# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this file,
# You can obtain one at http://mozilla.org/MPL/2.0/.
# The mozprocess ProcessHandler and ProcessHandlerMixin are typically used as
# an alternative to the python subprocess module. They have been used in many
# Mozilla test harnesses with some success -- but also with on-going concerns,
# especially regarding reliability and exception handling.
#
# New code should try to use the standard subprocess module, and only use
# this ProcessHandler if absolutely necessary.
import codecs
import errno
import io
import os
import signal
import subprocess
import sys
import threading
import time
import traceback
from datetime
import datetime
from queue
import Empty, Queue
import six
# Set the MOZPROCESS_DEBUG environment variable to 1 to see some debugging output
MOZPROCESS_DEBUG = os.getenv(
"MOZPROCESS_DEBUG")
INTERVAL_PROCESS_ALIVE_CHECK = 0.02
# We dont use mozinfo because it is expensive to import, see bug 933558.
isWin = os.name ==
"nt"
isPosix = os.name ==
"posix" # includes MacOS X
if isWin:
from ctypes
import WinError, addressof, byref, c_longlong, c_ulong, sizeof
from .
import winprocess
from .qijo
import (
IO_COUNTERS,
JOBOBJECT_ASSOCIATE_COMPLETION_PORT,
JOBOBJECT_BASIC_LIMIT_INFORMATION,
JOBOBJECT_EXTENDED_LIMIT_INFORMATION,
JobObjectAssociateCompletionPortInformation,
JobObjectExtendedLimitInformation,
)
class ProcessHandlerMixin(object):
"""
A
class for launching
and manipulating local processes.
:param cmd: command to run. May be a string
or a list.
If specified
as a list, the first
element will be interpreted
as the command,
and all additional elements will be interpreted
as arguments to that command.
:param args: list of arguments to
pass to the command (defaults to
None). Must
not be set when
`cmd`
is specified
as a list.
:param cwd: working directory
for command (defaults to
None).
:param env:
is the environment to use
for the process (defaults to os.environ).
:param ignore_children: causes system to ignore child processes when
True,
defaults to
False (which tracks child processes).
:param kill_on_timeout: when
True, the process will be killed when a timeout
is reached.
When
False, the caller
is responsible
for killing the process.
Failure to do so could cause a call to wait() to hang indefinitely. (Defaults to
True.)
:param processOutputLine: function
or list of functions to be called
for
each line of output produced by the process (defaults to an empty
list).
:param processStderrLine: function
or list of functions to be called
for each line of error output - stderr - produced by the process
(defaults to an empty list).
If this
is not specified, stderr lines
will be sent to the *processOutputLine* callbacks.
:param onTimeout: function
or list of functions to be called when the process times out.
:param onFinish: function
or list of functions to be called when the process terminates
normally without timing out.
:param kwargs: additional keyword args to
pass directly into Popen.
NOTE: Child processes will be tracked by default.
If for any reason
we are unable to track child processes
and ignore_children
is set to
False,
then we will fall back to only tracking the root process. The fallback
will be logged.
"""
class Process(subprocess.Popen):
"""
Represents our view of a subprocess.
It adds a kill() method which allows it to be stopped explicitly.
"""
MAX_IOCOMPLETION_PORT_NOTIFICATION_DELAY = 180
TIMEOUT_BEFORE_SIGKILL = 1.0
def __init__(
self,
args,
bufsize=0,
executable=
None,
stdin=
None,
stdout=
None,
stderr=
None,
preexec_fn=
None,
close_fds=
False,
shell=
False,
cwd=
None,
env=
None,
universal_newlines=
False,
startupinfo=
None,
creationflags=0,
ignore_children=
False,
encoding=
"utf-8",
):
# Parameter for whether or not we should attempt to track child processes
self._ignore_children = ignore_children
self._job =
None
self._io_port =
None
if isWin:
self._cleanup_lock = threading.Lock()
if not self._ignore_children
and not isWin:
# Set the process group id for linux systems
# Sets process group id to the pid of the parent process
# NOTE: This prevents you from using preexec_fn and managing
# child processes, TODO: Ideally, find a way around this
def setpgidfn():
os.setpgid(0, 0)
preexec_fn = setpgidfn
kwargs = {
"bufsize": bufsize,
"executable": executable,
"stdin": stdin,
"stdout": stdout,
"stderr": stderr,
"preexec_fn": preexec_fn,
"close_fds": close_fds,
"shell": shell,
"cwd": cwd,
"env": env,
"startupinfo": startupinfo,
"creationflags": creationflags,
}
if sys.version_info.minor >= 6
and universal_newlines:
kwargs[
"universal_newlines"] = universal_newlines
kwargs[
"encoding"] = encoding
try:
subprocess.Popen.__init__(self, args, **kwargs)
except OSError:
print(args, file=sys.stderr)
raise
def debug(self, msg):
if not MOZPROCESS_DEBUG:
return
thread = threading.current_thread().name
print(
"DBG::MOZPROC PID:{} ({}) | {}".format(self.pid, thread, msg))
def __del__(self):
if isWin:
_maxint = sys.maxsize
handle = getattr(self,
"_handle",
None)
if handle:
# _internal_poll is a Python3 built-in call and requires _handle to be an int on Windows
# It's only an AutoHANDLE for legacy Python2 reasons that are non-trivial to remove
self._handle = int(self._handle)
self._internal_poll(_deadstate=_maxint)
# Revert it back to the saved 'handle' (AutoHANDLE) for self._cleanup()
self._handle = handle
if handle
or self._job
or self._io_port:
self._cleanup()
else:
subprocess.Popen.__del__(self)
def send_signal(self, sig=
None):
if isWin:
try:
if not self._ignore_children
and self._handle
and self._job:
self.debug(
"calling TerminateJobObject")
winprocess.TerminateJobObject(
self._job, winprocess.ERROR_CONTROL_C_EXIT
)
elif self._handle:
self.debug(
"calling TerminateProcess")
winprocess.TerminateProcess(
self._handle, winprocess.ERROR_CONTROL_C_EXIT
)
except WindowsError:
self._cleanup()
traceback.print_exc()
raise OSError(
"Could not terminate process")
else:
def send_sig(sig, retries=0):
pid = self.detached_pid
or self.pid
if not self._ignore_children:
try:
os.killpg(pid, sig)
except BaseException
as e:
# On Mac OSX if the process group contains zombie
# processes, killpg results in an EPERM.
# In this case, zombie processes need to be reaped
# before continuing
# Note: A negative pid refers to the entire process
# group
if retries < 1
and getattr(e,
"errno",
None) == errno.EPERM:
try:
os.waitpid(-pid, 0)
finally:
return send_sig(sig, retries + 1)
# ESRCH is a "no such process" failure, which is fine because the
# application might already have been terminated itself. Any other
# error would indicate a problem in killing the process.
if getattr(e,
"errno",
None) != errno.ESRCH:
print(
"Could not terminate process: %s" % self.pid,
file=sys.stderr,
)
raise
else:
os.kill(pid, sig)
if sig
is None and isPosix:
# ask the process for termination and wait a bit
send_sig(signal.SIGTERM)
limit = time.time() + self.TIMEOUT_BEFORE_SIGKILL
while time.time() <= limit:
if self.poll()
is not None:
# process terminated nicely
break
time.sleep(INTERVAL_PROCESS_ALIVE_CHECK)
else:
# process did not terminate - send SIGKILL to force
send_sig(signal.SIGKILL)
else:
# a signal was explicitly set or not posix
send_sig(sig
or signal.SIGKILL)
def kill(self, sig=
None, timeout=
None):
self.send_signal(sig)
self.returncode = self.wait(timeout)
self._cleanup()
return self.returncode
def poll(self):
"""Popen.poll
Check
if child process has terminated. Set
and return returncode attribute.
"""
if isWin:
returncode = self._custom_wait(timeout=0)
else:
returncode = subprocess.Popen.poll(self)
if returncode
is not None:
self._cleanup()
return returncode
def wait(self, timeout=
None):
"""Popen.wait
Called to wait
for a running process to shut down
and return
its exit code
Returns the main process
's exit code
"""
# This call will be different for each OS
self.returncode = self._custom_wait(timeout=timeout)
if self.returncode
is not None:
self._cleanup()
return self.returncode
""" Private Members of Process class """
if isWin:
# Redefine the execute child so that we can track process groups
def _execute_child(self, *args_tuple):
(
args,
executable,
preexec_fn,
close_fds,
pass_fds,
cwd,
env,
startupinfo,
creationflags,
shell,
p2cread,
p2cwrite,
c2pread,
c2pwrite,
errread,
errwrite,
*_,
) = args_tuple
if not isinstance(args, six.string_types):
args = subprocess.list2cmdline(args)
# Always or in the create new process group
creationflags |= winprocess.CREATE_NEW_PROCESS_GROUP
if startupinfo
is None:
startupinfo = winprocess.STARTUPINFO()
if None not in (p2cread, c2pwrite, errwrite):
startupinfo.dwFlags |= winprocess.STARTF_USESTDHANDLES
startupinfo.hStdInput = int(p2cread)
startupinfo.hStdOutput = int(c2pwrite)
startupinfo.hStdError = int(errwrite)
if shell:
startupinfo.dwFlags |= winprocess.STARTF_USESHOWWINDOW
startupinfo.wShowWindow = winprocess.SW_HIDE
comspec = os.environ.get(
"COMSPEC",
"cmd.exe")
args = comspec +
" /c " + args
# Determine if we can create a job or create nested jobs.
can_create_job = winprocess.CanCreateJobObject()
can_nest_jobs = self._can_nest_jobs()
# Ensure we write a warning message if we are falling back
if not (can_create_job
or can_nest_jobs)
and not self._ignore_children:
# We can't create job objects AND the user wanted us to
# Warn the user about this.
print(
"ProcessManager UNABLE to use job objects to manage "
"child processes",
file=sys.stderr,
)
# set process creation flags
creationflags |= winprocess.CREATE_SUSPENDED
creationflags |= winprocess.CREATE_UNICODE_ENVIRONMENT
if can_create_job:
creationflags |= winprocess.CREATE_BREAKAWAY_FROM_JOB
if not (can_create_job
or can_nest_jobs):
# Since we've warned, we just log info here to inform you
# of the consequence of setting ignore_children = True
print(
"ProcessManager NOT managing child processes")
# create the process
hp, ht, pid, tid = winprocess.CreateProcess(
executable,
args,
None,
None,
# No special security
1,
# Must inherit handles!
creationflags,
winprocess.EnvironmentBlock(env),
cwd,
startupinfo,
)
self._child_created =
True
self._handle = hp
self._thread = ht
self.pid = pid
self.tid = tid
if not self._ignore_children
and (can_create_job
or can_nest_jobs):
try:
# We create a new job for this process, so that we can kill
# the process and any sub-processes
# Create the IO Completion Port
self._io_port = winprocess.CreateIoCompletionPort()
self._job = winprocess.CreateJobObject()
# Now associate the io comp port and the job object
joacp = JOBOBJECT_ASSOCIATE_COMPLETION_PORT(
winprocess.COMPKEY_JOBOBJECT, self._io_port
)
winprocess.SetInformationJobObject(
self._job,
JobObjectAssociateCompletionPortInformation,
addressof(joacp),
sizeof(joacp),
)
# Allow subprocesses to break away from us - necessary when
# Firefox restarts, or flash with protected mode
limit_flags = winprocess.JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE
if not can_nest_jobs:
# This allows sandbox processes to create their own job,
# and is necessary to set for older versions of Windows
# without nested job support.
limit_flags |= winprocess.JOB_OBJECT_LIMIT_BREAKAWAY_OK
jbli = JOBOBJECT_BASIC_LIMIT_INFORMATION(
c_longlong(0),
# per process time limit (ignored)
c_longlong(0),
# per job user time limit (ignored)
limit_flags,
0,
# min working set (ignored)
0,
# max working set (ignored)
0,
# active process limit (ignored)
None,
# affinity (ignored)
0,
# Priority class (ignored)
0,
# Scheduling class (ignored)
)
iocntr = IO_COUNTERS()
jeli = JOBOBJECT_EXTENDED_LIMIT_INFORMATION(
jbli,
# basic limit info struct
iocntr,
# io_counters (ignored)
0,
# process mem limit (ignored)
0,
# job mem limit (ignored)
0,
# peak process limit (ignored)
0,
)
# peak job limit (ignored)
winprocess.SetInformationJobObject(
self._job,
JobObjectExtendedLimitInformation,
addressof(jeli),
sizeof(jeli),
)
# Assign the job object to the process
winprocess.AssignProcessToJobObject(self._job, int(hp))
# It's overkill, but we use Queue to signal between threads
# because it handles errors more gracefully than event or condition.
self._process_events = Queue()
# Spin up our thread for managing the IO Completion Port
self._procmgrthread = threading.Thread(target=self._procmgr)
except Exception:
print(
"""Exception trying to use job objects;
falling back to
not using job objects
for managing child processes
""",
file=sys.stderr,
)
tb = traceback.format_exc()
print(tb, file=sys.stderr)
# Ensure no dangling handles left behind
self._cleanup_job_io_port()
else:
self._job =
None
winprocess.ResumeThread(int(ht))
if getattr(self,
"_procmgrthread",
None):
self._procmgrthread.start()
ht.Close()
for i
in (p2cread, c2pwrite, errwrite):
if i
is not None:
i.Close()
# Per:
# https://msdn.microsoft.com/en-us/library/windows/desktop/hh448388%28v=vs.85%29.aspx
# Nesting jobs came in with windows versions starting with 6.2 according to the table
# on this page:
# https://msdn.microsoft.com/en-us/library/ms724834%28v=vs.85%29.aspx
def _can_nest_jobs(self):
winver = sys.getwindowsversion()
return winver.major > 6
or winver.major == 6
and winver.minor >= 2
# Windows Process Manager - watches the IO Completion Port and
# keeps track of child processes
def _procmgr(self):
if not (self._io_port)
or not (self._job):
return
try:
self._poll_iocompletion_port()
except Exception:
traceback.print_exc()
# If _poll_iocompletion_port threw an exception for some unexpected reason,
# send an event that will make _custom_wait throw an Exception.
self._process_events.put({})
except KeyboardInterrupt:
raise KeyboardInterrupt
def _poll_iocompletion_port(self):
# Watch the IO Completion port for status
self._spawned_procs = {}
countdowntokill = 0
self.debug(
"start polling IO completion port")
while True:
msgid = c_ulong(0)
compkey = c_ulong(0)
pid = c_ulong(0)
portstatus = winprocess.GetQueuedCompletionStatus(
self._io_port, byref(msgid), byref(compkey), byref(pid), 5000
)
# If the countdowntokill has been activated, we need to check
# if we should start killing the children or not.
if countdowntokill != 0:
diff = datetime.now() - countdowntokill
# Arbitrarily wait 3 minutes for windows to get its act together
# Windows sometimes takes a small nap between notifying the
# IO Completion port and actually killing the children, and we
# don't want to mistake that situation for the situation of an unexpected
# parent abort (which is what we're looking for here).
if diff.seconds > self.MAX_IOCOMPLETION_PORT_NOTIFICATION_DELAY:
print(
"WARNING | IO Completion Port failed to signal "
"process shutdown",
file=sys.stderr,
)
print(
"Parent process %s exited with children alive:"
% self.pid,
file=sys.stderr,
)
print(
"PIDS: %s"
%
", ".join([str(i)
for i
in self._spawned_procs]),
file=sys.stderr,
)
print(
"Attempting to kill them, but no guarantee of success",
file=sys.stderr,
)
self.send_signal()
self._process_events.put({self.pid:
"FINISHED"})
break
if not portstatus:
# Check to see what happened
errcode = winprocess.GetLastError()
if errcode == winprocess.ERROR_ABANDONED_WAIT_0:
# Then something has killed the port, break the loop
print(
"IO Completion Port unexpectedly closed",
file=sys.stderr,
)
self._process_events.put({self.pid:
"FINISHED"})
break
elif errcode == winprocess.WAIT_TIMEOUT:
# Timeouts are expected, just keep on polling
continue
else:
print(
"Error Code %s trying to query IO Completion Port, "
"exiting" % errcode,
file=sys.stderr,
)
raise WinError(errcode)
break
if compkey.value == winprocess.COMPKEY_TERMINATE.value:
self.debug(
"compkeyterminate detected")
# Then we're done
break
# Check the status of the IO Port and do things based on it
if compkey.value == winprocess.COMPKEY_JOBOBJECT.value:
if msgid.value == winprocess.JOB_OBJECT_MSG_ACTIVE_PROCESS_ZERO:
# No processes left, time to shut down
# Signal anyone waiting on us that it is safe to shut down
self.debug(
"job object msg active processes zero")
self._process_events.put({self.pid:
"FINISHED"})
break
elif msgid.value == winprocess.JOB_OBJECT_MSG_NEW_PROCESS:
# New Process started
# Add the child proc to our list in case our parent flakes out on us
# without killing everything.
if pid.value != self.pid:
self._spawned_procs[pid.value] = 1
self.debug(
"new process detected with pid value: %s"
% pid.value
)
elif msgid.value == winprocess.JOB_OBJECT_MSG_EXIT_PROCESS:
self.debug(
"process id %s exited normally" % pid.value)
# One process exited normally
if pid.value == self.pid
and len(self._spawned_procs) > 0:
# Parent process dying, start countdown timer
countdowntokill = datetime.now()
elif pid.value
in self._spawned_procs:
# Child Process died remove from list
del self._spawned_procs[pid.value]
elif (
msgid.value
== winprocess.JOB_OBJECT_MSG_ABNORMAL_EXIT_PROCESS
):
# One process existed abnormally
self.debug(
"process id %s exited abnormally" % pid.value)
if pid.value == self.pid
and len(self._spawned_procs) > 0:
# Parent process dying, start countdown timer
countdowntokill = datetime.now()
elif pid.value
in self._spawned_procs:
# Child Process died remove from list
del self._spawned_procs[pid.value]
else:
# We don't care about anything else
self.debug(
"We got a message %s" % msgid.value)
pass
def _custom_wait(self, timeout=
None):
"""Custom implementation of wait.
- timeout: number of seconds before timing out.
If None,
will wait indefinitely.
"""
# First, check to see if the process is still running
if self._handle:
returncode = winprocess.GetExitCodeProcess(self._handle)
if returncode != winprocess.STILL_ACTIVE:
self.returncode = returncode
else:
# Dude, the process is like totally dead!
return self.returncode
# On Windows, an unlimited timeout prevents KeyboardInterrupt from
# being caught.
the_timeout = 0.1
if timeout
is None else timeout
if self._job:
self.debug(
"waiting with IO completion port")
# Then we are managing with IO Completion Ports
# wait on a signal so we know when we have seen the last
# process come through.
# We use queues to synchronize between the thread and this
# function because events just didn't have robust enough error
# handling on pre-2.7 versions
try:
while True:
try:
item = self._process_events.get(timeout=the_timeout)
except Empty:
# The timeout was not given by the user, we just have a
# timeout to allow KeyboardInterrupt, so retry.
if timeout
is None:
continue
else:
raise
break
# re-emit the event in case some other thread is also calling wait()
self._process_events.put(item)
if item[self.pid] ==
"FINISHED":
self.debug(
"received 'FINISHED' from _procmgrthread")
self._process_events.task_done()
except Empty:
# There was no event within the expected time.
pass
except Exception:
traceback.print_exc()
raise OSError(
"IO Completion Port failed to signal process shutdown"
)
finally:
if self._handle:
returncode = winprocess.GetExitCodeProcess(self._handle)
if returncode != winprocess.STILL_ACTIVE:
self.returncode = returncode
else:
# Not managing with job objects, so all we can reasonably do
# is call waitforsingleobject and hope for the best
self.debug(
"waiting without IO completion port")
if not self._ignore_children:
self.debug(
"NOT USING JOB OBJECTS!!!")
# First, make sure we have not already ended
if self.returncode
is not None:
return self.returncode
rc =
None
if self._handle:
# timeout for WaitForSingleObject is in ms
the_timeout = int(the_timeout * 1000)
while True:
rc = winprocess.WaitForSingleObject(
self._handle, the_timeout
)
# The timeout was not given by the user, we just have a
# timeout to allow KeyboardInterrupt, so retry.
if timeout
is None and rc == winprocess.WAIT_TIMEOUT:
continue
break
if rc == winprocess.WAIT_TIMEOUT:
# Timeout happened as asked.
pass
elif rc == winprocess.WAIT_OBJECT_0:
# We caught WAIT_OBJECT_0, which indicates all is well
print(
"Single process terminated successfully")
self.returncode = winprocess.GetExitCodeProcess(self._handle)
else:
# An error occured we should probably throw
rc = winprocess.GetLastError()
if rc:
raise WinError(rc)
return self.returncode
def _cleanup_job_io_port(self):
"""Do the job and IO port cleanup separately because there are
cases where we want to clean these without killing _handle
(i.e.
if we fail to create the job object
in the first place)
"""
if (
getattr(self,
"_job")
and self._job != winprocess.INVALID_HANDLE_VALUE
):
self._job.Close()
self._job =
None
else:
# If windows already freed our handle just set it to none
# (saw this intermittently while testing)
self._job =
None
if (
getattr(self,
"_io_port",
None)
and self._io_port != winprocess.INVALID_HANDLE_VALUE
):
self._io_port.Close()
self._io_port =
None
else:
self._io_port =
None
if getattr(self,
"_procmgrthread",
None):
self._procmgrthread =
None
def _cleanup(self):
self._cleanup_lock.acquire()
self._cleanup_job_io_port()
if self._thread
and self._thread != winprocess.INVALID_HANDLE_VALUE:
self._thread.Close()
self._thread =
None
else:
self._thread =
None
if self._handle
and self._handle != winprocess.INVALID_HANDLE_VALUE:
self._handle.Close()
self._handle =
None
else:
self._handle =
None
self._cleanup_lock.release()
else:
def _custom_wait(self, timeout=
None):
"""Haven't found any reason to differentiate between these platforms
so they all use the same wait callback.
If it
is necessary to
craft different styles of wait, then a new _custom_wait method
could be easily implemented.
"""
# For non-group wait, call base class
try:
subprocess.Popen.wait(self, timeout=timeout)
except subprocess.TimeoutExpired:
# We want to return None in this case
pass
return self.returncode
def _cleanup(self):
pass
def __init__(
self,
cmd,
args=
None,
cwd=
None,
env=
None,
ignore_children=
False,
kill_on_timeout=
True,
processOutputLine=(),
processStderrLine=(),
onTimeout=(),
onFinish=(),
**kwargs
):
self.cmd = cmd
self.args = args
self.cwd = cwd
self.didTimeout =
False
self.didOutputTimeout =
False
self._ignore_children = ignore_children
self.keywordargs = kwargs
self.read_buffer =
""
if env
is None:
env = os.environ.copy()
self.env = env
# handlers
def to_callable_list(arg):
if callable(arg):
arg = [arg]
return CallableList(arg)
processOutputLine = to_callable_list(processOutputLine)
processStderrLine = to_callable_list(processStderrLine)
onTimeout = to_callable_list(onTimeout)
onFinish = to_callable_list(onFinish)
def on_timeout():
self.didTimeout =
True
self.didOutputTimeout = self.reader.didOutputTimeout
if kill_on_timeout:
self.kill()
onTimeout.insert(0, on_timeout)
self._stderr = subprocess.STDOUT
if processStderrLine:
self._stderr = subprocess.PIPE
self.reader = ProcessReader(
stdout_callback=processOutputLine,
stderr_callback=processStderrLine,
finished_callback=onFinish,
timeout_callback=onTimeout,
)
# It is common for people to pass in the entire array with the cmd and
# the args together since this is how Popen uses it. Allow for that.
if isinstance(self.cmd, list):
if self.args
is not None:
raise TypeError(
"cmd and args must not both be lists")
(self.cmd, self.args) = (self.cmd[0], self.cmd[1:])
elif self.args
is None:
self.args = []
def debug(self, msg):
if not MOZPROCESS_DEBUG:
return
cmd = self.cmd.split(os.sep)[-1:]
print(
"DBG::MOZPROC ProcessHandlerMixin {} | {}".format(cmd, msg))
@property
def timedOut(self):
"""True if the process has timed out for any reason."""
return self.didTimeout
@property
def outputTimedOut(self):
"""True if the process has timed out for no output."""
return self.didOutputTimeout
@property
def commandline(self):
"""the string value of the command line (command + args)"""
return subprocess.list2cmdline([self.cmd] + self.args)
def run(self, timeout=
None, outputTimeout=
None):
"""
Starts the process.
If timeout
is not None, the process will be allowed to
continue for
that number of seconds before being killed.
If the process
is killed
due to a timeout, the onTimeout handler will be called.
If outputTimeout
is not None, the process will be allowed to
continue
for that number of seconds without producing any output before
being killed.
"""
self.didTimeout =
False
self.didOutputTimeout =
False
# default arguments
args = dict(
stdout=subprocess.PIPE,
stderr=self._stderr,
cwd=self.cwd,
env=self.env,
ignore_children=self._ignore_children,
)
# build process arguments
args.update(self.keywordargs)
# launch the process
self.proc = self.Process([self.cmd] + self.args, **args)
if isPosix:
# Keep track of the initial process group in case the process detaches itself
self.proc.pgid = self._getpgid(self.proc.pid)
self.proc.detached_pid =
None
self.processOutput(timeout=timeout, outputTimeout=outputTimeout)
def kill(self, sig=
None, timeout=
None):
"""
Kills the managed process.
If you created the process
with 'ignore_children=False' (the
default) then it will also also kill all child processes spawned by
it.
If you specified
'ignore_children=True' when creating the
process, only the root process will be killed.
Note that this does
not manage any state, save any output etc,
it immediately kills the process.
:param sig: Signal used to kill the process, defaults to SIGKILL
(has no effect on Windows)
"""
if not hasattr(self,
"proc"):
raise RuntimeError(
"Process hasn't been started yet")
self.proc.kill(sig=sig, timeout=timeout)
# When we kill the the managed process we also have to wait for the
# reader thread to be finished. Otherwise consumers would have to assume
# that it still has not completely shutdown.
rc = self.wait(0)
if rc
is None:
self.debug(
"kill: wait failed -- process is still alive")
return rc
def poll(self):
"""Check if child process has terminated
Returns the current returncode value:
-
None if the process hasn
't terminated yet
- A negative number
if the process was killed by signal N (Unix only)
-
'0' if the process ended without failures
"""
if not hasattr(self,
"proc"):
raise RuntimeError(
"Process hasn't been started yet")
# Ensure that we first check for the reader status. Otherwise
# we might mark the process as finished while output is still getting
# processed.
elif not self._ignore_children
and self.reader.is_alive():
return None
elif hasattr(self,
"returncode"):
return self.returncode
else:
return self.proc.poll()
def processOutput(self, timeout=
None, outputTimeout=
None):
"""
Handle process output until the process terminates
or times out.
If timeout
is not None, the process will be allowed to
continue for
that number of seconds before being killed.
If outputTimeout
is not None, the process will be allowed to
continue
for that number of seconds without producing any output before
being killed.
"""
# this method is kept for backward compatibility
if not hasattr(self,
"proc"):
self.run(timeout=timeout, outputTimeout=outputTimeout)
# self.run will call this again
return
if not self.reader.is_alive():
self.reader.timeout = timeout
self.reader.output_timeout = outputTimeout
self.reader.start(self.proc)
def wait(self, timeout=
None):
"""
Waits until all output has been read
and the process
is
terminated.
If timeout
is not None, will
return after timeout seconds.
This timeout only causes the wait function to
return and
does
not kill the process.
Returns the process exit code value:
-
None if the process hasn
't terminated yet
- A negative number
if the process was killed by signal N (Unix only)
-
'0' if the process ended without failures
"""
self.returncode = self.proc.wait(timeout)
if (
self.returncode
is not None
and self.reader.thread
and self.reader.thread
is not threading.current_thread()
):
# If children are ignored and a child is still running because it's
# been daemonized or something, the reader might still be attached
# to that child'd output... and joining will deadlock.
# So instead, we wait for there to be no more active reading still
# happening.
if self._ignore_children:
while self.reader.is_still_reading(timeout=0.1):
time.sleep(0.1)
else:
self.reader.join()
return self.returncode
@property
def pid(self):
if not hasattr(self,
"proc"):
raise RuntimeError(
"Process hasn't been started yet")
return self.proc.pid
@staticmethod
def pid_exists(pid):
if pid < 0:
return False
if isWin:
try:
process = winprocess.OpenProcess(
winprocess.PROCESS_QUERY_INFORMATION | winprocess.PROCESS_VM_READ,
False,
pid,
)
return winprocess.GetExitCodeProcess(process) == winprocess.STILL_ACTIVE
except WindowsError
as e:
# no such process
if e.winerror == winprocess.ERROR_INVALID_PARAMETER:
return False
# access denied
if e.winerror == winprocess.ERROR_ACCESS_DENIED:
return True
# re-raise for any other type of exception
raise
elif isPosix:
try:
os.kill(pid, 0)
except OSError
as e:
return e.errno == errno.EPERM
else:
return True
@classmethod
def _getpgid(cls, pid):
try:
return os.getpgid(pid)
except OSError
as e:
# Do not raise for "No such process"
if e.errno != errno.ESRCH:
raise
def check_for_detached(self, new_pid):
"""Check if the current process has been detached and mark it appropriately.
In case of application restarts the process can spawn itself into a new process group.
From now on the process can no longer be tracked by mozprocess anymore
and has to be
marked
as detached.
If the consumer of mozprocess still knows the new process id it could
check
for the detached state.
new_pid
is the new process id of the child process.
"""
if not hasattr(self,
"proc"):
raise RuntimeError(
"Process hasn't been started yet")
if isPosix:
new_pgid = self._getpgid(new_pid)
if new_pgid
and new_pgid != self.proc.pgid:
self.proc.detached_pid = new_pid
print(
'Child process with id "%s" has been marked as detached because it is no '
"longer in the managed process group. Keeping reference to the process id "
'"%s" which is the new child process.' % (self.pid, new_pid),
file=sys.stdout,
)
class CallableList(list):
def __call__(self, *args, **kwargs):
for e
in self:
e(*args, **kwargs)
def __add__(self, lst):
return CallableList(list.__add__(self, lst))
class ProcessReader(object):
def __init__(
self,
stdout_callback=
None,
stderr_callback=
None,
finished_callback=
None,
timeout_callback=
None,
timeout=
None,
output_timeout=
None,
):
self.stdout_callback = stdout_callback
or (
lambda line:
True)
self.stderr_callback = stderr_callback
or (
lambda line:
True)
self.finished_callback = finished_callback
or (
lambda:
True)
self.timeout_callback = timeout_callback
or (
lambda:
True)
self.timeout = timeout
self.output_timeout = output_timeout
self.thread =
None
self.got_data = threading.Event()
self.didOutputTimeout =
False
def debug(self, msg):
if not MOZPROCESS_DEBUG:
return
print(
"DBG::MOZPROC ProcessReader | {}".format(msg))
def _create_stream_reader(self, name, stream, queue, callback):
thread = threading.Thread(
name=name, target=self._read_stream, args=(stream, queue, callback)
)
thread.daemon =
True
thread.start()
return thread
def _read_stream(self, stream, queue, callback):
sentinel =
"" if isinstance(stream, io.TextIOBase)
else b
""
for line
in iter(stream.readline, sentinel):
queue.put((line, callback))
# Give a chance to the reading loop to exit without a timeout.
queue.put((b
"",
None))
stream.close()
def start(self, proc):
queue = Queue()
readers = 0
if proc.stdout:
self._create_stream_reader(
"ProcessReaderStdout", proc.stdout, queue, self.stdout_callback
)
readers += 1
if proc.stderr
and proc.stderr != proc.stdout:
self._create_stream_reader(
"ProcessReaderStderr", proc.stderr, queue, self.stderr_callback
)
readers += 1
self.thread = threading.Thread(
name=
"ProcessReader",
target=self._read,
args=(queue, readers),
)
self.thread.daemon =
True
self.thread.start()
self.debug(
"ProcessReader started")
def _read(self, queue, readers):
start_time = time.time()
timeout = self.timeout
if timeout
is not None:
timeout += start_time
output_timeout = self.output_timeout
def get_line():
queue_timeout =
None
if timeout:
queue_timeout = timeout - time.time()
if output_timeout:
if queue_timeout:
queue_timeout = min(queue_timeout, output_timeout)
else:
queue_timeout = output_timeout
return queue.get(timeout=queue_timeout)
try:
# We need to wait for as many `(b"", None)` sentinels as there are
# reader threads setup in start.
for n
in range(readers):
for line, callback
in iter(get_line, (b
"",
None)):
self.got_data.set()
try:
callback(line.rstrip())
except Exception:
traceback.print_exc()
try:
self.finished_callback()
except Exception:
traceback.print_exc()
except Empty:
if timeout
and time.time() < timeout
or not timeout:
self.didOutputTimeout =
True
try:
self.timeout_callback()
except Exception:
traceback.print_exc()
self.debug(
"_read exited")
def is_alive(self):
if self.thread:
return self.thread.is_alive()
return False
def is_still_reading(self, timeout):
self.got_data.clear()
return self.got_data.wait(timeout)
def join(self, timeout=
None):
if self.thread:
self.thread.join(timeout=timeout)
# default output handlers
# these should be callables that take the output line
class StoreOutput(object):
"""accumulate stdout"""
def __init__(self):
self.output = []
def __call__(self, line):
self.output.append(line)
class StreamOutput(object):
"""pass output to a stream and flush"""
def __init__(self, stream, text=
True):
self.stream = stream
self.text = text
def __call__(self, line):
ensure = six.ensure_text
if self.text
else six.ensure_binary
try:
self.stream.write(ensure(line, errors=
"ignore") + ensure(
"\n"))
except TypeError:
print(
"HEY! If you're reading this, you're about to encounter a "
"type error, probably as a result of a conversion from "
"Python 2 to Python 3. This is almost definitely because "
"you're trying to write binary data to a text-encoded "
"stream, or text data to a binary-encoded stream. Check how "
"you're instantiating your ProcessHandler and if the output "
"should be text-encoded, make sure you pass "
"universal_newlines=True.",
file=sys.stderr,
)
raise
self.stream.flush()
class LogOutput(StreamOutput):
"""pass output to a file"""
def __init__(self, filename):
self.file_obj = open(filename,
"a")
StreamOutput.__init__(self, self.file_obj,
True)
def __del__(self):
if self.file_obj
is not None:
self.file_obj.close()
# front end class with the default handlers
class ProcessHandler(ProcessHandlerMixin):
"""
Convenience
class for handling processes
with default output handlers.
By default, all output
is sent to stdout. This can be disabled by setting
the *stream* argument to
None.
If processOutputLine keyword argument
is specified the function
or the
list of functions specified by this argument will be called
for each line
of output; the output will
not be written to stdout automatically then
if stream
is True (the default).
If storeOutput==
True, the output produced by the process will be saved
as self.output.
If logfile
is not None, the output produced by the process will be
appended to the given file.
"""
def __init__(self, cmd, logfile=
None, stream=
True, storeOutput=
True, **kwargs):
kwargs.setdefault(
"processOutputLine", [])
if callable(kwargs[
"processOutputLine"]):
kwargs[
"processOutputLine"] = [kwargs[
"processOutputLine"]]
if logfile:
logoutput = LogOutput(logfile)
kwargs[
"processOutputLine"].append(logoutput)
text = kwargs.get(
"universal_newlines",
False)
or kwargs.get(
"text",
False)
if stream
is True:
if text:
# The encoding of stdout isn't guaranteed to be utf-8. Fix that.
stdout = codecs.getwriter(
"utf-8")(sys.stdout.buffer)
else:
stdout = sys.stdout.buffer
if not kwargs[
"processOutputLine"]:
kwargs[
"processOutputLine"].append(StreamOutput(stdout, text))
elif stream:
streamoutput = StreamOutput(stream, text)
kwargs[
"processOutputLine"].append(streamoutput)
self.output =
None
if storeOutput:
storeoutput = StoreOutput()
self.output = storeoutput.output
kwargs[
"processOutputLine"].append(storeoutput)
ProcessHandlerMixin.__init__(self, cmd, **kwargs)
