Cleanup for stow ---STOW
[dotfiles/.git] / .local / lib / python2.7 / site-packages / trollius / base_events.py
diff --git a/.local/lib/python2.7/site-packages/trollius/base_events.py b/.local/lib/python2.7/site-packages/trollius/base_events.py
new file mode 100644 (file)
index 0000000..c5e6eff
--- /dev/null
@@ -0,0 +1,1260 @@
+"""Base implementation of event loop.
+
+The event loop can be broken up into a multiplexer (the part
+responsible for notifying us of I/O events) and the event loop proper,
+which wraps a multiplexer with functionality for scheduling callbacks,
+immediately or at a given time in the future.
+
+Whenever a public API takes a callback, subsequent positional
+arguments will be passed to the callback if/when it is called.  This
+avoids the proliferation of trivial lambdas implementing closures.
+Keyword arguments for the callback are not supported; this is a
+conscious design decision, leaving the door open for keyword arguments
+to modify the meaning of the API call itself.
+"""
+
+
+import collections
+import heapq
+import inspect
+import logging
+import os
+import socket
+import subprocess
+import sys
+import traceback
+import warnings
+try:
+    from collections import OrderedDict
+except ImportError:
+    # Python 2.6: use ordereddict backport
+    from ordereddict import OrderedDict
+try:
+    from threading import get_ident as _get_thread_ident
+except ImportError:
+    # Python 2
+    from threading import _get_ident as _get_thread_ident
+
+from . import compat
+from . import coroutines
+from . import events
+from . import futures
+from . import tasks
+from .coroutines import coroutine, From, Return
+from .executor import get_default_executor
+from .log import logger
+from .time_monotonic import time_monotonic, time_monotonic_resolution
+
+
+__all__ = ['BaseEventLoop']
+
+
+# Argument for default thread pool executor creation.
+_MAX_WORKERS = 5
+
+# Minimum number of _scheduled timer handles before cleanup of
+# cancelled handles is performed.
+_MIN_SCHEDULED_TIMER_HANDLES = 100
+
+# Minimum fraction of _scheduled timer handles that are cancelled
+# before cleanup of cancelled handles is performed.
+_MIN_CANCELLED_TIMER_HANDLES_FRACTION = 0.5
+
+def _format_handle(handle):
+    cb = handle._callback
+    if inspect.ismethod(cb) and isinstance(cb.__self__, tasks.Task):
+        # format the task
+        return repr(cb.__self__)
+    else:
+        return str(handle)
+
+
+def _format_pipe(fd):
+    if fd == subprocess.PIPE:
+        return '<pipe>'
+    elif fd == subprocess.STDOUT:
+        return '<stdout>'
+    else:
+        return repr(fd)
+
+
+class _StopError(BaseException):
+    """Raised to stop the event loop."""
+
+
+def _check_resolved_address(sock, address):
+    # Ensure that the address is already resolved to avoid the trap of hanging
+    # the entire event loop when the address requires doing a DNS lookup.
+    #
+    # getaddrinfo() is slow (around 10 us per call): this function should only
+    # be called in debug mode
+    family = sock.family
+
+    if family == socket.AF_INET:
+        host, port = address
+    elif family == socket.AF_INET6:
+        host, port = address[:2]
+    else:
+        return
+
+    # On Windows, socket.inet_pton() is only available since Python 3.4
+    if hasattr(socket, 'inet_pton'):
+        # getaddrinfo() is slow and has known issue: prefer inet_pton()
+        # if available
+        try:
+            socket.inet_pton(family, host)
+        except socket.error as exc:
+            raise ValueError("address must be resolved (IP address), "
+                             "got host %r: %s"
+                             % (host, exc))
+    else:
+        # Use getaddrinfo(flags=AI_NUMERICHOST) to ensure that the address is
+        # already resolved.
+        type_mask = 0
+        if hasattr(socket, 'SOCK_NONBLOCK'):
+            type_mask |= socket.SOCK_NONBLOCK
+        if hasattr(socket, 'SOCK_CLOEXEC'):
+            type_mask |= socket.SOCK_CLOEXEC
+        try:
+            socket.getaddrinfo(host, port,
+                               family,
+                               (sock.type & ~type_mask),
+                               sock.proto,
+                               socket.AI_NUMERICHOST)
+        except socket.gaierror as err:
+            raise ValueError("address must be resolved (IP address), "
+                             "got host %r: %s"
+                             % (host, err))
+
+def _raise_stop_error(*args):
+    raise _StopError
+
+
+def _run_until_complete_cb(fut):
+    exc = fut._exception
+    if (isinstance(exc, BaseException)
+    and not isinstance(exc, Exception)):
+        # Issue #22429: run_forever() already finished, no need to
+        # stop it.
+        return
+    _raise_stop_error()
+
+
+class Server(events.AbstractServer):
+
+    def __init__(self, loop, sockets):
+        self._loop = loop
+        self.sockets = sockets
+        self._active_count = 0
+        self._waiters = []
+
+    def __repr__(self):
+        return '<%s sockets=%r>' % (self.__class__.__name__, self.sockets)
+
+    def _attach(self):
+        assert self.sockets is not None
+        self._active_count += 1
+
+    def _detach(self):
+        assert self._active_count > 0
+        self._active_count -= 1
+        if self._active_count == 0 and self.sockets is None:
+            self._wakeup()
+
+    def close(self):
+        sockets = self.sockets
+        if sockets is None:
+            return
+        self.sockets = None
+        for sock in sockets:
+            self._loop._stop_serving(sock)
+        if self._active_count == 0:
+            self._wakeup()
+
+    def _wakeup(self):
+        waiters = self._waiters
+        self._waiters = None
+        for waiter in waiters:
+            if not waiter.done():
+                waiter.set_result(waiter)
+
+    @coroutine
+    def wait_closed(self):
+        if self.sockets is None or self._waiters is None:
+            raise Return()
+        waiter = futures.Future(loop=self._loop)
+        self._waiters.append(waiter)
+        yield From(waiter)
+
+
+class BaseEventLoop(events.AbstractEventLoop):
+
+    def __init__(self):
+        self._timer_cancelled_count = 0
+        self._closed = False
+        self._ready = collections.deque()
+        self._scheduled = []
+        self._default_executor = None
+        self._internal_fds = 0
+        # Identifier of the thread running the event loop, or None if the
+        # event loop is not running
+        self._thread_id = None
+        self._clock_resolution = time_monotonic_resolution
+        self._exception_handler = None
+        self.set_debug(bool(os.environ.get('TROLLIUSDEBUG')))
+        # In debug mode, if the execution of a callback or a step of a task
+        # exceed this duration in seconds, the slow callback/task is logged.
+        self.slow_callback_duration = 0.1
+        self._current_handle = None
+        self._task_factory = None
+        self._coroutine_wrapper_set = False
+
+    def __repr__(self):
+        return ('<%s running=%s closed=%s debug=%s>'
+                % (self.__class__.__name__, self.is_running(),
+                   self.is_closed(), self.get_debug()))
+
+    def create_task(self, coro):
+        """Schedule a coroutine object.
+
+        Return a task object.
+        """
+        self._check_closed()
+        if self._task_factory is None:
+            task = tasks.Task(coro, loop=self)
+            if task._source_traceback:
+                del task._source_traceback[-1]
+        else:
+            task = self._task_factory(self, coro)
+        return task
+
+    def set_task_factory(self, factory):
+        """Set a task factory that will be used by loop.create_task().
+
+        If factory is None the default task factory will be set.
+
+        If factory is a callable, it should have a signature matching
+        '(loop, coro)', where 'loop' will be a reference to the active
+        event loop, 'coro' will be a coroutine object.  The callable
+        must return a Future.
+        """
+        if factory is not None and not callable(factory):
+            raise TypeError('task factory must be a callable or None')
+        self._task_factory = factory
+
+    def get_task_factory(self):
+        """Return a task factory, or None if the default one is in use."""
+        return self._task_factory
+
+    def _make_socket_transport(self, sock, protocol, waiter=None,
+                               extra=None, server=None):
+        """Create socket transport."""
+        raise NotImplementedError
+
+    def _make_ssl_transport(self, rawsock, protocol, sslcontext, waiter=None,
+                            server_side=False, server_hostname=None,
+                            extra=None, server=None):
+        """Create SSL transport."""
+        raise NotImplementedError
+
+    def _make_datagram_transport(self, sock, protocol,
+                                 address=None, waiter=None, extra=None):
+        """Create datagram transport."""
+        raise NotImplementedError
+
+    def _make_read_pipe_transport(self, pipe, protocol, waiter=None,
+                                  extra=None):
+        """Create read pipe transport."""
+        raise NotImplementedError
+
+    def _make_write_pipe_transport(self, pipe, protocol, waiter=None,
+                                   extra=None):
+        """Create write pipe transport."""
+        raise NotImplementedError
+
+    @coroutine
+    def _make_subprocess_transport(self, protocol, args, shell,
+                                   stdin, stdout, stderr, bufsize,
+                                   extra=None, **kwargs):
+        """Create subprocess transport."""
+        raise NotImplementedError
+
+    def _write_to_self(self):
+        """Write a byte to self-pipe, to wake up the event loop.
+
+        This may be called from a different thread.
+
+        The subclass is responsible for implementing the self-pipe.
+        """
+        raise NotImplementedError
+
+    def _process_events(self, event_list):
+        """Process selector events."""
+        raise NotImplementedError
+
+    def _check_closed(self):
+        if self._closed:
+            raise RuntimeError('Event loop is closed')
+
+    def run_forever(self):
+        """Run until stop() is called."""
+        self._check_closed()
+        if self.is_running():
+            raise RuntimeError('Event loop is running.')
+        self._set_coroutine_wrapper(self._debug)
+        self._thread_id = _get_thread_ident()
+        try:
+            while True:
+                try:
+                    self._run_once()
+                except _StopError:
+                    break
+        finally:
+            self._thread_id = None
+            self._set_coroutine_wrapper(False)
+
+    def run_until_complete(self, future):
+        """Run until the Future is done.
+
+        If the argument is a coroutine, it is wrapped in a Task.
+
+        WARNING: It would be disastrous to call run_until_complete()
+        with the same coroutine twice -- it would wrap it in two
+        different Tasks and that can't be good.
+
+        Return the Future's result, or raise its exception.
+        """
+        self._check_closed()
+
+        new_task = not isinstance(future, futures._FUTURE_CLASSES)
+        future = tasks.ensure_future(future, loop=self)
+        if new_task:
+            # An exception is raised if the future didn't complete, so there
+            # is no need to log the "destroy pending task" message
+            future._log_destroy_pending = False
+
+        future.add_done_callback(_run_until_complete_cb)
+        try:
+            self.run_forever()
+        except:
+            if new_task and future.done() and not future.cancelled():
+                # The coroutine raised a BaseException. Consume the exception
+                # to not log a warning, the caller doesn't have access to the
+                # local task.
+                future.exception()
+            raise
+        future.remove_done_callback(_run_until_complete_cb)
+        if not future.done():
+            raise RuntimeError('Event loop stopped before Future completed.')
+
+        return future.result()
+
+    def stop(self):
+        """Stop running the event loop.
+
+        Every callback scheduled before stop() is called will run. Callbacks
+        scheduled after stop() is called will not run. However, those callbacks
+        will run if run_forever is called again later.
+        """
+        self.call_soon(_raise_stop_error)
+
+    def close(self):
+        """Close the event loop.
+
+        This clears the queues and shuts down the executor,
+        but does not wait for the executor to finish.
+
+        The event loop must not be running.
+        """
+        if self.is_running():
+            raise RuntimeError("Cannot close a running event loop")
+        if self._closed:
+            return
+        if self._debug:
+            logger.debug("Close %r", self)
+        self._closed = True
+        self._ready.clear()
+        del self._scheduled[:]
+        executor = self._default_executor
+        if executor is not None:
+            self._default_executor = None
+            executor.shutdown(wait=False)
+
+    def is_closed(self):
+        """Returns True if the event loop was closed."""
+        return self._closed
+
+    # On Python 3.3 and older, objects with a destructor part of a reference
+    # cycle are never destroyed. It's not more the case on Python 3.4 thanks
+    # to the PEP 442.
+    if compat.PY34:
+        def __del__(self):
+            if not self.is_closed():
+                warnings.warn("unclosed event loop %r" % self, ResourceWarning)
+                if not self.is_running():
+                    self.close()
+
+    def is_running(self):
+        """Returns True if the event loop is running."""
+        return (self._thread_id is not None)
+
+    def time(self):
+        """Return the time according to the event loop's clock.
+
+        This is a float expressed in seconds since an epoch, but the
+        epoch, precision, accuracy and drift are unspecified and may
+        differ per event loop.
+        """
+        return time_monotonic()
+
+    def call_later(self, delay, callback, *args):
+        """Arrange for a callback to be called at a given time.
+
+        Return a Handle: an opaque object with a cancel() method that
+        can be used to cancel the call.
+
+        The delay can be an int or float, expressed in seconds.  It is
+        always relative to the current time.
+
+        Each callback will be called exactly once.  If two callbacks
+        are scheduled for exactly the same time, it undefined which
+        will be called first.
+
+        Any positional arguments after the callback will be passed to
+        the callback when it is called.
+        """
+        timer = self.call_at(self.time() + delay, callback, *args)
+        if timer._source_traceback:
+            del timer._source_traceback[-1]
+        return timer
+
+    def call_at(self, when, callback, *args):
+        """Like call_later(), but uses an absolute time.
+
+        Absolute time corresponds to the event loop's time() method.
+        """
+        if (coroutines.iscoroutine(callback)
+        or coroutines.iscoroutinefunction(callback)):
+            raise TypeError("coroutines cannot be used with call_at()")
+        self._check_closed()
+        if self._debug:
+            self._check_thread()
+        timer = events.TimerHandle(when, callback, args, self)
+        if timer._source_traceback:
+            del timer._source_traceback[-1]
+        heapq.heappush(self._scheduled, timer)
+        timer._scheduled = True
+        return timer
+
+    def call_soon(self, callback, *args):
+        """Arrange for a callback to be called as soon as possible.
+
+        This operates as a FIFO queue: callbacks are called in the
+        order in which they are registered.  Each callback will be
+        called exactly once.
+
+        Any positional arguments after the callback will be passed to
+        the callback when it is called.
+        """
+        if self._debug:
+            self._check_thread()
+        handle = self._call_soon(callback, args)
+        if handle._source_traceback:
+            del handle._source_traceback[-1]
+        return handle
+
+    def _call_soon(self, callback, args):
+        if (coroutines.iscoroutine(callback)
+        or coroutines.iscoroutinefunction(callback)):
+            raise TypeError("coroutines cannot be used with call_soon()")
+        self._check_closed()
+        handle = events.Handle(callback, args, self)
+        if handle._source_traceback:
+            del handle._source_traceback[-1]
+        self._ready.append(handle)
+        return handle
+
+    def _check_thread(self):
+        """Check that the current thread is the thread running the event loop.
+
+        Non-thread-safe methods of this class make this assumption and will
+        likely behave incorrectly when the assumption is violated.
+
+        Should only be called when (self._debug == True).  The caller is
+        responsible for checking this condition for performance reasons.
+        """
+        if self._thread_id is None:
+            return
+        thread_id = _get_thread_ident()
+        if thread_id != self._thread_id:
+            raise RuntimeError(
+                "Non-thread-safe operation invoked on an event loop other "
+                "than the current one")
+
+    def call_soon_threadsafe(self, callback, *args):
+        """Like call_soon(), but thread-safe."""
+        handle = self._call_soon(callback, args)
+        if handle._source_traceback:
+            del handle._source_traceback[-1]
+        self._write_to_self()
+        return handle
+
+    def run_in_executor(self, executor, func, *args):
+        if (coroutines.iscoroutine(func)
+        or coroutines.iscoroutinefunction(func)):
+            raise TypeError("coroutines cannot be used with run_in_executor()")
+        self._check_closed()
+        if isinstance(func, events.Handle):
+            assert not args
+            assert not isinstance(func, events.TimerHandle)
+            if func._cancelled:
+                f = futures.Future(loop=self)
+                f.set_result(None)
+                return f
+            func, args = func._callback, func._args
+        if executor is None:
+            executor = self._default_executor
+            if executor is None:
+                executor = get_default_executor()
+                self._default_executor = executor
+        return futures.wrap_future(executor.submit(func, *args), loop=self)
+
+    def set_default_executor(self, executor):
+        self._default_executor = executor
+
+    def _getaddrinfo_debug(self, host, port, family, type, proto, flags):
+        msg = ["%s:%r" % (host, port)]
+        if family:
+            msg.append('family=%r' % family)
+        if type:
+            msg.append('type=%r' % type)
+        if proto:
+            msg.append('proto=%r' % proto)
+        if flags:
+            msg.append('flags=%r' % flags)
+        msg = ', '.join(msg)
+        logger.debug('Get address info %s', msg)
+
+        t0 = self.time()
+        addrinfo = socket.getaddrinfo(host, port, family, type, proto, flags)
+        dt = self.time() - t0
+
+        msg = ('Getting address info %s took %.3f ms: %r'
+               % (msg, dt * 1e3, addrinfo))
+        if dt >= self.slow_callback_duration:
+            logger.info(msg)
+        else:
+            logger.debug(msg)
+        return addrinfo
+
+    def getaddrinfo(self, host, port,
+                    family=0, type=0, proto=0, flags=0):
+        if self._debug:
+            return self.run_in_executor(None, self._getaddrinfo_debug,
+                                        host, port, family, type, proto, flags)
+        else:
+            return self.run_in_executor(None, socket.getaddrinfo,
+                                        host, port, family, type, proto, flags)
+
+    def getnameinfo(self, sockaddr, flags=0):
+        return self.run_in_executor(None, socket.getnameinfo, sockaddr, flags)
+
+    @coroutine
+    def create_connection(self, protocol_factory, host=None, port=None,
+                          ssl=None, family=0, proto=0, flags=0, sock=None,
+                          local_addr=None, server_hostname=None):
+        """Connect to a TCP server.
+
+        Create a streaming transport connection to a given Internet host and
+        port: socket family AF_INET or socket.AF_INET6 depending on host (or
+        family if specified), socket type SOCK_STREAM. protocol_factory must be
+        a callable returning a protocol instance.
+
+        This method is a coroutine which will try to establish the connection
+        in the background.  When successful, the coroutine returns a
+        (transport, protocol) pair.
+        """
+        if server_hostname is not None and not ssl:
+            raise ValueError('server_hostname is only meaningful with ssl')
+
+        if server_hostname is None and ssl:
+            # Use host as default for server_hostname.  It is an error
+            # if host is empty or not set, e.g. when an
+            # already-connected socket was passed or when only a port
+            # is given.  To avoid this error, you can pass
+            # server_hostname='' -- this will bypass the hostname
+            # check.  (This also means that if host is a numeric
+            # IP/IPv6 address, we will attempt to verify that exact
+            # address; this will probably fail, but it is possible to
+            # create a certificate for a specific IP address, so we
+            # don't judge it here.)
+            if not host:
+                raise ValueError('You must set server_hostname '
+                                 'when using ssl without a host')
+            server_hostname = host
+
+        if host is not None or port is not None:
+            if sock is not None:
+                raise ValueError(
+                    'host/port and sock can not be specified at the same time')
+
+            f1 = self.getaddrinfo(
+                host, port, family=family,
+                type=socket.SOCK_STREAM, proto=proto, flags=flags)
+            fs = [f1]
+            if local_addr is not None:
+                f2 = self.getaddrinfo(
+                    *local_addr, family=family,
+                    type=socket.SOCK_STREAM, proto=proto, flags=flags)
+                fs.append(f2)
+            else:
+                f2 = None
+
+            yield From(tasks.wait(fs, loop=self))
+
+            infos = f1.result()
+            if not infos:
+                raise socket.error('getaddrinfo() returned empty list')
+            if f2 is not None:
+                laddr_infos = f2.result()
+                if not laddr_infos:
+                    raise socket.error('getaddrinfo() returned empty list')
+
+            exceptions = []
+            for family, type, proto, cname, address in infos:
+                try:
+                    sock = socket.socket(family=family, type=type, proto=proto)
+                    sock.setblocking(False)
+                    if f2 is not None:
+                        for _, _, _, _, laddr in laddr_infos:
+                            try:
+                                sock.bind(laddr)
+                                break
+                            except socket.error as exc:
+                                exc = socket.error(
+                                    exc.errno, 'error while '
+                                    'attempting to bind on address '
+                                    '{0!r}: {1}'.format(
+                                        laddr, exc.strerror.lower()))
+                                exceptions.append(exc)
+                        else:
+                            sock.close()
+                            sock = None
+                            continue
+                    if self._debug:
+                        logger.debug("connect %r to %r", sock, address)
+                    yield From(self.sock_connect(sock, address))
+                except socket.error as exc:
+                    if sock is not None:
+                        sock.close()
+                    exceptions.append(exc)
+                except:
+                    if sock is not None:
+                        sock.close()
+                    raise
+                else:
+                    break
+            else:
+                if len(exceptions) == 1:
+                    raise exceptions[0]
+                else:
+                    # If they all have the same str(), raise one.
+                    model = str(exceptions[0])
+                    if all(str(exc) == model for exc in exceptions):
+                        raise exceptions[0]
+                    # Raise a combined exception so the user can see all
+                    # the various error messages.
+                    raise socket.error('Multiple exceptions: {0}'.format(
+                        ', '.join(str(exc) for exc in exceptions)))
+
+        elif sock is None:
+            raise ValueError(
+                'host and port was not specified and no sock specified')
+
+        sock.setblocking(False)
+
+        transport, protocol = yield From(self._create_connection_transport(
+            sock, protocol_factory, ssl, server_hostname))
+        if self._debug:
+            # Get the socket from the transport because SSL transport closes
+            # the old socket and creates a new SSL socket
+            sock = transport.get_extra_info('socket')
+            logger.debug("%r connected to %s:%r: (%r, %r)",
+                         sock, host, port, transport, protocol)
+        raise Return(transport, protocol)
+
+    @coroutine
+    def _create_connection_transport(self, sock, protocol_factory, ssl,
+                                     server_hostname):
+        protocol = protocol_factory()
+        waiter = futures.Future(loop=self)
+        if ssl:
+            sslcontext = None if isinstance(ssl, bool) else ssl
+            transport = self._make_ssl_transport(
+                sock, protocol, sslcontext, waiter,
+                server_side=False, server_hostname=server_hostname)
+        else:
+            transport = self._make_socket_transport(sock, protocol, waiter)
+
+        try:
+            yield From(waiter)
+        except:
+            transport.close()
+            raise
+
+        raise Return(transport, protocol)
+
+    @coroutine
+    def create_datagram_endpoint(self, protocol_factory,
+                                 local_addr=None, remote_addr=None,
+                                 family=0, proto=0, flags=0):
+        """Create datagram connection."""
+        if not (local_addr or remote_addr):
+            if family == 0:
+                raise ValueError('unexpected address family')
+            addr_pairs_info = (((family, proto), (None, None)),)
+        else:
+            # join address by (family, protocol)
+            addr_infos = OrderedDict()
+            for idx, addr in ((0, local_addr), (1, remote_addr)):
+                if addr is not None:
+                    assert isinstance(addr, tuple) and len(addr) == 2, (
+                        '2-tuple is expected')
+
+                    infos = yield From(self.getaddrinfo(
+                        *addr, family=family, type=socket.SOCK_DGRAM,
+                        proto=proto, flags=flags))
+                    if not infos:
+                        raise socket.error('getaddrinfo() returned empty list')
+
+                    for fam, _, pro, _, address in infos:
+                        key = (fam, pro)
+                        if key not in addr_infos:
+                            addr_infos[key] = [None, None]
+                        addr_infos[key][idx] = address
+
+            # each addr has to have info for each (family, proto) pair
+            addr_pairs_info = [
+                (key, addr_pair) for key, addr_pair in addr_infos.items()
+                if not ((local_addr and addr_pair[0] is None) or
+                        (remote_addr and addr_pair[1] is None))]
+
+            if not addr_pairs_info:
+                raise ValueError('can not get address information')
+
+        exceptions = []
+
+        for ((family, proto),
+             (local_address, remote_address)) in addr_pairs_info:
+            sock = None
+            r_addr = None
+            try:
+                sock = socket.socket(
+                    family=family, type=socket.SOCK_DGRAM, proto=proto)
+                sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+                sock.setblocking(False)
+
+                if local_addr:
+                    sock.bind(local_address)
+                if remote_addr:
+                    yield From(self.sock_connect(sock, remote_address))
+                    r_addr = remote_address
+            except socket.error as exc:
+                if sock is not None:
+                    sock.close()
+                exceptions.append(exc)
+            except:
+                if sock is not None:
+                    sock.close()
+                raise
+            else:
+                break
+        else:
+            raise exceptions[0]
+
+        protocol = protocol_factory()
+        waiter = futures.Future(loop=self)
+        transport = self._make_datagram_transport(sock, protocol, r_addr,
+                                                  waiter)
+        if self._debug:
+            if local_addr:
+                logger.info("Datagram endpoint local_addr=%r remote_addr=%r "
+                            "created: (%r, %r)",
+                            local_addr, remote_addr, transport, protocol)
+            else:
+                logger.debug("Datagram endpoint remote_addr=%r created: "
+                             "(%r, %r)",
+                             remote_addr, transport, protocol)
+
+        try:
+            yield From(waiter)
+        except:
+            transport.close()
+            raise
+
+        raise Return(transport, protocol)
+
+    @coroutine
+    def create_server(self, protocol_factory, host=None, port=None,
+                      family=socket.AF_UNSPEC,
+                      flags=socket.AI_PASSIVE,
+                      sock=None,
+                      backlog=100,
+                      ssl=None,
+                      reuse_address=None):
+        """Create a TCP server bound to host and port.
+
+        Return a Server object which can be used to stop the service.
+
+        This method is a coroutine.
+        """
+        if isinstance(ssl, bool):
+            raise TypeError('ssl argument must be an SSLContext or None')
+        if host is not None or port is not None:
+            if sock is not None:
+                raise ValueError(
+                    'host/port and sock can not be specified at the same time')
+
+            AF_INET6 = getattr(socket, 'AF_INET6', 0)
+            if reuse_address is None:
+                reuse_address = os.name == 'posix' and sys.platform != 'cygwin'
+            sockets = []
+            if host == '':
+                host = None
+
+            infos = yield From(self.getaddrinfo(
+                host, port, family=family,
+                type=socket.SOCK_STREAM, proto=0, flags=flags))
+            if not infos:
+                raise socket.error('getaddrinfo() returned empty list')
+
+            completed = False
+            try:
+                for res in infos:
+                    af, socktype, proto, canonname, sa = res
+                    try:
+                        sock = socket.socket(af, socktype, proto)
+                    except socket.error:
+                        # Assume it's a bad family/type/protocol combination.
+                        if self._debug:
+                            logger.warning('create_server() failed to create '
+                                           'socket.socket(%r, %r, %r)',
+                                           af, socktype, proto, exc_info=True)
+                        continue
+                    sockets.append(sock)
+                    if reuse_address:
+                        sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR,
+                                        True)
+                    # Disable IPv4/IPv6 dual stack support (enabled by
+                    # default on Linux) which makes a single socket
+                    # listen on both address families.
+                    if af == AF_INET6 and hasattr(socket, 'IPPROTO_IPV6'):
+                        sock.setsockopt(socket.IPPROTO_IPV6,
+                                        socket.IPV6_V6ONLY,
+                                        True)
+                    try:
+                        sock.bind(sa)
+                    except socket.error as err:
+                        raise socket.error(err.errno,
+                                           'error while attempting '
+                                           'to bind on address %r: %s'
+                                           % (sa, err.strerror.lower()))
+                completed = True
+            finally:
+                if not completed:
+                    for sock in sockets:
+                        sock.close()
+        else:
+            if sock is None:
+                raise ValueError('Neither host/port nor sock were specified')
+            sockets = [sock]
+
+        server = Server(self, sockets)
+        for sock in sockets:
+            sock.listen(backlog)
+            sock.setblocking(False)
+            self._start_serving(protocol_factory, sock, ssl, server)
+        if self._debug:
+            logger.info("%r is serving", server)
+        raise Return(server)
+
+    @coroutine
+    def connect_read_pipe(self, protocol_factory, pipe):
+        protocol = protocol_factory()
+        waiter = futures.Future(loop=self)
+        transport = self._make_read_pipe_transport(pipe, protocol, waiter)
+
+        try:
+            yield From(waiter)
+        except:
+            transport.close()
+            raise
+
+        if self._debug:
+            logger.debug('Read pipe %r connected: (%r, %r)',
+                         pipe.fileno(), transport, protocol)
+        raise Return(transport, protocol)
+
+    @coroutine
+    def connect_write_pipe(self, protocol_factory, pipe):
+        protocol = protocol_factory()
+        waiter = futures.Future(loop=self)
+        transport = self._make_write_pipe_transport(pipe, protocol, waiter)
+
+        try:
+            yield From(waiter)
+        except:
+            transport.close()
+            raise
+
+        if self._debug:
+            logger.debug('Write pipe %r connected: (%r, %r)',
+                         pipe.fileno(), transport, protocol)
+        raise Return(transport, protocol)
+
+    def _log_subprocess(self, msg, stdin, stdout, stderr):
+        info = [msg]
+        if stdin is not None:
+            info.append('stdin=%s' % _format_pipe(stdin))
+        if stdout is not None and stderr == subprocess.STDOUT:
+            info.append('stdout=stderr=%s' % _format_pipe(stdout))
+        else:
+            if stdout is not None:
+                info.append('stdout=%s' % _format_pipe(stdout))
+            if stderr is not None:
+                info.append('stderr=%s' % _format_pipe(stderr))
+        logger.debug(' '.join(info))
+
+    @coroutine
+    def subprocess_shell(self, protocol_factory, cmd, stdin=subprocess.PIPE,
+                         stdout=subprocess.PIPE, stderr=subprocess.PIPE,
+                         universal_newlines=False, shell=True, bufsize=0,
+                         **kwargs):
+        if not isinstance(cmd, compat.string_types):
+            raise ValueError("cmd must be a string")
+        if universal_newlines:
+            raise ValueError("universal_newlines must be False")
+        if not shell:
+            raise ValueError("shell must be True")
+        if bufsize != 0:
+            raise ValueError("bufsize must be 0")
+        protocol = protocol_factory()
+        if self._debug:
+            # don't log parameters: they may contain sensitive information
+            # (password) and may be too long
+            debug_log = 'run shell command %r' % cmd
+            self._log_subprocess(debug_log, stdin, stdout, stderr)
+        transport = yield From(self._make_subprocess_transport(
+            protocol, cmd, True, stdin, stdout, stderr, bufsize, **kwargs))
+        if self._debug:
+            logger.info('%s: %r' % (debug_log, transport))
+        raise Return(transport, protocol)
+
+    @coroutine
+    def subprocess_exec(self, protocol_factory, program, *args, **kwargs):
+        stdin = kwargs.pop('stdin', subprocess.PIPE)
+        stdout = kwargs.pop('stdout', subprocess.PIPE)
+        stderr = kwargs.pop('stderr', subprocess.PIPE)
+        universal_newlines = kwargs.pop('universal_newlines', False)
+        shell = kwargs.pop('shell', False)
+        bufsize = kwargs.pop('bufsize', 0)
+        if universal_newlines:
+            raise ValueError("universal_newlines must be False")
+        if shell:
+            raise ValueError("shell must be False")
+        if bufsize != 0:
+            raise ValueError("bufsize must be 0")
+        popen_args = (program,) + args
+        for arg in popen_args:
+            if not isinstance(arg, compat.string_types ):
+                raise TypeError("program arguments must be "
+                                "a bytes or text string, not %s"
+                                % type(arg).__name__)
+        protocol = protocol_factory()
+        if self._debug:
+            # don't log parameters: they may contain sensitive information
+            # (password) and may be too long
+            debug_log = 'execute program %r' % program
+            self._log_subprocess(debug_log, stdin, stdout, stderr)
+        transport = yield From(self._make_subprocess_transport(
+            protocol, popen_args, False, stdin, stdout, stderr,
+            bufsize, **kwargs))
+        if self._debug:
+            logger.info('%s: %r' % (debug_log, transport))
+        raise Return(transport, protocol)
+
+    def set_exception_handler(self, handler):
+        """Set handler as the new event loop exception handler.
+
+        If handler is None, the default exception handler will
+        be set.
+
+        If handler is a callable object, it should have a
+        signature matching '(loop, context)', where 'loop'
+        will be a reference to the active event loop, 'context'
+        will be a dict object (see `call_exception_handler()`
+        documentation for details about context).
+        """
+        if handler is not None and not callable(handler):
+            raise TypeError('A callable object or None is expected, '
+                            'got {0!r}'.format(handler))
+        self._exception_handler = handler
+
+    def default_exception_handler(self, context):
+        """Default exception handler.
+
+        This is called when an exception occurs and no exception
+        handler is set, and can be called by a custom exception
+        handler that wants to defer to the default behavior.
+
+        The context parameter has the same meaning as in
+        `call_exception_handler()`.
+        """
+        message = context.get('message')
+        if not message:
+            message = 'Unhandled exception in event loop'
+
+        exception = context.get('exception')
+        if exception is not None:
+            if hasattr(exception, '__traceback__'):
+                # Python 3
+                tb = exception.__traceback__
+            else:
+                # call_exception_handler() is usually called indirectly
+                # from an except block. If it's not the case, the traceback
+                # is undefined...
+                tb = sys.exc_info()[2]
+            exc_info = (type(exception), exception, tb)
+        else:
+            exc_info = False
+
+        if ('source_traceback' not in context
+        and self._current_handle is not None
+        and self._current_handle._source_traceback):
+            context['handle_traceback'] = self._current_handle._source_traceback
+
+        log_lines = [message]
+        for key in sorted(context):
+            if key in ('message', 'exception'):
+                continue
+            value = context[key]
+            if key == 'source_traceback':
+                tb = ''.join(traceback.format_list(value))
+                value = 'Object created at (most recent call last):\n'
+                value += tb.rstrip()
+            elif key == 'handle_traceback':
+                tb = ''.join(traceback.format_list(value))
+                value = 'Handle created at (most recent call last):\n'
+                value += tb.rstrip()
+            else:
+                value = repr(value)
+            log_lines.append('{0}: {1}'.format(key, value))
+
+        logger.error('\n'.join(log_lines), exc_info=exc_info)
+
+    def call_exception_handler(self, context):
+        """Call the current event loop's exception handler.
+
+        The context argument is a dict containing the following keys:
+
+        - 'message': Error message;
+        - 'exception' (optional): Exception object;
+        - 'future' (optional): Future instance;
+        - 'handle' (optional): Handle instance;
+        - 'protocol' (optional): Protocol instance;
+        - 'transport' (optional): Transport instance;
+        - 'socket' (optional): Socket instance.
+
+        New keys maybe introduced in the future.
+
+        Note: do not overload this method in an event loop subclass.
+        For custom exception handling, use the
+        `set_exception_handler()` method.
+        """
+        if self._exception_handler is None:
+            try:
+                self.default_exception_handler(context)
+            except Exception:
+                # Second protection layer for unexpected errors
+                # in the default implementation, as well as for subclassed
+                # event loops with overloaded "default_exception_handler".
+                logger.error('Exception in default exception handler',
+                             exc_info=True)
+        else:
+            try:
+                self._exception_handler(self, context)
+            except Exception as exc:
+                # Exception in the user set custom exception handler.
+                try:
+                    # Let's try default handler.
+                    self.default_exception_handler({
+                        'message': 'Unhandled error in exception handler',
+                        'exception': exc,
+                        'context': context,
+                    })
+                except Exception:
+                    # Guard 'default_exception_handler' in case it is
+                    # overloaded.
+                    logger.error('Exception in default exception handler '
+                                 'while handling an unexpected error '
+                                 'in custom exception handler',
+                                 exc_info=True)
+
+    def _add_callback(self, handle):
+        """Add a Handle to _scheduled (TimerHandle) or _ready."""
+        assert isinstance(handle, events.Handle), 'A Handle is required here'
+        if handle._cancelled:
+            return
+        assert not isinstance(handle, events.TimerHandle)
+        self._ready.append(handle)
+
+    def _add_callback_signalsafe(self, handle):
+        """Like _add_callback() but called from a signal handler."""
+        self._add_callback(handle)
+        self._write_to_self()
+
+    def _timer_handle_cancelled(self, handle):
+        """Notification that a TimerHandle has been cancelled."""
+        if handle._scheduled:
+            self._timer_cancelled_count += 1
+
+    def _run_once(self):
+        """Run one full iteration of the event loop.
+
+        This calls all currently ready callbacks, polls for I/O,
+        schedules the resulting callbacks, and finally schedules
+        'call_later' callbacks.
+        """
+
+        sched_count = len(self._scheduled)
+        if (sched_count > _MIN_SCHEDULED_TIMER_HANDLES and
+            float(self._timer_cancelled_count) / sched_count >
+                _MIN_CANCELLED_TIMER_HANDLES_FRACTION):
+            # Remove delayed calls that were cancelled if their number
+            # is too high
+            new_scheduled = []
+            for handle in self._scheduled:
+                if handle._cancelled:
+                    handle._scheduled = False
+                else:
+                    new_scheduled.append(handle)
+
+            heapq.heapify(new_scheduled)
+            self._scheduled = new_scheduled
+            self._timer_cancelled_count = 0
+        else:
+            # Remove delayed calls that were cancelled from head of queue.
+            while self._scheduled and self._scheduled[0]._cancelled:
+                self._timer_cancelled_count -= 1
+                handle = heapq.heappop(self._scheduled)
+                handle._scheduled = False
+
+        timeout = None
+        if self._ready:
+            timeout = 0
+        elif self._scheduled:
+            # Compute the desired timeout.
+            when = self._scheduled[0]._when
+            timeout = max(0, when - self.time())
+
+        if self._debug and timeout != 0:
+            t0 = self.time()
+            event_list = self._selector.select(timeout)
+            dt = self.time() - t0
+            if dt >= 1.0:
+                level = logging.INFO
+            else:
+                level = logging.DEBUG
+            nevent = len(event_list)
+            if timeout is None:
+                logger.log(level, 'poll took %.3f ms: %s events',
+                           dt * 1e3, nevent)
+            elif nevent:
+                logger.log(level,
+                           'poll %.3f ms took %.3f ms: %s events',
+                           timeout * 1e3, dt * 1e3, nevent)
+            elif dt >= 1.0:
+                logger.log(level,
+                           'poll %.3f ms took %.3f ms: timeout',
+                           timeout * 1e3, dt * 1e3)
+        else:
+            event_list = self._selector.select(timeout)
+        self._process_events(event_list)
+
+        # Handle 'later' callbacks that are ready.
+        end_time = self.time() + self._clock_resolution
+        while self._scheduled:
+            handle = self._scheduled[0]
+            if handle._when >= end_time:
+                break
+            handle = heapq.heappop(self._scheduled)
+            handle._scheduled = False
+            self._ready.append(handle)
+
+        # This is the only place where callbacks are actually *called*.
+        # All other places just add them to ready.
+        # Note: We run all currently scheduled callbacks, but not any
+        # callbacks scheduled by callbacks run this time around --
+        # they will be run the next time (after another I/O poll).
+        # Use an idiom that is thread-safe without using locks.
+        ntodo = len(self._ready)
+        for i in range(ntodo):
+            handle = self._ready.popleft()
+            if handle._cancelled:
+                continue
+            if self._debug:
+                try:
+                    self._current_handle = handle
+                    t0 = self.time()
+                    handle._run()
+                    dt = self.time() - t0
+                    if dt >= self.slow_callback_duration:
+                        logger.warning('Executing %s took %.3f seconds',
+                                       _format_handle(handle), dt)
+                finally:
+                    self._current_handle = None
+            else:
+                handle._run()
+        handle = None  # Needed to break cycles when an exception occurs.
+
+    def _set_coroutine_wrapper(self, enabled):
+        try:
+            set_wrapper = sys.set_coroutine_wrapper
+            get_wrapper = sys.get_coroutine_wrapper
+        except AttributeError:
+            return
+
+        enabled = bool(enabled)
+        if self._coroutine_wrapper_set == enabled:
+            return
+
+        wrapper = coroutines.debug_wrapper
+        current_wrapper = get_wrapper()
+
+        if enabled:
+            if current_wrapper not in (None, wrapper):
+                warnings.warn(
+                    "loop.set_debug(True): cannot set debug coroutine "
+                    "wrapper; another wrapper is already set %r" %
+                    current_wrapper, RuntimeWarning)
+            else:
+                set_wrapper(wrapper)
+                self._coroutine_wrapper_set = True
+        else:
+            if current_wrapper not in (None, wrapper):
+                warnings.warn(
+                    "loop.set_debug(False): cannot unset debug coroutine "
+                    "wrapper; another wrapper was set %r" %
+                    current_wrapper, RuntimeWarning)
+            else:
+                set_wrapper(None)
+                self._coroutine_wrapper_set = False
+
+    def get_debug(self):
+        return self._debug
+
+    def set_debug(self, enabled):
+        self._debug = enabled
+
+        if self.is_running():
+            self._set_coroutine_wrapper(enabled)