]> git-server-git.apps.pok.os.sepia.ceph.com Git - remoto.git/commitdiff
removed orig file left behind
authorAlfredo Deza <alfredo@deza.pe>
Tue, 25 Feb 2014 14:50:28 +0000 (09:50 -0500)
committerAlfredo Deza <alfredo@deza.pe>
Tue, 25 Feb 2014 14:50:28 +0000 (09:50 -0500)
remoto/lib/execnet/gateway_base.py.orig [deleted file]

diff --git a/remoto/lib/execnet/gateway_base.py.orig b/remoto/lib/execnet/gateway_base.py.orig
deleted file mode 100644 (file)
index 9dc8dcd..0000000
+++ /dev/null
@@ -1,1474 +0,0 @@
-"""
-base execnet gateway code send to the other side for bootstrapping.
-
-NOTE: aims to be compatible to Python 2.5-3.X, Jython and IronPython
-
-(C) 2004-2013 Holger Krekel, Armin Rigo, Benjamin Peterson, Ronny Pfannschmidt and others
-"""
-from __future__ import with_statement
-import sys, os, weakref
-import traceback, struct
-
-# NOTE that we want to avoid try/except style importing
-# to avoid setting sys.exc_info() during import
-#
-
-ISPY3 = sys.version_info >= (3, 0)
-if ISPY3:
-    from io import BytesIO
-    exec("def do_exec(co, loc): exec(co, loc)\n"
-         "def reraise(cls, val, tb): raise val\n")
-    unicode = str
-    _long_type = int
-    from _thread import interrupt_main
-else:
-    from StringIO import StringIO as BytesIO
-    exec("def do_exec(co, loc): exec co in loc\n"
-         "def reraise(cls, val, tb): raise cls, val, tb\n")
-    bytes = str
-    _long_type = long
-    try:
-        from thread import interrupt_main
-    except ImportError:
-        interrupt_main = None
-
-class EmptySemaphore:
-    acquire = release = lambda self: None
-
-def get_execmodel(backend):
-    if hasattr(backend, "backend"):
-        return backend
-    if backend == "thread":
-        importdef = {
-            'get_ident': ['thread::get_ident', '_thread::get_ident'],
-            '_start_new_thread': ['thread::start_new_thread',
-                                  '_thread::start_new_thread'],
-            'threading': ["threading",],
-            'queue': ["queue", "Queue"],
-            'sleep': ['time::sleep'],
-            'subprocess': ['subprocess'],
-            'socket': ['socket'],
-            '_fdopen': ['os::fdopen'],
-            '_lock': ['threading'],
-            '_event': ['threading'],
-        }
-        def exec_start(self, func, args=()):
-            self._start_new_thread(func, args)
-
-    elif backend == "eventlet":
-        importdef = {
-            'get_ident': ['eventlet.green.thread::get_ident'],
-            '_spawn_n': ['eventlet::spawn_n'],
-            'threading': ['eventlet.green.threading'],
-            'queue': ["eventlet.queue"],
-            'sleep': ['eventlet::sleep'],
-            'subprocess': ['eventlet.green.subprocess'],
-            'socket': ['eventlet.green.socket'],
-            '_fdopen': ['eventlet.green.os::fdopen'],
-            '_lock': ['eventlet.green.threading'],
-            '_event': ['eventlet.green.threading'],
-        }
-        def exec_start(self, func, args=()):
-            self._spawn_n(func, *args)
-    elif backend == "gevent":
-        importdef = {
-            'get_ident': ['gevent.thread::get_ident'],
-            '_spawn_n': ['gevent::spawn'],
-            'threading': ['threading'],
-            'queue': ["gevent.queue"],
-            'sleep': ['gevent::sleep'],
-            'subprocess': ['gevent.subprocess'],
-            'socket': ['gevent.socket'],
-            # XXX
-            '_fdopen': ['gevent.fileobject::FileObjectThread'],
-            '_lock': ['gevent.lock'],
-            '_event': ['gevent.event'],
-        }
-        def exec_start(self, func, args=()):
-            self._spawn_n(func, *args)
-    else:
-        raise ValueError("unknown execmodel %r" %(backend,))
-
-    class ExecModel:
-        def __init__(self, name):
-            self._importdef = importdef
-            self.backend = name
-            self._count = 0
-
-        def __repr__(self):
-            return "<ExecModel %r>" % self.backend
-
-        def __getattr__(self, name):
-            locs = self._importdef.get(name)
-            if locs is None:
-                raise AttributeError(name)
-            for loc in locs:
-                parts = loc.split("::")
-                loc = parts.pop(0)
-                try:
-                    mod = __import__(loc, None, None, "__doc__")
-                except ImportError:
-                    pass
-                else:
-                    if parts:
-                        mod = getattr(mod, parts[0])
-                    setattr(self, name, mod)
-                    return mod
-            raise AttributeError(name)
-
-        start = exec_start
-
-        def fdopen(self, fd, mode, bufsize=1):
-            return self._fdopen(fd, mode, bufsize)
-
-        def WorkerPool(self, size=None, hasprimary=False):
-            return WorkerPool(self, size, hasprimary=hasprimary)
-
-        def Semaphore(self, size=None):
-            if size is None:
-                return EmptySemaphore()
-            return self._lock.Semaphore(size)
-
-        def Lock(self):
-            return self._lock.RLock()
-
-        def RLock(self):
-            return self._lock.RLock()
-
-        def Event(self):
-            event = self._event.Event()
-            if sys.version_info < (2,7):
-                # patch wait function to return event state instead of None
-                real_wait = event.wait
-                def wait(timeout=None):
-                    real_wait(timeout=timeout)
-                    return event.isSet()
-                event.wait = wait
-            return event
-
-        def PopenPiped(self, args):
-            PIPE = self.subprocess.PIPE
-            return self.subprocess.Popen(args, stdout=PIPE, stdin=PIPE)
-
-
-    return ExecModel(backend)
-
-
-class Reply(object):
-    """ reply instances provide access to the result
-        of a function execution that got dispatched
-        through WorkerPool.spawn()
-    """
-    def __init__(self, task, threadmodel):
-        self.task = task
-        self._result_ready = threadmodel.Event()
-        self.running = True
-
-    def get(self, timeout=None):
-        """ get the result object from an asynchronous function execution.
-            if the function execution raised an exception,
-            then calling get() will reraise that exception
-            including its traceback.
-        """
-        self.waitfinish(timeout)
-        try:
-            return self._result
-        except AttributeError:
-            reraise(*(self._excinfo[:3]))  # noqa
-
-    def waitfinish(self, timeout=None):
-        if not self._result_ready.wait(timeout):
-            raise IOError("timeout waiting for %r" %(self.task, ))
-
-    def run(self):
-        func, args, kwargs = self.task
-        try:
-            try:
-                self._result = func(*args, **kwargs)
-            except:
-                self._excinfo = sys.exc_info()
-        finally:
-            self._result_ready.set()
-            self.running = False
-
-
-class WorkerPool(object):
-    """ A WorkerPool allows to spawn function executions
-        to threads, returning a reply object on which you
-        can ask for the result (and get exceptions reraised)
-    """
-    def __init__(self, execmodel, size=None, hasprimary=False):
-        """ by default allow unlimited number of spawns. """
-        self.execmodel = execmodel
-        self._size = size
-        self._running_lock = self.execmodel.Lock()
-        self._sem = self.execmodel.Semaphore(size)
-        self._running = set()
-        self._shutdown_event = self.execmodel.Event()
-        if hasprimary:
-            if self.execmodel.backend != "thread":
-                raise ValueError("hasprimary=True requires thread model")
-            self._primary_thread_event = self.execmodel.Event()
-
-    def integrate_as_primary_thread(self):
-        """ integrate the thread with which we are called as a primary
-        thread to dispatch to when spawn is called.
-        """
-        assert self.execmodel.backend == "thread", self.execmodel
-        # XXX insert check if we really are in the main thread
-        primary_thread_event = self._primary_thread_event
-        # interacts with code at REF1
-        while not self._shutdown_event.isSet():
-            primary_thread_event.wait()
-            func, args, kwargs = self._primary_thread_task
-            if func is None:  # waitall() woke us up to finish the loop
-                break
-            func(*args, **kwargs)
-            primary_thread_event.clear()
-
-    def shutdown(self):
-        self._shutdown_event.set()
-
-    def wait_for_shutdown(self, timeout=None):
-        return self._shutdown_event.wait(timeout=timeout)
-
-    def active_count(self):
-        return len(self._running)
-
-    def spawn(self, func, *args, **kwargs):
-        """ return Reply object for the asynchronous dispatch
-            of the given func(*args, **kwargs).
-        """
-        reply = Reply((func, args, kwargs), self.execmodel)
-        def run_and_release():
-            reply.run()
-            with self._running_lock:
-                self._running.remove(reply)
-                self._sem.release()
-                if not self._running:
-                    try:
-                        self._waitall_event.set()
-                    except AttributeError:
-                        pass
-        self._sem.acquire()
-        with self._running_lock:
-            self._running.add(reply)
-            # REF1 in 'thread' model we give priority to running in main thread
-            primary_thread_event = getattr(self, "_primary_thread_event", None)
-            if primary_thread_event is not None:
-                if not primary_thread_event.isSet():
-                    self._primary_thread_task = run_and_release, (), {}
-                    primary_thread_event.set()
-                    return reply
-            self.execmodel.start(run_and_release, ())
-        return reply
-
-    def waitall(self, timeout=None):
-        """ wait until all previosuly spawns have terminated. """
-        # if it exists signal primary_thread to finish its loop
-        self._primary_thread_task = None, None, None
-        try:
-            self._primary_thread_event.set()
-        except AttributeError:
-            pass
-        with self._running_lock:
-            if not self._running:
-                return True
-            # if a Reply still runs, we let run_and_release
-            # signal us -- note that we are still holding the
-            # _running_lock to avoid race conditions
-            self._waitall_event = self.execmodel.Event()
-        return self._waitall_event.wait(timeout=timeout)
-
-
-sysex = (KeyboardInterrupt, SystemExit)
-
-
-DEBUG = os.environ.get('EXECNET_DEBUG')
-pid = os.getpid()
-if DEBUG == '2':
-    def trace(*msg):
-        try:
-            line = " ".join(map(str, msg))
-            sys.stderr.write("[%s] %s\n" % (pid, line))
-            sys.stderr.flush()
-        except Exception:
-            pass # nothing we can do, likely interpreter-shutdown
-elif DEBUG:
-    import tempfile, os.path
-    fn = os.path.join(tempfile.gettempdir(), 'execnet-debug-%d' % pid)
-    #sys.stderr.write("execnet-debug at %r" %(fn,))
-    debugfile = open(fn, 'w')
-    def trace(*msg):
-        try:
-            line = " ".join(map(str, msg))
-            debugfile.write(line + "\n")
-            debugfile.flush()
-        except Exception:
-            try:
-                v = sys.exc_info()[1]
-                sys.stderr.write(
-                    "[%s] exception during tracing: %r\n" % (pid, v))
-            except Exception:
-                pass # nothing we can do, likely interpreter-shutdown
-else:
-    notrace = trace = lambda *msg: None
-
-class Popen2IO:
-    error = (IOError, OSError, EOFError)
-
-    def __init__(self, outfile, infile, execmodel):
-        # we need raw byte streams
-        self.outfile, self.infile = outfile, infile
-        if sys.platform == "win32":
-            import msvcrt
-            try:
-                msvcrt.setmode(infile.fileno(), os.O_BINARY)
-                msvcrt.setmode(outfile.fileno(), os.O_BINARY)
-            except (AttributeError, IOError):
-                pass
-        self._read = getattr(infile, "buffer", infile).read
-        self._write = getattr(outfile, "buffer", outfile).write
-        self.execmodel = execmodel
-
-    def read(self, numbytes):
-        """Read exactly 'numbytes' bytes from the pipe. """
-        # a file in non-blocking mode may return less bytes, so we loop
-        buf = bytes()
-        while numbytes > len(buf):
-            data = self._read(numbytes-len(buf))
-            if not data:
-                raise EOFError("expected %d bytes, got %d" %(numbytes, len(buf)))
-            buf += data
-        return buf
-
-    def write(self, data):
-        """write out all data bytes. """
-        assert isinstance(data, bytes)
-        self._write(data)
-        self.outfile.flush()
-
-    def close_read(self):
-        self.infile.close()
-
-    def close_write(self):
-        self.outfile.close()
-
-class Message:
-    """ encapsulates Messages and their wire protocol. """
-    _types = []
-
-    def __init__(self, msgcode, channelid=0, data=''):
-        self.msgcode = msgcode
-        self.channelid = channelid
-        self.data = data
-
-    @staticmethod
-    def from_io(io):
-        try:
-            header = io.read(9) # type 1, channel 4, payload 4
-            if not header:
-                raise EOFError("empty read")
-        except EOFError:
-            e = sys.exc_info()[1]
-            raise EOFError('couldnt load message header, ' + e.args[0])
-        msgtype, channel, payload = struct.unpack('!bii', header)
-        return Message(msgtype, channel, io.read(payload))
-
-    def to_io(self, io):
-        header = struct.pack('!bii', self.msgcode, self.channelid,
-                             len(self.data))
-        io.write(header+self.data)
-
-    def received(self, gateway):
-        self._types[self.msgcode](self, gateway)
-
-    def __repr__(self):
-        class FakeChannel(object):
-            _strconfig = False, False # never transform, never fail
-            def __init__(self, id):
-                self.id = id
-            def __repr__(self):
-                return '<Channel %s>' % self.id
-        FakeChannel.new = FakeChannel
-        FakeChannel.gateway = FakeChannel
-        name = self._types[self.msgcode].__name__.upper()
-        try:
-            data = loads_internal(self.data, FakeChannel)
-        except LoadError:
-            data = self.data
-        r = repr(data)
-        if len(r) > 90:
-            return "<Message.%s channelid=%d len=%d>" %(name,
-                        self.channelid, len(r))
-        else:
-            return "<Message.%s channelid=%d %s>" %(name,
-                        self.channelid, r)
-
-def _setupmessages():
-    def status(message, gateway):
-        # we use the channelid to send back information
-        # but don't instantiate a channel object
-        d = {'numchannels': len(gateway._channelfactory._channels),
-             'numexecuting': gateway._execpool.active_count(),
-             'execmodel': gateway.execmodel.backend,
-        }
-        gateway._send(Message.CHANNEL_DATA, message.channelid, dumps_internal(d))
-        gateway._send(Message.CHANNEL_CLOSE, message.channelid)
-
-    def channel_exec(message, gateway):
-        channel = gateway._channelfactory.new(message.channelid)
-        gateway._local_schedulexec(channel=channel,sourcetask=message.data)
-
-    def channel_data(message, gateway):
-        gateway._channelfactory._local_receive(message.channelid, message.data)
-
-    def channel_close(message, gateway):
-        gateway._channelfactory._local_close(message.channelid)
-
-    def channel_close_error(message, gateway):
-        remote_error = RemoteError(loads_internal(message.data))
-        gateway._channelfactory._local_close(message.channelid, remote_error)
-
-    def channel_last_message(message, gateway):
-        gateway._channelfactory._local_close(message.channelid, sendonly=True)
-
-    def gateway_terminate(message, gateway):
-        # wake up and terminate any execution waiting to receive something
-        gateway._channelfactory._finished_receiving()
-        # then try harder to terminate execution
-        gateway._terminate_execution()
-
-    def reconfigure(message, gateway):
-        if message.channelid == 0:
-            target = gateway
-        else:
-            target = gateway._channelfactory.new(message.channelid)
-        target._strconfig = loads_internal(message.data, gateway)
-
-    types = [
-        status, reconfigure, gateway_terminate,
-        channel_exec, channel_data, channel_close,
-        channel_close_error, channel_last_message,
-    ]
-    for i, handler in enumerate(types):
-        Message._types.append(handler)
-        setattr(Message, handler.__name__.upper(), i)
-
-_setupmessages()
-
-def geterrortext(excinfo,
-    format_exception=traceback.format_exception, sysex=sysex):
-    try:
-        l = format_exception(*excinfo)
-        errortext = "".join(l)
-    except sysex:
-        raise
-    except:
-        errortext = '%s: %s' % (excinfo[0].__name__,
-                                excinfo[1])
-    return errortext
-
-class RemoteError(Exception):
-    """ Exception containing a stringified error from the other side. """
-    def __init__(self, formatted):
-        self.formatted = formatted
-        Exception.__init__(self)
-
-    def __str__(self):
-        return self.formatted
-
-    def __repr__(self):
-        return "%s: %s" %(self.__class__.__name__, self.formatted)
-
-    def warn(self):
-        if self.formatted != INTERRUPT_TEXT:
-            # XXX do this better
-            sys.stderr.write("[%s] Warning: unhandled %r\n"
-                             % (os.getpid(), self,))
-
-class TimeoutError(IOError):
-    """ Exception indicating that a timeout was reached. """
-
-
-NO_ENDMARKER_WANTED = object()
-
-class Channel(object):
-    """Communication channel between two Python Interpreter execution points."""
-    RemoteError = RemoteError
-    TimeoutError = TimeoutError
-    Message = Message
-    _INTERNALWAKEUP = 1000
-    _executing = False
-
-    def __init__(self, gateway, id):
-        assert isinstance(id, int)
-        self.gateway = gateway
-        #XXX: defaults copied from Unserializer
-        self._strconfig = getattr(gateway, '_strconfig', (True, False))
-        self.id = id
-        self._items = self.gateway.execmodel.queue.Queue()
-        self._closed = False
-        self._receiveclosed = self.gateway.execmodel.Event()
-        self._remoteerrors = []
-
-    def _trace(self, *msg):
-        self.gateway._trace(self.id, *msg)
-
-    def setcallback(self, callback, endmarker=NO_ENDMARKER_WANTED):
-        """ set a callback function for receiving items.
-
-            All already queued items will immediately trigger the callback.
-            Afterwards the callback will execute in the receiver thread
-            for each received data item and calls to ``receive()`` will
-            raise an error.
-            If an endmarker is specified the callback will eventually
-            be called with the endmarker when the channel closes.
-        """
-        _callbacks = self.gateway._channelfactory._callbacks
-        _receivelock = self.gateway._receivelock
-        _receivelock.acquire()
-        try:
-            if self._items is None:
-                raise IOError("%r has callback already registered" %(self,))
-            items = self._items
-            self._items = None
-            while 1:
-                try:
-                    olditem = items.get(block=False)
-                except self.gateway.execmodel.queue.Empty:
-                    if not (self._closed or self._receiveclosed.isSet()):
-                        _callbacks[self.id] = (
-                            callback,
-                            endmarker,
-                            self._strconfig,
-                        )
-                    break
-                else:
-                    if olditem is ENDMARKER:
-                        items.put(olditem) # for other receivers
-                        if endmarker is not NO_ENDMARKER_WANTED:
-                            callback(endmarker)
-                        break
-                    else:
-                        callback(olditem)
-        finally:
-            _receivelock.release()
-
-    def __repr__(self):
-        flag = self.isclosed() and "closed" or "open"
-        return "<Channel id=%d %s>" % (self.id, flag)
-
-    def __del__(self):
-        if self.gateway is None:   # can be None in tests
-            return
-        self._trace("channel.__del__")
-        # no multithreading issues here, because we have the last ref to 'self'
-        if self._closed:
-            # state transition "closed" --> "deleted"
-            for error in self._remoteerrors:
-                error.warn()
-        elif self._receiveclosed.isSet():
-            # state transition "sendonly" --> "deleted"
-            # the remote channel is already in "deleted" state, nothing to do
-            pass
-        else:
-            # state transition "opened" --> "deleted"
-            if self._items is None:    # has_callback
-                msgcode = self.Message.CHANNEL_LAST_MESSAGE
-            else:
-                msgcode = self.Message.CHANNEL_CLOSE
-            try:
-                self.gateway._send(msgcode, self.id)
-            except (IOError, ValueError): # ignore problems with sending
-                pass
-
-    def _getremoteerror(self):
-        try:
-            return self._remoteerrors.pop(0)
-        except IndexError:
-            try:
-                return self.gateway._error
-            except AttributeError:
-                pass
-            return None
-
-    #
-    # public API for channel objects
-    #
-    def isclosed(self):
-        """ return True if the channel is closed. A closed
-            channel may still hold items.
-        """
-        return self._closed
-
-    def makefile(self, mode='w', proxyclose=False):
-        """ return a file-like object.
-            mode can be 'w' or 'r' for writeable/readable files.
-            if proxyclose is true file.close() will also close the channel.
-        """
-        if mode == "w":
-            return ChannelFileWrite(channel=self, proxyclose=proxyclose)
-        elif mode == "r":
-            return ChannelFileRead(channel=self, proxyclose=proxyclose)
-        raise ValueError("mode %r not availabe" %(mode,))
-
-    def close(self, error=None):
-        """ close down this channel with an optional error message.
-            Note that closing of a channel tied to remote_exec happens
-            automatically at the end of execution and cannot
-            be done explicitely.
-        """
-        if self._executing:
-            raise IOError("cannot explicitly close channel within remote_exec")
-        if self._closed:
-            self.gateway._trace(self, "ignoring redundant call to close()")
-        if not self._closed:
-            # state transition "opened/sendonly" --> "closed"
-            # threads warning: the channel might be closed under our feet,
-            # but it's never damaging to send too many CHANNEL_CLOSE messages
-            # however, if the other side triggered a close already, we
-            # do not send back a closed message.
-            if not self._receiveclosed.isSet():
-                put = self.gateway._send
-                if error is not None:
-                    put(self.Message.CHANNEL_CLOSE_ERROR, self.id, dumps_internal(error))
-                else:
-                    put(self.Message.CHANNEL_CLOSE, self.id)
-                self._trace("sent channel close message")
-            if isinstance(error, RemoteError):
-                self._remoteerrors.append(error)
-            self._closed = True         # --> "closed"
-            self._receiveclosed.set()
-            queue = self._items
-            if queue is not None:
-                queue.put(ENDMARKER)
-            self.gateway._channelfactory._no_longer_opened(self.id)
-
-    def waitclose(self, timeout=None):
-        """ wait until this channel is closed (or the remote side
-        otherwise signalled that no more data was being sent).
-        The channel may still hold receiveable items, but not receive
-        any more after waitclose() has returned.  Exceptions from executing
-        code on the other side are reraised as local channel.RemoteErrors.
-        EOFError is raised if the reading-connection was prematurely closed,
-        which often indicates a dying process.
-        self.TimeoutError is raised after the specified number of seconds
-        (default is None, i.e. wait indefinitely).
-        """
-        self._receiveclosed.wait(timeout=timeout) # wait for non-"opened" state
-        if not self._receiveclosed.isSet():
-            raise self.TimeoutError("Timeout after %r seconds" % timeout)
-        error = self._getremoteerror()
-        if error:
-            raise error
-
-    def send(self, item):
-        """sends the given item to the other side of the channel,
-        possibly blocking if the sender queue is full.
-        The item must be a simple python type and will be
-        copied to the other side by value.  IOError is
-        raised if the write pipe was prematurely closed.
-        """
-        if self.isclosed():
-            raise IOError("cannot send to %r" %(self,))
-        self.gateway._send(self.Message.CHANNEL_DATA, self.id, dumps_internal(item))
-
-    def receive(self, timeout=None):
-        """receive a data item that was sent from the other side.
-        timeout: None [default] blocked waiting.  A positive number
-        indicates the number of seconds after which a channel.TimeoutError
-        exception will be raised if no item was received.
-        Note that exceptions from the remotely executing code will be
-        reraised as channel.RemoteError exceptions containing
-        a textual representation of the remote traceback.
-        """
-        itemqueue = self._items
-        if itemqueue is None:
-            raise IOError("cannot receive(), channel has receiver callback")
-        try:
-            x = itemqueue.get(timeout=timeout)
-        except self.gateway.execmodel.queue.Empty:
-            raise self.TimeoutError("no item after %r seconds" %(timeout))
-        if x is ENDMARKER:
-            itemqueue.put(x)  # for other receivers
-            raise self._getremoteerror() or EOFError()
-        else:
-            return x
-
-    def __iter__(self):
-        return self
-
-    def next(self):
-        try:
-            return self.receive()
-        except EOFError:
-            raise StopIteration
-    __next__ = next
-
-
-    def reconfigure(self, py2str_as_py3str=True, py3str_as_py2str=False):
-        """
-        set the string coercion for this channel
-        the default is to try to convert py2 str as py3 str,
-        but not to try and convert py3 str to py2 str
-        """
-        self._strconfig = (py2str_as_py3str, py3str_as_py2str)
-        data = dumps_internal(self._strconfig)
-        self.gateway._send(self.Message.RECONFIGURE, self.id, data=data)
-
-ENDMARKER = object()
-INTERRUPT_TEXT = "keyboard-interrupted"
-
-class ChannelFactory(object):
-    def __init__(self, gateway, startcount=1):
-        self._channels = weakref.WeakValueDictionary()
-        self._callbacks = {}
-        self._writelock = gateway.execmodel.Lock()
-        self.gateway = gateway
-        self.count = startcount
-        self.finished = False
-        self._list = list # needed during interp-shutdown
-
-    def new(self, id=None):
-        """ create a new Channel with 'id' (or create new id if None). """
-        self._writelock.acquire()
-        try:
-            if self.finished:
-                raise IOError("connexion already closed: %s" % (self.gateway,))
-            if id is None:
-                id = self.count
-                self.count += 2
-            try:
-                channel = self._channels[id]
-            except KeyError:
-                channel = self._channels[id] = Channel(self.gateway, id)
-            return channel
-        finally:
-            self._writelock.release()
-
-    def channels(self):
-        return self._list(self._channels.values())
-
-    #
-    # internal methods, called from the receiver thread
-    #
-    def _no_longer_opened(self, id):
-        try:
-            del self._channels[id]
-        except KeyError:
-            pass
-        try:
-            callback, endmarker, strconfig = self._callbacks.pop(id)
-        except KeyError:
-            pass
-        else:
-            if endmarker is not NO_ENDMARKER_WANTED:
-                callback(endmarker)
-
-    def _local_close(self, id, remoteerror=None, sendonly=False):
-        channel = self._channels.get(id)
-        if channel is None:
-            # channel already in "deleted" state
-            if remoteerror:
-                remoteerror.warn()
-            self._no_longer_opened(id)
-        else:
-            # state transition to "closed" state
-            if remoteerror:
-                channel._remoteerrors.append(remoteerror)
-            queue = channel._items
-            if queue is not None:
-                queue.put(ENDMARKER)
-            self._no_longer_opened(id)
-            if not sendonly: # otherwise #--> "sendonly"
-                channel._closed = True          # --> "closed"
-            channel._receiveclosed.set()
-
-    def _local_receive(self, id, data):
-        # executes in receiver thread
-        try:
-            callback, endmarker, strconfig= self._callbacks[id]
-            channel = self._channels.get(id)
-        except KeyError:
-            channel = self._channels.get(id)
-            queue = channel and channel._items
-            if queue is None:
-                pass    # drop data
-            else:
-                item = loads_internal(data, channel)
-                queue.put(item)
-        else:
-            try:
-                data = loads_internal(data, channel, strconfig)
-                callback(data)   # even if channel may be already closed
-            except KeyboardInterrupt:
-                raise
-            except:
-                excinfo = sys.exc_info()
-                self.gateway._trace("exception during callback: %s" % excinfo[1])
-                errortext = self.gateway._geterrortext(excinfo)
-                self.gateway._send(Message.CHANNEL_CLOSE_ERROR, id, dumps_internal(errortext))
-                self._local_close(id, errortext)
-
-    def _finished_receiving(self):
-        self.gateway._trace("finished receiving")
-        self._writelock.acquire()
-        try:
-            self.finished = True
-        finally:
-            self._writelock.release()
-        for id in self._list(self._channels):
-            self._local_close(id, sendonly=True)
-        for id in self._list(self._callbacks):
-            self._no_longer_opened(id)
-
-class ChannelFile(object):
-    def __init__(self, channel, proxyclose=True):
-        self.channel = channel
-        self._proxyclose = proxyclose
-
-    def isatty(self):
-        return False
-
-    def close(self):
-        if self._proxyclose:
-            self.channel.close()
-
-    def __repr__(self):
-        state = self.channel.isclosed() and 'closed' or 'open'
-        return '<ChannelFile %d %s>' %(self.channel.id, state)
-
-class ChannelFileWrite(ChannelFile):
-    def write(self, out):
-        self.channel.send(out)
-
-    def flush(self):
-        pass
-
-class ChannelFileRead(ChannelFile):
-    def __init__(self, channel, proxyclose=True):
-        super(ChannelFileRead, self).__init__(channel, proxyclose)
-        self._buffer = None
-
-    def read(self, n):
-        try:
-            if self._buffer is None:
-                self._buffer = self.channel.receive()
-            while len(self._buffer) < n:
-                self._buffer += self.channel.receive()
-        except EOFError:
-            self.close()
-        if self._buffer is None:
-            ret = ""
-        else:
-            ret = self._buffer[:n]
-            self._buffer = self._buffer[n:]
-        return ret
-
-    def readline(self):
-        if self._buffer is not None:
-            i = self._buffer.find("\n")
-            if i != -1:
-                return self.read(i+1)
-            line = self.read(len(self._buffer)+1)
-        else:
-            line = self.read(1)
-        while line and line[-1] != "\n":
-            c = self.read(1)
-            if not c:
-                break
-            line += c
-        return line
-
-class BaseGateway(object):
-    exc_info = sys.exc_info
-    _sysex = sysex
-    id = "<slave>"
-
-    def __init__(self, io, id, _startcount=2):
-        self.execmodel = io.execmodel
-        self._io = io
-        self.id = id
-        self._strconfig = (Unserializer.py2str_as_py3str,
-                           Unserializer.py3str_as_py2str)
-        self._channelfactory = ChannelFactory(self, _startcount)
-        self._receivelock = self.execmodel.RLock()
-        # globals may be NONE at process-termination
-        self.__trace = trace
-        self._geterrortext = geterrortext
-        self._workerpool = self.execmodel.WorkerPool(1)
-
-    def _trace(self, *msg):
-        self.__trace(self.id, *msg)
-
-    def _initreceive(self):
-        self._receiverthread = self._workerpool.spawn(self._thread_receiver)
-
-    def _thread_receiver(self):
-        def log(*msg):
-            self._trace("[receiver-thread]", *msg)
-        log("RECEIVERTHREAD: starting to run")
-        eof = False
-        io = self._io
-        try:
-            try:
-                while 1:
-                    msg = Message.from_io(io)
-                    log("received", msg)
-                    _receivelock = self._receivelock
-                    _receivelock.acquire()
-                    try:
-                        msg.received(self)
-                        del msg
-                    finally:
-                        _receivelock.release()
-            except self._sysex:
-                log("io.close_read()")
-                self._io.close_read()
-            except EOFError:
-                log("got EOF")
-                #log("traceback was: ", self._geterrortext(self.exc_info()))
-                self._error = self.exc_info()[1]
-                eof = True
-            except:
-                log(self._geterrortext(self.exc_info()))
-        finally:
-            try:
-                log('entering finalization')
-                # wake up any execution waiting to receive something
-                self._channelfactory._finished_receiving()
-                if eof:
-                    self._terminate_execution()
-                log('leaving finalization')
-            except:
-                pass # XXX be silent at interp-shutdown
-
-    def _terminate_execution(self):
-        pass
-
-    def _send(self, msgcode, channelid=0, data=bytes()):
-        message = Message(msgcode, channelid, data)
-        try:
-            message.to_io(self._io)
-            self._trace('sent', message)
-        except (IOError, ValueError):
-            e = sys.exc_info()[1]
-            self._trace('failed to send', message, e)
-            raise
-
-
-    def _local_schedulexec(self, channel, sourcetask):
-        channel.close("execution disallowed")
-
-    # _____________________________________________________________________
-    #
-    # High Level Interface
-    # _____________________________________________________________________
-    #
-    def newchannel(self):
-        """ return a new independent channel. """
-        return self._channelfactory.new()
-
-    def join(self, timeout=None):
-        """ Wait for receiverthread to terminate. """
-        self._trace("waiting for receiver thread to finish")
-        self._receiverthread.waitfinish()
-
-class SlaveGateway(BaseGateway):
-
-    def _local_schedulexec(self, channel, sourcetask):
-        sourcetask = loads_internal(sourcetask)
-        self._execpool.spawn(self.executetask, ((channel, sourcetask)))
-
-    def _terminate_execution(self):
-        # called from receiverthread
-        self._trace("shutting down execution pool")
-        self._execpool.shutdown()
-        if not self._execpool.waitall(1.0):
-            self._trace("execution ongoing, trying interrupt_main")
-            # We try hard to terminate execution based on the assumption
-            # that there is only one gateway object running per-process.
-            if sys.platform != "win32":
-                self._trace("sending ourselves a SIGINT")
-                os.kill(os.getpid(), 2)  # send ourselves a SIGINT
-            elif interrupt_main is not None:
-                self._trace("calling interrupt_main()")
-                interrupt_main()
-            if not self._execpool.waitall(10.0):
-                self._trace("execution did not finish in 10 secs, "
-                            "calling os._exit()")
-                os._exit(1)
-
-    def serve(self):
-        trace = lambda msg: self._trace("[serve] " + msg)
-        hasprimary = self.execmodel.backend == "thread"
-        self._execpool = self.execmodel.WorkerPool(hasprimary=hasprimary)
-        trace("spawning receiver thread")
-        self._initreceive()
-        try:
-            try:
-                if hasprimary:
-                    trace("integrating as main primary exec thread")
-                    self._execpool.integrate_as_primary_thread()
-                else:
-                    trace("waiting for execution to finish")
-                    self._execpool.wait_for_shutdown()
-            finally:
-                trace("execution finished, closing io write stream")
-                self._io.close_write()
-            trace("joining receiver thread")
-            self.join()
-        except KeyboardInterrupt:
-            # in the slave we can't really do anything sensible
-            trace("swallowing keyboardinterrupt, serve finished")
-
-    def executetask(self, item):
-        try:
-            channel, (source, call_name, kwargs) = item
-            if not ISPY3 and kwargs:
-                # some python2 versions do not accept unicode keyword params
-                # note: Unserializer generally turns py2-str to py3-str objects
-                newkwargs = {}
-                for name, value in kwargs.items():
-                    if isinstance(name, unicode):
-                        name = name.encode('ascii')
-                    newkwargs[name] = value
-                kwargs = newkwargs
-            loc = {'channel' : channel, '__name__': '__channelexec__'}
-            self._trace("execution starts[%s]: %s" %
-                            (channel.id, repr(source)[:50]))
-            channel._executing = True
-            try:
-                co = compile(source+'\n', '<remote exec>', 'exec')
-                do_exec(co, loc) # noqa
-                if call_name:
-                    self._trace('calling %s(**%60r)' % (call_name, kwargs))
-                    function = loc[call_name]
-                    function(channel, **kwargs)
-            finally:
-                channel._executing = False
-                self._trace("execution finished")
-        except KeyboardInterrupt:
-            channel.close(INTERRUPT_TEXT)
-            raise
-        except:
-            excinfo = self.exc_info()
-            if not isinstance(excinfo[1], EOFError):
-                if not channel.gateway._channelfactory.finished:
-                    self._trace("got exception: %r" % (excinfo[1],))
-                    errortext = self._geterrortext(excinfo)
-                    channel.close(errortext)
-                    return
-            self._trace("ignoring EOFError because receiving finished")
-        channel.close()
-
-#
-# Cross-Python pickling code, tested from test_serializer.py
-#
-
-class DataFormatError(Exception):
-    pass
-
-class DumpError(DataFormatError):
-    """Error while serializing an object."""
-
-class LoadError(DataFormatError):
-    """Error while unserializing an object."""
-
-if ISPY3:
-    def bchr(n):
-        return bytes([n])
-else:
-    bchr = chr
-
-DUMPFORMAT_VERSION = bchr(1)
-
-FOUR_BYTE_INT_MAX = 2147483647
-
-FLOAT_FORMAT = "!d"
-FLOAT_FORMAT_SIZE = struct.calcsize(FLOAT_FORMAT)
-
-class _Stop(Exception):
-    pass
-
-class Unserializer(object):
-    num2func = {} # is filled after this class definition
-    py2str_as_py3str = True # True
-    py3str_as_py2str = False  # false means py2 will get unicode
-
-    def __init__(self, stream, channel_or_gateway=None, strconfig=None):
-        gateway = getattr(channel_or_gateway, 'gateway', channel_or_gateway)
-        strconfig = getattr(channel_or_gateway, '_strconfig', strconfig)
-        if strconfig:
-            self.py2str_as_py3str, self.py3str_as_py2str = strconfig
-        self.stream = stream
-        self.channelfactory = getattr(gateway, '_channelfactory', gateway)
-
-    def load(self, versioned=False):
-        if versioned:
-            ver = self.stream.read(1)
-            if ver != DUMPFORMAT_VERSION:
-                raise LoadError("wrong dumpformat version")
-        self.stack = []
-        try:
-            while True:
-                opcode = self.stream.read(1)
-                if not opcode:
-                    raise EOFError
-                try:
-                    loader = self.num2func[opcode]
-                except KeyError:
-                    raise LoadError("unkown opcode %r - "
-                        "wire protocol corruption?" % (opcode,))
-                loader(self)
-        except _Stop:
-            if len(self.stack) != 1:
-                raise LoadError("internal unserialization error")
-            return self.stack.pop(0)
-        else:
-            raise LoadError("didn't get STOP")
-
-    def load_none(self):
-        self.stack.append(None)
-
-    def load_true(self):
-        self.stack.append(True)
-
-    def load_false(self):
-        self.stack.append(False)
-
-    def load_int(self):
-        i = self._read_int4()
-        self.stack.append(i)
-
-    def load_longint(self):
-        s = self._read_byte_string()
-        self.stack.append(int(s))
-
-    if ISPY3:
-        load_long = load_int
-        load_longlong = load_longint
-    else:
-        def load_long(self):
-            i = self._read_int4()
-            self.stack.append(long(i))
-
-        def load_longlong(self):
-            l = self._read_byte_string()
-            self.stack.append(long(l))
-
-    def load_float(self):
-        binary = self.stream.read(FLOAT_FORMAT_SIZE)
-        self.stack.append(struct.unpack(FLOAT_FORMAT, binary)[0])
-
-    def _read_int4(self):
-        return struct.unpack("!i", self.stream.read(4))[0]
-
-    def _read_byte_string(self):
-        length = self._read_int4()
-        as_bytes = self.stream.read(length)
-        return as_bytes
-
-    def load_py3string(self):
-        as_bytes = self._read_byte_string()
-        if not ISPY3 and self.py3str_as_py2str:
-            # XXX Should we try to decode into latin-1?
-            self.stack.append(as_bytes)
-        else:
-            self.stack.append(as_bytes.decode("utf-8"))
-
-    def load_py2string(self):
-        as_bytes = self._read_byte_string()
-        if ISPY3 and self.py2str_as_py3str:
-            s = as_bytes.decode("latin-1")
-        else:
-            s = as_bytes
-        self.stack.append(s)
-
-    def load_bytes(self):
-        s = self._read_byte_string()
-        self.stack.append(s)
-
-    def load_unicode(self):
-        self.stack.append(self._read_byte_string().decode("utf-8"))
-
-    def load_newlist(self):
-        length = self._read_int4()
-        self.stack.append([None] * length)
-
-    def load_setitem(self):
-        if len(self.stack) < 3:
-            raise LoadError("not enough items for setitem")
-        value = self.stack.pop()
-        key = self.stack.pop()
-        self.stack[-1][key] = value
-
-    def load_newdict(self):
-        self.stack.append({})
-
-    def _load_collection(self, type_):
-        length = self._read_int4()
-        if length:
-            res = type_(self.stack[-length:])
-            del self.stack[-length:]
-            self.stack.append(res)
-        else:
-            self.stack.append(type_())
-
-    def load_buildtuple(self):
-        self._load_collection(tuple)
-
-    def load_set(self):
-        self._load_collection(set)
-
-    def load_frozenset(self):
-        self._load_collection(frozenset)
-
-    def load_stop(self):
-        raise _Stop
-
-    def load_channel(self):
-        id = self._read_int4()
-        newchannel = self.channelfactory.new(id)
-        self.stack.append(newchannel)
-
-# automatically build opcodes and byte-encoding
-
-class opcode:
-    """ container for name -> num mappings. """
-
-def _buildopcodes():
-    l = []
-    for name, func in Unserializer.__dict__.items():
-        if name.startswith("load_"):
-            opname = name[5:].upper()
-            l.append((opname, func))
-    l.sort()
-    for i,(opname, func) in enumerate(l):
-        assert i < 26, "xxx"
-        i = bchr(64+i)
-        Unserializer.num2func[i] = func
-        setattr(opcode, opname, i)
-
-_buildopcodes()
-
-def dumps(obj):
-    """ return a serialized bytestring of the given obj.
-
-    The obj and all contained objects must be of a builtin
-    python type (so nested dicts, sets, etc. are all ok but
-    not user-level instances).
-    """
-    return _Serializer().save(obj, versioned=True)
-
-def dump(byteio, obj):
-    """ write a serialized bytestring of the given obj to the given stream. """
-    _Serializer(write=byteio.write).save(obj, versioned=True)
-
-def loads(bytestring, py2str_as_py3str=False, py3str_as_py2str=False):
-    """ return the object as deserialized from the given bytestring.
-
-    py2str_as_py3str: if true then string (str) objects previously
-                      dumped on Python2 will be loaded as Python3
-                      strings which really are text objects.
-    py3str_as_py2str: if true then string (str) objects previously
-                      dumped on Python3 will be loaded as Python2
-                      strings instead of unicode objects.
-
-    if the bytestring was dumped with an incompatible protocol
-    version or if the bytestring is corrupted, the
-    ``execnet.DataFormatError`` will be raised.
-    """
-    io = BytesIO(bytestring)
-    return load(io, py2str_as_py3str=py2str_as_py3str,
-                    py3str_as_py2str=py3str_as_py2str)
-
-def load(io, py2str_as_py3str=False, py3str_as_py2str=False):
-    """ derserialize an object form the specified stream.
-
-    Behaviour and parameters are otherwise the same as with ``loads``
-    """
-    strconfig=(py2str_as_py3str, py3str_as_py2str)
-    return Unserializer(io, strconfig=strconfig).load(versioned=True)
-
-def loads_internal(bytestring, channelfactory=None, strconfig=None):
-    io = BytesIO(bytestring)
-    return Unserializer(io, channelfactory, strconfig).load()
-
-def dumps_internal(obj):
-    return _Serializer().save(obj)
-
-
-class _Serializer(object):
-    _dispatch = {}
-
-    def __init__(self, write=None):
-        if write is None:
-            self._streamlist = []
-            write = self._streamlist.append
-        self._write = write
-
-    def save(self, obj, versioned=False):
-        # calling here is not re-entrant but multiple instances
-        # may write to the same stream because of the common platform
-        # atomic-write guaruantee (concurrent writes each happen atomicly)
-        if versioned:
-            self._write(DUMPFORMAT_VERSION)
-        self._save(obj)
-        self._write(opcode.STOP)
-        try:
-            streamlist = self._streamlist
-        except AttributeError:
-            return None
-        return type(streamlist[0])().join(streamlist)
-
-    def _save(self, obj):
-        tp = type(obj)
-        try:
-            dispatch = self._dispatch[tp]
-        except KeyError:
-            methodname = 'save_' + tp.__name__
-            meth = getattr(self.__class__, methodname, None)
-            if meth is None:
-                raise DumpError("can't serialize %s" % (tp,))
-            dispatch = self._dispatch[tp] = meth
-        dispatch(self, obj)
-
-    def save_NoneType(self, non):
-        self._write(opcode.NONE)
-
-    def save_bool(self, boolean):
-        if boolean:
-            self._write(opcode.TRUE)
-        else:
-            self._write(opcode.FALSE)
-
-    def save_bytes(self, bytes_):
-        self._write(opcode.BYTES)
-        self._write_byte_sequence(bytes_)
-
-    if ISPY3:
-        def save_str(self, s):
-            self._write(opcode.PY3STRING)
-            self._write_unicode_string(s)
-    else:
-        def save_str(self, s):
-            self._write(opcode.PY2STRING)
-            self._write_byte_sequence(s)
-
-        def save_unicode(self, s):
-            self._write(opcode.UNICODE)
-            self._write_unicode_string(s)
-
-    def _write_unicode_string(self, s):
-        try:
-            as_bytes = s.encode("utf-8")
-        except UnicodeEncodeError:
-            raise DumpError("strings must be utf-8 encodable")
-        self._write_byte_sequence(as_bytes)
-
-    def _write_byte_sequence(self, bytes_):
-        self._write_int4(len(bytes_), "string is too long")
-        self._write(bytes_)
-
-    def _save_integral(self, i, short_op, long_op):
-        if i <= FOUR_BYTE_INT_MAX:
-            self._write(short_op)
-            self._write_int4(i)
-        else:
-            self._write(long_op)
-            self._write_byte_sequence(str(i).rstrip("L").encode("ascii"))
-
-    def save_int(self, i):
-        self._save_integral(i, opcode.INT, opcode.LONGINT)
-
-    def save_long(self, l):
-        self._save_integral(l, opcode.LONG, opcode.LONGLONG)
-
-    def save_float(self, flt):
-        self._write(opcode.FLOAT)
-        self._write(struct.pack(FLOAT_FORMAT, flt))
-
-    def _write_int4(self, i, error="int must be less than %i" %
-                    (FOUR_BYTE_INT_MAX,)):
-        if i > FOUR_BYTE_INT_MAX:
-            raise DumpError(error)
-        self._write(struct.pack("!i", i))
-
-    def save_list(self, L):
-        self._write(opcode.NEWLIST)
-        self._write_int4(len(L), "list is too long")
-        for i, item in enumerate(L):
-            self._write_setitem(i, item)
-
-    def _write_setitem(self, key, value):
-        self._save(key)
-        self._save(value)
-        self._write(opcode.SETITEM)
-
-    def save_dict(self, d):
-        self._write(opcode.NEWDICT)
-        for key, value in d.items():
-            self._write_setitem(key, value)
-
-    def save_tuple(self, tup):
-        for item in tup:
-            self._save(item)
-        self._write(opcode.BUILDTUPLE)
-        self._write_int4(len(tup), "tuple is too long")
-
-    def _write_set(self, s, op):
-        for item in s:
-            self._save(item)
-        self._write(op)
-        self._write_int4(len(s), "set is too long")
-
-    def save_set(self, s):
-        self._write_set(s, opcode.SET)
-
-    def save_frozenset(self, s):
-        self._write_set(s, opcode.FROZENSET)
-
-    def save_Channel(self, channel):
-        self._write(opcode.CHANNEL)
-        self._write_int4(channel.id)
-
-def init_popen_io(execmodel):
-    if not hasattr(os, 'dup'): # jython
-        io = Popen2IO(sys.stdout, sys.stdin, execmodel)
-        import tempfile
-        sys.stdin = tempfile.TemporaryFile('r')
-        sys.stdout = tempfile.TemporaryFile('w')
-    else:
-        try:
-            devnull = os.devnull
-        except AttributeError:
-            if os.name == 'nt':
-                devnull = 'NUL'
-            else:
-                devnull = '/dev/null'
-        # stdin
-        stdin  = execmodel.fdopen(os.dup(0), 'r', 1)
-        fd = os.open(devnull, os.O_RDONLY)
-        os.dup2(fd, 0)
-        os.close(fd)
-
-        # stdout
-        stdout = execmodel.fdopen(os.dup(1), 'w', 1)
-        fd = os.open(devnull, os.O_WRONLY)
-        os.dup2(fd, 1)
-
-        # stderr for win32
-        if os.name == 'nt':
-            sys.stderr = execmodel.fdopen(os.dup(2), 'w', 1)
-            os.dup2(fd, 2)
-        os.close(fd)
-        io = Popen2IO(stdout, stdin, execmodel)
-        sys.stdin = execmodel.fdopen(0, 'r', 1)
-        sys.stdout = execmodel.fdopen(1, 'w', 1)
-    return io
-
-def serve(io, id):
-    trace("creating slavegateway on %r" %(io,))
-    SlaveGateway(io=io, id=id, _startcount=2).serve()