = {
'debug': 'Debug',
'release': 'Release',
'product': 'Product',
}
# Mapping table between build mode and build configuration.
BUILD_SANITIZERS = {
None: '',
'none': '',
'asan': 'ASAN',
'lsan': 'LSAN',
'msan': 'MSAN',
'tsan': 'TSAN',
'ubsan': 'UBSAN',
}
# Mapping table between OS and build output location.
BUILD_ROOT = {
'win32': 'out',
'linux': 'out',
'freebsd': 'out',
'macos': 'xcodebuild',
}
# Note: gn expects these to be lower case.
ARCH_FAMILY = {
'ia32': 'ia32',
'x64': 'ia32',
'arm': 'arm',
'armv6': 'arm',
'arm64': 'arm',
'arm_x64': 'arm',
'simarm': 'ia32',
'simarmv6': 'ia32',
'simarm64': 'ia32',
'simarm_x64': 'ia32',
}
BASE_DIR = os.path.abspath(os.path.join(os.curdir, '..'))
DART_DIR = os.path.abspath(os.path.join(__file__, '..', '..'))
VERSION_FILE = os.path.join(DART_DIR, 'tools', 'VERSION')
def GetArchFamily(arch):
return ARCH_FAMILY[arch]
def GetBuildDir(host_os):
return BUILD_ROOT[host_os]
def GetBuildMode(mode):
return BUILD_MODES[mode]
def GetBuildSanitizer(sanitizer):
return BUILD_SANITIZERS[sanitizer]
def GetBaseDir():
return BASE_DIR
def GetBotUtils(repo_path=DART_DIR):
'''Dynamically load the tools/bots/bot_utils.py python module.'''
return imp.load_source(
'bot_utils', os.path.join(repo_path, 'tools', 'bots', 'bot_utils.py'))
def GetMinidumpUtils(repo_path=DART_DIR):
'''Dynamically load the tools/minidump.py python module.'''
return imp.load_source('minidump',
os.path.join(repo_path, 'tools', 'minidump.py'))
class Version(object):
def __init__(self, channel, major, minor, patch, prerelease,
prerelease_patch, abi_version, oldest_supported_abi_version):
self.channel = channel
self.major = major
self.minor = minor
self.patch = patch
self.prerelease = prerelease
self.prerelease_patch = prerelease_patch
self.abi_version = abi_version
self.oldest_supported_abi_version = oldest_supported_abi_version
# Try to guess the host operating system.
def GuessOS():
os_id = platform.system()
if os_id == 'Linux':
return 'linux'
elif os_id == 'Darwin':
return 'macos'
elif os_id == 'Windows' or os_id == 'Microsoft':
# On Windows Vista platform.system() can return 'Microsoft' with some
# versions of Python, see http://bugs.python.org/issue1082 for details.
return 'win32'
elif os_id == 'FreeBSD':
return 'freebsd'
elif os_id == 'OpenBSD':
return 'openbsd'
elif os_id == 'SunOS':
return 'solaris'
return None
# Try to guess the host architecture.
def GuessArchitecture():
os_id = platform.machine()
if os_id.startswith('armv6'):
return 'armv6'
elif os_id.startswith('arm'):
return 'arm'
elif os_id.startswith('aarch64'):
return 'arm64'
elif '64' in os_id:
return 'x64'
elif (not os_id) or (not re.match('(x|i[3-6])86', os_id) is None):
return 'ia32'
elif os_id == 'i86pc':
return 'ia32'
guess_os = GuessOS()
print('Warning: Guessing architecture {} based on os {}\n'.format(
os_id, guess_os))
if guess_os == 'win32':
return 'ia32'
return None
# Try to guess the number of cpus on this machine.
def GuessCpus():
if os.getenv('DART_NUMBER_OF_CORES') is not None:
return int(os.getenv('DART_NUMBER_OF_CORES'))
if os.path.exists('/proc/cpuinfo'):
return int(
subprocess.check_output(
'grep -E \'^processor\' /proc/cpuinfo | wc -l', shell=True))
if os.path.exists('/usr/bin/hostinfo'):
return int(
subprocess.check_output(
'/usr/bin/hostinfo |'
' grep "processors are logically available." |'
' awk "{ print \$1 }"',
shell=True))
win_cpu_count = os.getenv("NUMBER_OF_PROCESSORS")
if win_cpu_count:
return int(win_cpu_count)
return 2
def GetWindowsRegistryKeyName(name):
import win32process
# Check if python process is 64-bit or if it's 32-bit running in 64-bit OS.
# We need to know whether host is 64-bit so that we are looking in right
# registry for Visual Studio path.
if sys.maxsize > 2**32 or win32process.IsWow64Process():
wow6432Node = 'Wow6432Node\\'
else:
wow6432Node = ''
return r'SOFTWARE\{}{}'.format(wow6432Node, name)
# Try to guess Visual Studio location when buiding on Windows.
def GuessVisualStudioPath():
defaultPath = r'C:\Program Files (x86)\Microsoft Visual Studio 14.0\Common7' \
r'\IDE'
defaultExecutable = 'devenv.com'
if not IsWindows():
return (defaultPath, defaultExecutable)
keyNamesAndExecutables = [
# Pair for non-Express editions.
(GetWindowsRegistryKeyName(r'Microsoft\VisualStudio'), 'devenv.com'),
# Pair for 2012 Express edition.
(GetWindowsRegistryKeyName(r'Microsoft\VSWinExpress'),
'VSWinExpress.exe'),
# Pair for pre-2012 Express editions.
(GetWindowsRegistryKeyName(r'Microsoft\VCExpress'), 'VCExpress.exe')
]
bestGuess = (0.0, (defaultPath, defaultExecutable))
import _winreg
for (keyName, executable) in keyNamesAndExecutables:
try:
key = _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, keyName)
except WindowsError:
# Can't find this key - moving on the next one.
continue
try:
subkeyCounter = 0
while True:
try:
subkeyName = _winreg.EnumKey(key, subkeyCounter)
subkeyCounter += 1
except WindowsError:
# Reached end of enumeration. Moving on the next key.
break
match = re.match(r'^\d+\.\d+$', subkeyName)
if match:
with _winreg.OpenKey(key, subkeyName) as subkey:
try:
(installDir, registrytype) = _winreg.QueryValueEx(
subkey, 'InstallDir')
except WindowsError:
# Can't find value under the key - continue to the next key.
continue
isExpress = executable != 'devenv.com'
if not isExpress and subkeyName == '14.0':
# Stop search since if we found non-Express VS2015 version
# installed, which is preferred version.
return installDir, executable
version = float(subkeyName)
# We prefer higher version of Visual Studio and given equal
# version numbers we prefer non-Express edition.
if version > bestGuess[0]:
bestGuess = (version, (installDir, executable))
finally:
_winreg.CloseKey(key)
return bestGuess[1]
# Returns true if we're running under Windows.
def IsWindows():
return GuessOS() == 'win32'
# Reads a text file into an array of strings - one for each
# line. Strips comments in the process.
def ReadLinesFrom(name):
result = []
for line in open(name):
if '#' in line:
line = line[:line.find('#')]
line = line.strip()
if len(line) == 0:
continue
result.append(line)
return result
# Filters out all arguments until the next '--' argument
# occurs.
def ListArgCallback(option, value, parser):
if value is None:
value = []
for arg in parser.rargs:
if arg[:2].startswith('--'):
break
value.append(arg)
del parser.rargs[:len(value)]
setattr(parser.values, option.dest, value)
# Filters out all argument until the first non '-' or the
# '--' argument occurs.
def ListDartArgCallback(option, value, parser):
if value is None:
value = []
for arg in parser.rargs:
if arg[:2].startswith('--') or arg[0] != '-':
break
value.append(arg)
del parser.rargs[:len(value)]
setattr(parser.values, option.dest, value)
def IsCrossBuild(target_os, arch):
host_arch = GuessArchitecture()
return ((GetArchFamily(host_arch) != GetArchFamily(arch)) or
(target_os != GuessOS()))
def GetBuildConf(mode, arch, conf_os=None, sanitizer=None):
if conf_os is not None and conf_os != GuessOS() and conf_os != 'host':
return '{}{}{}'.format(GetBuildMode(mode), conf_os.title(),
arch.upper())
# Ask for a cross build if the host and target architectures don't match.
host_arch = GuessArchitecture()
cross_build = ''
if GetArchFamily(host_arch) != GetArchFamily(arch):
cross_build = 'X'
return '{}{}{}{}'.format(GetBuildMode(mode), GetBuildSanitizer(sanitizer),
cross_build, arch.upper())
def GetBuildRoot(host_os, mode=None, arch=None, target_os=None, sanitizer=None):
build_root = GetBuildDir(host_os)
if mode:
build_root = os.path.join(
build_root, GetBuildConf(mode, arch, target_os, sanitizer))
return build_root
def GetBuildSdkBin(host_os, mode=None, arch=None, target_os=None):
build_root = GetBuildRoot(host_os, mode, arch, target_os)
return os.path.join(build_root, 'dart-sdk', 'bin')
def GetShortVersion(version_file=None):
version = ReadVersionFile(version_file)
return ('{}.{}.{}.{}.{}'.format(version.major, version.minor, version.patch,
version.prerelease,
version.prerelease_patch))
def GetSemanticSDKVersion(no_git_hash=False,
version_file=None,
git_revision_file=None):
version = ReadVersionFile(version_file)
if not version:
return None
suffix = ''
if version.channel == 'be':
suffix = '-edge' if no_git_hash else '-edge.{}'.format(
GetGitRevision(git_revision_file))
elif version.channel in ('beta', 'dev'):
suffix = '-{}.{}.{}'.format(version.prerelease,
version.prerelease_patch, version.channel)
else:
assert version.channel == 'stable'
return '{}.{}.{}{}'.format(version.major, version.minor, version.patch,
suffix)
def GetVersion(no_git_hash=False, version_file=None, git_revision_file=None):
return GetSemanticSDKVersion(no_git_hash, version_file, git_revision_file)
# The editor used to produce the VERSION file put on gcs. We now produce this
# in the bots archiving the sdk.
# The content looks like this:
#{
# "date": "2015-05-28",
# "version": "1.11.0-edge.131653",
# "revision": "535394c2657ede445142d8a92486d3899bbf49b5"
#}
def GetVersionFileContent():
result = {
'date': str(datetime.date.today()),
'version': GetVersion(),
'revision': GetGitRevision()
}
return json.dumps(result, indent=2)
def GetChannel(version_file=None):
version = ReadVersionFile(version_file)
return version.channel
def GetUserName():
key = 'USER'
if sys.platform == 'win32':
key = 'USERNAME'
return os.environ.get(key, '')
def GetAbiVersion(version_file=None):
version = ReadVersionFile(version_file)
return version.abi_version
def GetOldestSupportedAbiVersion(version_file=None):
version = ReadVersionFile(version_file)
return version.oldest_supported_abi_version
def ReadVersionFile(version_file=None):
def match_against(pattern, file_content):
match = re.search(pattern, file_content, flags=re.MULTILINE)
if match:
return match.group(1)
return None
if version_file == None:
version_file = VERSION_FILE
content = None
try:
with open(version_file) as fd:
content = fd.read()
except:
print('Warning: Could not read VERSION file ({})'.format(version_file))
return None
channel = match_against('^CHANNEL ([A-Za-z0-9]+)$', content)
major = match_against('^MAJOR (\d+)$', content)
minor = match_against('^MINOR (\d+)$', content)
patch = match_against('^PATCH (\d+)$', content)
prerelease = match_against('^PRERELEASE (\d+)$', content)
prerelease_patch = match_against('^PRERELEASE_PATCH (\d+)$', content)
abi_version = match_against('^ABI_VERSION (\d+)$', content)
oldest_supported_abi_version = match_against(
'^OLDEST_SUPPORTED_ABI_VERSION (\d+)$', content)
if (channel and major and minor and prerelease and prerelease_patch and
abi_version and oldest_supported_abi_version):
return Version(channel, major, minor, patch, prerelease,
prerelease_patch, abi_version,
oldest_supported_abi_version)
print('Warning: VERSION file ({}) has wrong format'.format(version_file))
return None
# Our schema for releases and archiving is based on an increasing
# sequence of numbers. In the svn world this was simply the revision of a
# commit, which would always give us a one to one mapping between the number
# and the commit. This was true across branches as well, so a number used
# to archive a build was always unique and unambiguous.
# In git there is no such global number, so we loosen the requirement a bit.
# We only use numbers on the master branch (bleeding edge). On branches
# we use the version number instead for archiving purposes.
# The number on master is the count of commits on the master branch.
def GetArchiveVersion(version_file=None):
version = ReadVersionFile(version_file=None)
if not version:
raise 'Could not get the archive version, parsing the version file failed'
if version.channel in ['be', 'integration']:
return GetGitNumber()
return GetSemanticSDKVersion()
def GetGitRevision(git_revision_file=None, repo_path=DART_DIR):
# When building from tarball use tools/GIT_REVISION
if git_revision_file is None:
git_revision_file = os.path.join(repo_path, 'tools', 'GIT_REVISION')
try:
with open(git_revision_file) as fd:
return fd.read().decode('utf-8').strip()
except:
pass
p = subprocess.Popen(['git', 'rev-parse', 'HEAD'],
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
shell=IsWindows(),
cwd=repo_path)
output, _ = p.communicate()
revision = output.decode('utf-8').strip()
# We expect a full git hash
if len(revision) != 40:
print('Warning: Could not parse git commit, output was {}'.format(
revision),
file=sys.stderr)
return None
return revision
def GetShortGitHash(repo_path=DART_DIR):
p = subprocess.Popen(['git', 'rev-parse', '--short', 'HEAD'],
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
shell=IsWindows(),
cwd=repo_path)
output, _ = p.communicate()
revision = output.decode('utf-8').strip()
if p.wait() != 0:
return None
return revision
def GetLatestDevTag(repo_path=DART_DIR):
# We used the old, pre-git2.13 refname:strip here since lstrip will fail on
# older git versions. strip is an alias for lstrip in later versions.
cmd = [
'git',
'for-each-ref',
'refs/tags/*dev*',
'--sort=-taggerdate',
"--format=%(refname:strip=2)",
'--count=1',
]
p = subprocess.Popen(cmd,
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
shell=IsWindows(),
cwd=repo_path)
output, _ = p.communicate()
tag = output.decode('utf-8').strip()
if p.wait() != 0:
print('Warning: Could not get the most recent dev branch tag {}'.format(
tag),
file=sys.stderr)
return None
return tag
def GetGitTimestamp(repo_path=DART_DIR):
p = subprocess.Popen(['git', 'log', '-n', '1', '--pretty=format:%cd'],
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
shell=IsWindows(),
cwd=repo_path)
output, _ = p.communicate()
timestamp = output.decode('utf-8').strip()
if p.wait() != 0:
return None
return timestamp
def GetGitNumber(repo_path=DART_DIR):
p = subprocess.Popen(['git', 'rev-list', 'HEAD', '--count'],
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
shell=IsWindows(),
cwd=repo_path)
output, _ = p.communicate()
number = output.decode('utf-8').strip()
try:
number = int(number)
return number + GIT_NUMBER_BASE
except:
print(
'Warning: Could not parse git count, output was {}'.format(number),
file=sys.stderr)
return None
def ParseGitInfoOutput(output):
"""Given a git log, determine the latest corresponding svn revision."""
for line in output.split('\n'):
tokens = line.split()
if len(tokens) > 0 and tokens[0] == 'git-svn-id:':
return tokens[1].split('@')[1]
return None
def ParseSvnInfoOutput(output):
revision_match = re.search('Last Changed Rev: (\d+)', output)
if revision_match:
return revision_match.group(1)
return None
def RewritePathSeparator(path, workspace):
# Paths in test files are always specified using '/'
# as the path separator. Replace with the actual
# path separator before use.
if '/' in path:
split_path = path.split('/')
path = os.sep.join(split_path)
path = os.path.join(workspace, path)
if not os.path.exists(path):
raise Exception(path)
return path
def ParseTestOptions(pattern, source, workspace):
match = pattern.search(source)
if match:
return [
RewritePathSeparator(o, workspace)
for o in match.group(1).split(' ')
]
return None
def ParseTestOptionsMultiple(pattern, source, workspace):
matches = pattern.findall(source)
if matches:
result = []
for match in matches:
if len(match) > 0:
result.append([
RewritePathSeparator(o, workspace) for o in match.split(' ')
])
else:
result.append([])
return result
return None
def Daemonize():
"""
Create a detached background process (daemon). Returns True for
the daemon, False for the parent process.
See: http://www.faqs.org/faqs/unix-faq/programmer/faq/
"1.7 How do I get my program to act like a daemon?"
"""
if os.fork() > 0:
return False
os.setsid()
if os.fork() > 0:
exit(0)
# TODO: What is this supposed to do? Didn't we already exit?
raise
return True
def CheckedUnlink(name):
"""Unlink a file without throwing an exception."""
try:
os.unlink(name)
except OSError as e:
sys.stderr.write('os.unlink() ' + str(e))
sys.stderr.write('\n')
# TODO(42528): Can we remove this? It's basically just an alias for Exception.
class Error(Exception):
pass
class ToolError(Exception):
"""Deprecated exception, use Error instead."""
def __init__(self, value):
self.value = value
def __str__(self):
return repr(self.value)
def IsCrashExitCode(exit_code):
if IsWindows():
return 0x80000000 & exit_code
return exit_code < 0
def DiagnoseExitCode(exit_code, command):
if IsCrashExitCode(exit_code):
sys.stderr.write(
'Command: {}\nCRASHED with exit code {} (0x{:x})\n'.format(
' '.join(command), exit_code, exit_code & 0xffffffff))
def Touch(name):
with file(name, 'a'):
os.utime(name, None)
def ExecuteCommand(cmd):
"""Execute a command in a subprocess."""
print('Executing: ' + ' '.join(cmd))
pipe = subprocess.Popen(cmd,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
shell=IsWindows())
output = pipe.communicate()
if pipe.returncode != 0:
raise Exception('Execution failed: ' + str(output))
return pipe.returncode, output
# The checked-in SDKs are documented at
# https://github.com/dart-lang/sdk/wiki/The-checked-in-SDK-in-tools
def CheckedInSdkPath():
# We don't use the normal macos, linux, win32 directory names here, instead,
# we use the names that the download_from_google_storage script uses.
osdict = {'Darwin': 'mac', 'Linux': 'linux', 'Windows': 'win'}
system = platform.system()
try:
osname = osdict[system]
except KeyError:
sys.stderr.write(
'WARNING: platform "{}" not supported\n'.format(system))
return None
tools_dir = os.path.dirname(os.path.realpath(__file__))
return os.path.join(tools_dir, 'sdks', 'dart-sdk')
def CheckedInSdkExecutable():
name = 'dart'
if IsWindows():
name = 'dart.exe'
return os.path.join(CheckedInSdkPath(), 'bin', name)
def CheckedInSdkCheckExecutable():
executable = CheckedInSdkExecutable()
canary_script = os.path.join(
os.path.dirname(os.path.realpath(__file__)), 'canary.dart')
try:
with open(os.devnull, 'wb') as silent_sink:
if 0 == subprocess.call([executable, canary_script],
stdout=silent_sink):
return True
except OSError as e:
pass
return False
def CheckLinuxCoreDumpPattern(fatal=False):
core_pattern_file = '/proc/sys/kernel/core_pattern'
core_pattern = open(core_pattern_file).read()
expected_core_pattern = 'core.%p'
if core_pattern.strip() != expected_core_pattern:
message = (
'Invalid core_pattern configuration. '
'The configuration of core dump handling is *not* correct for '
'a buildbot. The content of {0} must be "{1}" instead of "{2}".'.
format(core_pattern_file, expected_core_pattern, core_pattern))
if fatal:
raise Exception(message)
print(message)
return False
return True
class TempDir(object):
def __init__(self, prefix=''):
self._temp_dir = None
self._prefix = prefix
def __enter__(self):
self._temp_dir = tempfile.mkdtemp(self._prefix)
return self._temp_dir
def __exit__(self, *_):
shutil.rmtree(self._temp_dir, ignore_errors=True)
class ChangedWorkingDirectory(object):
def __init__(self, working_directory):
self._working_directory = working_directory
def __enter__(self):
self._old_cwd = os.getcwd()
print('Enter directory = ', self._working_directory)
os.chdir(self._working_directory)
def __exit__(self, *_):
print('Enter directory = ', self._old_cwd)
os.chdir(self._old_cwd)
class UnexpectedCrash(object):
def __init__(self, test, pid, *binaries):
self.test = test
self.pid = pid
self.binaries = binaries
def __str__(self):
return 'Crash({}: {} {})'.format(self.test, self.pid,
', '.join(self.binaries))
class PosixCoreDumpEnabler(object):
def __init__(self):
self._old_limits = None
def __enter__(self):
self._old_limits = resource.getrlimit(resource.RLIMIT_CORE)
resource.setrlimit(resource.RLIMIT_CORE, (-1, -1))
def __exit__(self, *_):
if self._old_limits != None:
resource.setrlimit(resource.RLIMIT_CORE, self._old_limits)
class LinuxCoreDumpEnabler(PosixCoreDumpEnabler):
def __enter__(self):
# Bump core limits to unlimited if core_pattern is correctly configured.
if CheckLinuxCoreDumpPattern(fatal=False):
super(LinuxCoreDumpEnabler, self).__enter__()
def __exit__(self, *args):
CheckLinuxCoreDumpPattern(fatal=False)
super(LinuxCoreDumpEnabler, self).__exit__(*args)
class WindowsCoreDumpEnabler(object):
"""This enabler assumes that Dart binary was built with Crashpad support.
In this case DART_CRASHPAD_CRASHES_DIR environment variable allows to
specify the location of Crashpad crashes database. Actual minidumps will
be written into reports subfolder of the database.
"""
CRASHPAD_DB_FOLDER = os.path.join(DART_DIR, 'crashes')
DUMPS_FOLDER = os.path.join(CRASHPAD_DB_FOLDER, 'reports')
def __init__(self):
pass
def __enter__(self):
print('INFO: Enabling coredump archiving into {}'.format(
WindowsCoreDumpEnabler.CRASHPAD_DB_FOLDER))
os.environ[
'DART_CRASHPAD_CRASHES_DIR'] = WindowsCoreDumpEnabler.CRASHPAD_DB_FOLDER
def __exit__(self, *_):
del os.environ['DART_CRASHPAD_CRASHES_DIR']
def TryUnlink(file):
try:
os.unlink(file)
except Exception as error:
print('ERROR: Failed to remove {}: {}'.format(file, error))
class BaseCoreDumpArchiver(object):
"""This class reads coredumps file written by UnexpectedCrashDumpArchiver
into the current working directory and uploads all cores and binaries
listed in it into Cloud Storage (see
pkg/test_runner/lib/src/test_progress.dart).
"""
# test.dart will write a line for each unexpected crash into this file.
_UNEXPECTED_CRASHES_FILE = 'unexpected-crashes'
def __init__(self, search_dir, output_directory):
self._bucket = 'dart-temp-crash-archive'
self._binaries_dir = os.getcwd()
self._search_dir = search_dir
self._output_directory = output_directory
def _safe_cleanup(self):
try:
return self._cleanup()
except Exception as error:
print('ERROR: Failure during cleanup: {}'.format(error))
return False
def __enter__(self):
print('INFO: Core dump archiving is activated')
# Cleanup any stale files
if self._safe_cleanup():
print('WARNING: Found and removed stale coredumps')
def __exit__(self, *_):
try:
crashes = self._find_unexpected_crashes()
if crashes:
# If we get a ton of crashes, only archive 10 dumps.
archive_crashes = crashes[:10]
print('Archiving coredumps for crash (if possible):')
for crash in archive_crashes:
print('----> {}'.format(crash))
sys.stdout.flush()
self._archive(archive_crashes)
else:
print('INFO: No unexpected crashes recorded')
dumps = self._find_all_coredumps()
if dumps:
print('INFO: However there are {} core dumps found'.format(
len(dumps)))
for dump in dumps:
print('INFO: -> {}'.format(dump))
print()
except Exception as error:
print('ERROR: Failed to archive crashes: {}'.format(error))
raise
finally:
self._safe_cleanup()
def _archive(self, crashes):
files = set()
missing = []
for crash in crashes:
files.update(crash.binaries)
core = self._find_coredump_file(crash)
if core:
files.add(core)
else:
missing.append(crash)
if self._output_directory is not None and self._is_shard():
print(
"INFO: Moving collected dumps and binaries into output directory\n"
"INFO: They will be uploaded to isolate server. Look for \"isolated"
" out\" under the failed step on the build page.\n"
"INFO: For more information see runtime/docs/infra/coredumps.md"
)
self._move(files)
else:
print(
"INFO: Uploading collected dumps and binaries into Cloud Storage\n"
"INFO: Use `gsutil.py cp from-url to-path` to download them.\n"
"INFO: For more information see runtime/docs/infra/coredumps.md"
)
self._upload(files)
if missing:
self._report_missing_crashes(missing, throw=True)
# todo(athom): move the logic to decide where to copy core dumps into the recipes.
def _is_shard(self):
return 'BUILDBOT_BUILDERNAME' not in os.environ
def _report_missing_crashes(self, missing, throw=True):
missing_as_string = ', '.join([str(c) for c in missing])
other_files = list(glob.glob(os.path.join(self._search_dir, '*')))
sys.stderr.write(
"Could not find crash dumps for '{}' in search directory '{}'.\n"
"Existing files which *did not* match the pattern inside the search "
"directory are are:\n {}\n".format(missing_as_string,
self._search_dir,
'\n '.join(other_files)))
# TODO: Figure out why windows coredump generation does not work.
# See http://dartbug.com/36469
if throw and GuessOS() != 'win32':
raise Exception(
'Missing crash dumps for: {}'.format(missing_as_string))
def _get_file_name(self, file):
# Sanitize the name: actual cores follow 'core.%d' pattern, crashed
# binaries are copied next to cores and named
# 'binary.<mode>_<arch>_<binary_name>'.
# This should match the code in testing/dart/test_progress.dart
name = os.path.basename(file)
(prefix, suffix) = name.split('.', 1)
is_binary = prefix == 'binary'
if is_binary:
(mode, arch, binary_name) = suffix.split('_', 2)
name = binary_name
return (name, is_binary)
def _move(self, files):
for file in files:
print('+++ Moving {} to output_directory ({})'.format(
file, self._output_directory))
(name, is_binary) = self._get_file_name(file)
destination = os.path.join(self._output_directory, name)
shutil.move(file, destination)
if is_binary and os.path.exists(file + '.pdb'):
# Also move a PDB file if there is one.
pdb = os.path.join(self._output_directory, name + '.pdb')
shutil.move(file + '.pdb', pdb)
def _tar(self, file):
(name, is_binary) = self._get_file_name(file)
tarname = '{}.tar.gz'.format(name)
# Compress the file.
tar = tarfile.open(tarname, mode='w:gz')
tar.add(file, arcname=name)
if is_binary and os.path.exists(file + '.pdb'):
# Also add a PDB file if there is one.
tar.add(file + '.pdb', arcname=name + '.pdb')
tar.close()
return tarname
def _upload(self, files):
bot_utils = GetBotUtils()
gsutil = bot_utils.GSUtil()
storage_path = '{}/{}/'.format(self._bucket, uuid.uuid4())
gs_prefix = 'gs://{}'.format(storage_path)
http_prefix = 'https://storage.cloud.google.com/{}'.format(storage_path)
print('\n--- Uploading into {} ({}) ---'.format(gs_prefix, http_prefix))
for file in files:
tarname = self._tar(file)
# Remove / from absolute path to not have // in gs path.
gs_url = '{}{}'.format(gs_prefix, tarname)
http_url = '{}{}'.format(http_prefix, tarname)
try:
gsutil.upload(tarname, gs_url)
print('+++ Uploaded {} ({})'.format(gs_url, http_url))
except Exception as error:
print('!!! Failed to upload {}, error: {}'.format(
tarname, error))
TryUnlink(tarname)
print('--- Done ---\n')
def _find_all_coredumps(self):
"""Return coredumps that were recorded (if supported by the platform).
This method will be overriden by concrete platform specific implementations.
"""
return []
def _find_unexpected_crashes(self):
"""Load coredumps file. Each line has the following format:
test-name,pid,binary-file1,binary-file2,...
"""
try:
with open(BaseCoreDumpArchiver._UNEXPECTED_CRASHES_FILE) as f:
return [
UnexpectedCrash(*ln.strip('\n').split(','))
for ln in f.readlines()
]
except:
return []
def _cleanup(self):
found = False
if os.path.exists(BaseCoreDumpArchiver._UNEXPECTED_CRASHES_FILE):
os.unlink(BaseCoreDumpArchiver._UNEXPECTED_CRASHES_FILE)
found = True
for binary in glob.glob(os.path.join(self._binaries_dir, 'binary.*')):
found = True
TryUnlink(binary)
return found
class PosixCoreDumpArchiver(BaseCoreDumpArchiver):
def __init__(self, search_dir, output_directory):
super(PosixCoreDumpArchiver, self).__init__(search_dir,
output_directory)
def _cleanup(self):
found = super(PosixCoreDumpArchiver, self)._cleanup()
for core in glob.glob(os.path.join(self._search_dir, 'core.*')):
found = True
TryUnlink(core)
return found
def _find_coredump_file(self, crash):
core_filename = os.path.join(self._search_dir,
'core.{}'.format(crash.pid))
if os.path.exists(core_filename):
return core_filename
class LinuxCoreDumpArchiver(PosixCoreDumpArchiver):
def __init__(self, output_directory):
super(LinuxCoreDumpArchiver, self).__init__(os.getcwd(),
output_directory)
class MacOSCoreDumpArchiver(PosixCoreDumpArchiver):
def __init__(self, output_directory):
super(MacOSCoreDumpArchiver, self).__init__('/cores', output_directory)
class WindowsCoreDumpArchiver(BaseCoreDumpArchiver):
def __init__(self, output_directory):
super(WindowsCoreDumpArchiver, self).__init__(
WindowsCoreDumpEnabler.DUMPS_FOLDER, output_directory)
self._dumps_by_pid = None
# Find CDB.exe in the win_toolchain that we are using.
def _find_cdb(self):
win_toolchain_json_path = os.path.join(DART_DIR, 'build',
'win_toolchain.json')
if not os.path.exists(win_toolchain_json_path):
return None
with open(win_toolchain_json_path, 'r') as f:
win_toolchain_info = json.loads(f.read())
win_sdk_path = win_toolchain_info['win_sdk']
# We assume that we are running on 64-bit Windows.
# Note: x64 CDB can work with both X64 and IA32 dumps.
cdb_path = os.path.join(win_sdk_path, 'Debuggers', 'x64', 'cdb.exe')
if not os.path.exists(cdb_path):
return None
return cdb_path
CDBG_PROMPT_RE = re.compile(r'^\d+:\d+>')
def _dump_all_stacks(self):
# On Windows due to crashpad integration crashes do not produce any
# stacktraces. Dump stack traces from dumps Crashpad collected using
# CDB (if available).
cdb_path = self._find_cdb()
if cdb_path is None:
return
dumps = self._find_all_coredumps()
if not dumps:
return
print('### Collected {} crash dumps'.format(len(dumps)))
for dump in dumps:
print()
print('### Dumping stacks from {} using CDB'.format(dump))
cdb_output = subprocess.check_output(
'"{}" -z "{}" -kqm -c "!uniqstack -b -v -p;qd"'.format(
cdb_path, dump),
stderr=subprocess.STDOUT)
# Extract output of uniqstack from the whole output of CDB.
output = False
for line in cdb_output.split('\n'):
if re.match(WindowsCoreDumpArchiver.CDBG_PROMPT_RE, line):
output = True
elif line.startswith('quit:'):
break
elif output:
print(line)
print()
print('#############################################')
print()
def __exit__(self, *args):
try:
self._dump_all_stacks()
except Exception as error:
print('ERROR: Unable to dump stacks from dumps: {}'.format(error))
super(WindowsCoreDumpArchiver, self).__exit__(*args)
def _cleanup(self):
found = super(WindowsCoreDumpArchiver, self)._cleanup()
for core in glob.glob(os.path.join(self._search_dir, '*')):
found = True
TryUnlink(core)
return found
def _find_all_coredumps(self):
pattern = os.path.join(self._search_dir, '*.dmp')
return [core_filename for core_filename in glob.glob(pattern)]
def _find_coredump_file(self, crash):
if self._dumps_by_pid is None:
# If this function is invoked the first time then look through the directory
# that contains crashes for all dump files and collect pid -> filename
# mapping.
self._dumps_by_pid = {}
minidump = GetMinidumpUtils()
pattern = os.path.join(self._search_dir, '*.dmp')
for core_filename in glob.glob(pattern):
pid = minidump.GetProcessIdFromDump(core_filename)
if pid != -1:
self._dumps_by_pid[str(pid)] = core_filename
if crash.pid in self._dumps_by_pid:
return self._dumps_by_pid[crash.pid]
def _report_missing_crashes(self, missing, throw=True):
# Let's only print the debugging information and not throw. We'll do more
# validation for werfault.exe and throw afterwards.
super(WindowsCoreDumpArchiver, self)._report_missing_crashes(
missing, throw=False)
if throw:
missing_as_string = ', '.join([str(c) for c in missing])
raise Exception(
'Missing crash dumps for: {}'.format(missing_as_string))
class IncreasedNumberOfFileDescriptors(object):
def __init__(self, nofiles):
self._old_limits = None
self._limits = (nofiles, nofiles)
def __enter__(self):
self._old_limits = resource.getrlimit(resource.RLIMIT_NOFILE)
resource.setrlimit(resource.RLIMIT_NOFILE, self._limits)
def __exit__(self, *_):
resource.setrlimit(resource.RLIMIT_CORE, self._old_limits)
@contextlib.contextmanager
def NooptContextManager():
yield
def CoreDumpArchiver(args):
enabled = '--copy-coredumps' in args
prefix = '--output-directory='
output_directory = next(
(arg[len(prefix):] for arg in args if arg.startswith(prefix)), None)
if not enabled:
return NooptContextManager()
osname = GuessOS()
if osname == 'linux':
return contextlib.nested(LinuxCoreDumpEnabler(),
LinuxCoreDumpArchiver(output_directory))
elif osname == 'macos':
return contextlib.nested(PosixCoreDumpEnabler(),
MacOSCoreDumpArchiver(output_directory))
elif osname == 'win32':
return contextlib.nested(WindowsCoreDumpEnabler(),
WindowsCoreDumpArchiver(output_directory))
# We don't have support for MacOS yet.
return NooptContextManager()
def FileDescriptorLimitIncreaser():
osname = GuessOS()
if osname == 'macos':
return IncreasedNumberOfFileDescriptors(nofiles=10000)
assert osname in ('linux', 'win32')
# We don't have support for MacOS yet.
return NooptContextManager()
def Main():
print('GuessOS() -> ', GuessOS())
print('GuessArchitecture() -> ', GuessArchitecture())
print('GuessCpus() -> ', GuessCpus())
print('IsWindows() -> ', IsWindows())
print('GuessVisualStudioPath() -> ', GuessVisualStudioPath())
print('GetGitRevision() -> ', GetGitRevision())
print('GetGitTimestamp() -> ', GetGitTimestamp())
print('GetVersionFileContent() -> ', GetVersionFileContent())
print('GetGitNumber() -> ', GetGitNumber())
if __name__ == '__main__':
Main()