Your IP : 172.28.240.42
"""Manage outgoing and incoming messages when communicating with the server.
The protocol to communicate between the client and the server has been designed
to be very robust so that messages are not lost. In addition it is (vaguely)
symmetric, as the client and server need to send messages both ways.
Client->Server Payload
======================
All message payloads are bpickled with L{landscape.lib.bpickle.dumps}. Client
to server payloads are C{dict}s of the form::
{'server-api': SERVER_API_VERSION,
'client-api': CLIENT_API_VERSION,
'sequence': SEQUENCE_NUMBER,
'accepted-types': SERVER_ACCEPTED_TYPES_DIGEST,
'messages': MESSAGES,
'total-messages': TOTAL_COUNT_OF_PENDING_MESSAGES,
'next-expected-sequence': EXPECTED_SEQUENCE_NUMBER,
'client-accepted-types': CLIENT_ACCEPTED_TYPES (optional)}
The values have the following semantics:
- C{SERVER_API_VERSION}: The API version that is required on the server
in order to process the messages in this payload (the schema and semantics
of message types are usually different for different API versions).
- C{CLIENT_API_VERSION}: The API version of the client, hinting the server
about the schema and semantics of the messages types accepted by the client
(see below).
- C{SEQUENCE_NUMBER}: A monotonically increasing nonnegative integer. The
meaning of this is described below.
- C{SERVER_ACCEPTED_TYPES_DIGEST}: A hash of the message types that the
client thinks are currently accepted by the server. The server can use it
to know whether to send the client a new up-to-date list of accepted
message types.
- C{MESSAGES}: A python list of messages, described below.
- C{TOTAL_COUNT_OF_PENDING_MESSAGES}: The total number of messages in the
client outgoing queue. This is includes the number of messages being sent
in this payload, plus any other messages still pending and not included
here.
- C{EXPECTED_SEQUENCE_NUMBER}: The sequence number which the client expects
the next message sent from the server to have.
- C{CLIENT_ACCEPTED_TYPES}: Optionally, a list of message types that the
client accepts. The server is supposed to send the client only messages of
this type. It will be inlcuded in the payload only if the hash that the
server sends us is out-of-date. This behavior is simmetric with respect to
the C{SERVER_ACCEPTED_TYPES_DIGEST} field described above.
Server->Client Payload
======================
The payloads that the server sends to not-yet-registered clients (i.e. clients
that don't provide a secure ID associated with a computer) are C{dict}s of the
form::
{'server-uuid': SERVER_UUID,
'server-api': SERVER_API,
'messages': MESSAGES}
where:
- C{SERVER_UUID}: A string identifying the particular Landscape server the
client is talking to.
- C{SERVER_API}: The version number of the highest server API that this
particular server is able to handle. It can be used by the client to
implement backward compatibility with old servers, knowing what message
schemas the server expects (since schemas can change from version to
version).
- C{MESSAGES}: A python list of messages, described below.
Additionally, payloads to registered clients will include these fields::
{'next-expected-sequence': EXPECTED_SEQUENCE_NUMBER,
'next-expected-token': EXPECTED_EXCHANGE_TOKEN,
'client-accepted-types-hash': CLIENT_ACCEPTED_TYPES_DIGEST,
where:
- C{EXPECTED_SEQUENCE_NUMBER}: The sequence number which the server expects
the next message sent from the client to have.
- C{EXPECTED_EXCHANGE_TOKEN}: The token (UUID string) that the server expects
to receive back the next time the client performs an exchange. Since the
client receives a new token at each exchange, this can be used by the
server to detect cloned clients (either the orignal client or the cloned
client will eventually send an expired token). The token is sent by the
client as a special HTTP header (see L{landscape.broker.transport}).
- C{CLIENT_ACCEPTED_TYPES_DIGEST}: A hash of the message types that the
server thinks are currently accepted by the client. The client can use it
to know whether to send to the server an up-to-date list the message types
it now accepts (see CLIENT_ACCEPTED_TYPES in the client->server payload).
Individual Messages
===================
A message is a C{dict} with required and optional keys. Messages are packed
into Python lists and set as the value of the 'messages' key in the payload.
The C{dict} of a single message is of the form::
{'type': MESSAGE_TYPE,
...}
where:
- C{MESSAGE_TYPE}: A simple string, which lets the server decide what handler
to dispatch the message to, also considering the SERVER_API_VERSION value.
- C{...}: Other entries specific to the type of message.
This format is the same for messages sent by the server to the client and for
messages sent by the client to the server. In addition, messages sent by the
client to the server will contain also the following extra fields::
{...
'api': SERVER_API,
'timestamp': TIMESTAMP,
...}
where:
- C{SERVER_API}: The server API that the client was targeting when it
generated the message. In single exchange the client will only include
messages targeted to the same server API.
- C{TIMESTAMP}: A timestamp indicating when the message was generated.
Message Sequencing
==================
A message numbering system is built in to the protocol to ensure robustness of
client/server communication. The way this works is not totally symmetrical, as
the client must connect to the server via HTTP, but the ordering that things
happen in over the course of many connections remains the same (see also
L{landscape.broker.store} for more concrete examples):
- Receiver tells Sender which sequence number it expects the next batch of
messages to start with.
- Sender gives some messages to Receiver, specifying the sequence number of
the first message. If the expected and actual sequence numbers are out of
synch, Sender resynchronizes in a certain way.
The client and server must play the part of *both* of these roles on every
interaction, but it simplifies things to talk about them in terms of a single
role at a time.
When the client connects to the server, it does the following things acting
in the role of Sender (which is by far its more burdened role):
- Send a payload containing messages and a sequence number. The sequence
number should be the same number that the server gave as
next-expected-sequence in the prior connection, or 0 if there was no
previous connection.
- Get back a next-expected-sequence from the server. If that value is is not
len(messages) + previous-next-expected, then resynchronize.
It does the following when acting as Receiver:
- Send a payload containing a next-expected-sequence, which should be the
sequence number of the first message that the server responds with. This
value should be previous-next-expected + len(previous_messages).
- Receive some messages from the server, and process them immediately.
When the server is acting as Sender, it does the following:
- Wait for a payload with next-expected-sequence from the client.
- Perhaps resynchronize if next-expected-sequence is unexpected.
- Respond with a payload of messages to the client. No sequence identifier
is given for this payload of messages, because it would be redundant with
data that has already passed over the wire (received from the client)
during the very same TCP connection.
When the server is acting as a Receiver, it does the following:
- Wait for a payload with a sequence identifier and a load of messages.
- Respond with a next-expected-sequence.
There are two interesting exceptional cases which must be handled with
resynchronization:
1. Messages received with sequence numbers less than the next expected
sequence number should be discarded, and further messages starting at
the expected sequence numbers should be processed.
2. If the sequence number is higher than what the receiver expected, then
no messages are processed and the receiver responds with the same
{'next-expected-sequence': N}, so that the sender can resynchronize
itself.
This implies that the receiver must record the sequence number of the last
successfully processed message, in order for it to respond to the sender
with that number. In addition, the sender must save outbound messages even
after they have been delivered over the transport, until the sender receives
a next-expected-sequence higher than the outbound message. The details of
this logic are described in L{landscape.broker.store}.
Exchange Sequence
=================
Diagram::
1. BrokerService --> MessageExchange : Start
2. MessageExchange --> MessageExchange : Schedule exchange
3. [event] <-- MessageExchange : Fire "pre-exchange"
4. [optional] : Do registration
(See L{landscape.broker.registration}) : sequence
5. MessageExchange --> MessageStore : Request pending
: messages
6. MessageExchange <-- MessageStore : return( Messages )
7. MessageExchange --> HTTPTransport : Exchange
8. HTTPTransport --> {Server}LandscapeMessageSystem
: HTTP POST
9. [Scope: Server]
|
| 9.1 LandscapeMessageSystem --> ComputerMessageAPI : run
|
| 9.2 ComputerMessageAPI --> FunctionHandler : handle
|
| 9.3 FunctionHandler --> Callable : call
| ( See also server code at:
| - C{canonical.landscape.message.handlers}
| - C{canonical.message.handler.FunctionHandler} )
|
|
| 9.4 [If: the callable raises ConsistencyError]
| |
| | 9.4.1 ComputerMessageAPI --> Computer : request
| | : Resynchronize
| |
| | 9.4.2 Computer --> Computer : Create
| | : ResynchronizeRequest
| | : activity
| |
| --[End If]
|
| 9.5 ComputerMessageAPI --> Computer : get deliverable
| : activities
|
| 9.6 ComputerMessageAPI <-- Computer : return activities
|
| 9.7 [Loop over activities]
| |
| | 9.7.1 ComputerMessageAPI --> Activity : deliver
| |
| | 9.7.2 Activity --> MessageStore : add activity message
| |
| --[End Loop]
|
| 9.8 ComputerMessageAPI --> MessageStore : get pending messages
|
| 9.9 ComputerMessageAPI <-- MessageStore : return messages
|
| 9.10 LandscapeMessageSystem <-- ComputerMessageAPI : return payload
| : (See below)
|
-- [End Scope]
10. HTTPTransport <-- {Server}LandscapeMessageSystem
: HTTP response
: with payload
11. MessageExchange <-- HTTPTransport : response
12. [If: server says it expects a very old message]
|
| 12.1 [event] <-- MessageExchange : event
| (See L{landscape.broker.server}) : "resynchronize-
| : clients"
|
-- [End if]
13. [Loop: over messages in payload]
|
| 13.1 [event] <-- MessageExchange : event
| : message (message)
|
| 13.2 [Switch: on message type]
| |
| |- 13.2.1 [Case: message type is "accepted-types"]
| | |
| | | 13.2.1.1 MessageExchange -> MessageStore
| | | : set accepted types
| | |
| | | 13.2.1.2 MessageExchange -> MessageExchange
| | | : schedule urgent
| | | : exchange
| | --[End Case]
| |
| |- 13.2.2 [Case: message type is "resynchronize"]
| | |
| | | 13.2.2.1 [event] <- MessageExchange
| | | (See L{landscape.broker.server})
| | | : event
| | | : "resynchronize-
| | | : clients"
| | |
| | | 13.2.2.2 MessageExchange -> MessageStore
| | | : add "resynchronize"
| | | : message
| | |
| | | 13.2.2.3 MessageExchange -> MessageExchange
| | | : schedule urgent
| | | : exchange
| | |
| | --[End Case]
| |
| |- 13.2.3 [Case: message type is "set-intervals"]
| | |
| | | 13.2.3.1 MessageExchange -> BrokerConfiguration
| | | : set exchange
| | | : interval
| | |
| | --[End Case]
| |
| -- [End Switch]
|
-- [End Loop]
14. Schedule exchange
"""
import time
import logging
from landscape.lib.hashlib import md5
from twisted.internet.defer import Deferred, succeed
from landscape.lib.fetch import HTTPCodeError
from landscape.lib.message import got_next_expected, ANCIENT
from landscape.lib.versioning import is_version_higher, sort_versions
from landscape.log import format_delta
from landscape import DEFAULT_SERVER_API, SERVER_API, CLIENT_API
class MessageExchange(object):
"""Schedule and handle message exchanges with the server.
The L{MessageExchange} is the place where messages are sent to go out
to the Landscape server. It accumulates messages in its L{MessageStore}
and periodically delivers them to the server.
It is also the place where messages coming from the server are handled. For
each message type the L{MessageExchange} supports setting an handler that
will be invoked when a message of the that type is received.
An exchange is performed with an HTTP POST request, whose body contains
outgoing messages and whose response contains incoming messages.
"""
# The highest server API that we are capable of speaking
_api = SERVER_API
def __init__(self, reactor, store, transport, registration_info,
exchange_store, config, max_messages=100):
"""
@param reactor: The L{LandscapeReactor} used to fire events in response
to messages received by the server.
@param store: The L{MessageStore} used to queue outgoing messages.
@param transport: The L{HTTPTransport} used to deliver messages.
@param registration_info: The L{Identity} storing our secure ID.
@param config: The L{BrokerConfiguration} with the `exchange_interval`
and `urgent_exchange_interval` parameters, respectively holding
the time interval between subsequent exchanges of non-urgent
messages, and the time interval between subsequent exchanges
of urgent messages.
"""
self._reactor = reactor
self._message_store = store
self._transport = transport
self._registration_info = registration_info
self._config = config
self._exchange_interval = config.exchange_interval
self._urgent_exchange_interval = config.urgent_exchange_interval
self._max_messages = max_messages
self._notification_id = None
self._exchange_id = None
self._exchanging = False
self._urgent_exchange = False
self._client_accepted_types = set()
self._client_accepted_types_hash = None
self._message_handlers = {}
self._exchange_store = exchange_store
self._stopped = False
self.register_message("accepted-types", self._handle_accepted_types)
self.register_message("resynchronize", self._handle_resynchronize)
self.register_message("set-intervals", self._handle_set_intervals)
reactor.call_on("resynchronize-clients", self._resynchronize)
def _message_is_obsolete(self, message):
"""Returns C{True} if message is obsolete.
A message is considered obsolete if the secure ID changed since it was
received.
"""
if 'operation-id' not in message:
return False
operation_id = message['operation-id']
context = self._exchange_store.get_message_context(operation_id)
if context is None:
logging.warning(
"No message context for message with operation-id: %s"
% operation_id)
return False
# Compare the current secure ID with the one that was in effect when
# the request message was received.
result = self._registration_info.secure_id != context.secure_id
context.remove()
return result
def send(self, message, urgent=False):
"""Include a message to be sent in an exchange.
If urgent is True, an exchange with the server will be
scheduled urgently.
@param message: Same as in L{MessageStore.add}.
"""
if self._message_is_obsolete(message):
logging.info(
"Response message with operation-id %s was discarded "
"because the client's secure ID has changed in the meantime"
% message.get('operation-id'))
return None
if "timestamp" not in message:
message["timestamp"] = int(self._reactor.time())
message_id = self._message_store.add(message)
if urgent:
self.schedule_exchange(urgent=True)
return message_id
def start(self):
"""Start scheduling exchanges. The first one will be urgent."""
self.schedule_exchange(urgent=True)
def stop(self):
"""Stop scheduling exchanges."""
if self._exchange_id is not None:
# Cancel the next scheduled exchange
self._reactor.cancel_call(self._exchange_id)
self._exchange_id = None
if self._notification_id is not None:
# Cancel the next scheduled notification of an impending exchange
self._reactor.cancel_call(self._notification_id)
self._notification_id = None
self._stopped = True
def _handle_accepted_types(self, message):
"""
When the server updates us about the types of message it
accepts, update our message store.
If this makes existing held messages available for sending,
urgently exchange messages.
If new types are made available or old types are dropped a
C{("message-type-acceptance-changed", type, bool)} reactor
event will be fired.
"""
old_types = set(self._message_store.get_accepted_types())
new_types = set(message["types"])
diff = get_accepted_types_diff(old_types, new_types)
self._message_store.set_accepted_types(new_types)
logging.info("Accepted types changed: %s", diff)
if self._message_store.get_pending_messages(1):
self.schedule_exchange(urgent=True)
for type in old_types - new_types:
self._reactor.fire("message-type-acceptance-changed", type, False)
for type in new_types - old_types:
self._reactor.fire("message-type-acceptance-changed", type, True)
def _handle_resynchronize(self, message):
opid = message["operation-id"]
scopes = message.get("scopes")
self.send({"type": "resynchronize", "operation-id": opid})
self._reactor.fire("resynchronize-clients", scopes=scopes)
def _resynchronize(self, scopes=None):
# When re-synchronisation occurs we don't want any previous messages
# being sent to the server, dropping the existing session_ids means
# that messages sent with those IDs will be dropped by the broker.
self._message_store.drop_session_ids(scopes)
self.schedule_exchange(urgent=True)
def _handle_set_intervals(self, message):
if "exchange" in message:
self._config.exchange_interval = message["exchange"]
logging.info("Exchange interval set to %d seconds." %
self._config.exchange_interval)
if "urgent-exchange" in message:
self._config.urgent_exchange_interval = message["urgent-exchange"]
logging.info("Urgent exchange interval set to %d seconds." %
self._config.urgent_exchange_interval)
self._config.write()
def exchange(self):
"""Send pending messages to the server and process responses.
A C{pre-exchange} reactor event will be emitted just before the
actual exchange takes place.
An C{exchange-done} or C{exchange-failed} reactor event will be
emitted after a successful or failed exchange.
@return: A L{Deferred} that is fired when exchange has completed.
"""
if self._exchanging:
return succeed(None)
self._exchanging = True
self._reactor.fire("pre-exchange")
payload = self._make_payload()
start_time = time.time()
if self._urgent_exchange:
logging.info("Starting urgent message exchange with %s."
% self._transport.get_url())
else:
logging.info("Starting message exchange with %s."
% self._transport.get_url())
deferred = Deferred()
def exchange_completed():
self.schedule_exchange(force=True)
self._reactor.fire("exchange-done")
logging.info("Message exchange completed in %s.",
format_delta(time.time() - start_time))
deferred.callback(None)
def handle_result(result):
self._exchanging = False
if result:
if self._urgent_exchange:
logging.info("Switching to normal exchange mode.")
self._urgent_exchange = False
self._handle_result(payload, result)
self._message_store.record_success(int(self._reactor.time()))
else:
self._reactor.fire("exchange-failed")
logging.info("Message exchange failed.")
exchange_completed()
def handle_failure(error_class, error, traceback):
self._exchanging = False
if isinstance(error, HTTPCodeError) and error.http_code == 404:
# If we got a 404 HTTP error it could be that we're trying to
# speak a server API version that the server does not support,
# e.g. this client was pointed at a different server. We'll to
# downgrade to the least possible server API version and try
# again.
if self._message_store.get_server_api() != DEFAULT_SERVER_API:
self._message_store.set_server_api(DEFAULT_SERVER_API)
self.exchange()
return
self._reactor.fire("exchange-failed")
self._message_store.record_failure(int(self._reactor.time()))
logging.info("Message exchange failed.")
exchange_completed()
self._reactor.call_in_thread(handle_result, handle_failure,
self._transport.exchange, payload,
self._registration_info.secure_id,
self._get_exchange_token(),
payload.get("server-api"))
return deferred
def is_urgent(self):
"""Return a bool showing whether there is an urgent exchange scheduled.
"""
return self._urgent_exchange
def schedule_exchange(self, urgent=False, force=False):
"""Schedule an exchange to happen.
The exchange will occur after some time based on whether C{urgent} is
True. An C{impending-exchange} reactor event will be emitted
approximately 10 seconds before the exchange is started.
@param urgent: If true, ensure an exchange happens within the
urgent interval. This will reschedule the exchange if necessary.
If another urgent exchange is already scheduled, nothing happens.
@param force: If true, an exchange will necessarily be scheduled,
even if it was already scheduled before.
"""
if self._stopped:
return
# The 'not self._exchanging' check below is currently untested.
# It's a bit tricky to test as it is preventing rehooking 'exchange'
# while there's a background thread doing the exchange itself.
if (not self._exchanging and
(force or self._exchange_id is None or
urgent and not self._urgent_exchange)):
if urgent:
self._urgent_exchange = True
if self._exchange_id:
self._reactor.cancel_call(self._exchange_id)
if self._urgent_exchange:
interval = self._config.urgent_exchange_interval
else:
interval = self._config.exchange_interval
if self._notification_id is not None:
self._reactor.cancel_call(self._notification_id)
notification_interval = interval - 10
self._notification_id = self._reactor.call_later(
notification_interval, self._notify_impending_exchange)
self._exchange_id = self._reactor.call_later(
interval, self.exchange)
def _get_exchange_token(self):
"""Get the token given us by the server at the last exchange.
It will be C{None} if we are not fully registered yet or if something
bad happened during the last exchange and we could not get the token
that the server had given us.
"""
exchange_token = self._message_store.get_exchange_token()
# Before starting the exchange set the saved token to None. This will
# prevent us from locking ourselves out if the exchange fails or if we
# crash badly, while the server has saved a new token that we couldn't
# receive or persist (this works because if the token is None the
# server will be forgiving and will authenticate us based only on the
# secure ID we provide).
self._message_store.set_exchange_token(None)
self._message_store.commit()
return exchange_token
def _notify_impending_exchange(self):
self._reactor.fire("impending-exchange")
def _make_payload(self):
"""Return a dict representing the complete exchange payload.
The payload will contain all pending messages eligible for
delivery, up to a maximum of C{max_messages} as passed to
the L{__init__} method.
"""
store = self._message_store
accepted_types_digest = self._hash_types(store.get_accepted_types())
messages = store.get_pending_messages(self._max_messages)
total_messages = store.count_pending_messages()
if messages:
# Each message is tagged with the API that the client was
# using at the time the message got added to the store. The
# logic below will make sure that all messages which are added
# to the payload being built will have the same api, and any
# other messages will be postponed to the next exchange.
server_api = messages[0].get("api")
for i, message in enumerate(messages):
if message.get("api") != server_api:
break
else:
i = None
if i is not None:
del messages[i:]
else:
server_api = store.get_server_api()
payload = {"server-api": server_api,
"client-api": CLIENT_API,
"sequence": store.get_sequence(),
"accepted-types": accepted_types_digest,
"messages": messages,
"total-messages": total_messages,
"next-expected-sequence": store.get_server_sequence()}
accepted_client_types = self.get_client_accepted_message_types()
accepted_client_types_hash = self._hash_types(accepted_client_types)
if accepted_client_types_hash != self._client_accepted_types_hash:
payload["client-accepted-types"] = accepted_client_types
return payload
def _hash_types(self, types):
accepted_types_str = ";".join(types)
return md5(accepted_types_str).digest()
def _handle_result(self, payload, result):
"""Handle a response from the server.
Called by L{exchange} after a batch of messages has been
successfully delivered to the server.
If the C{server_uuid} changed, a C{"server-uuid-changed"} event
will be fired.
Call L{handle_message} for each message in C{result}.
@param payload: The payload that was sent to the server.
@param result: The response got in reply to the C{payload}.
"""
message_store = self._message_store
self._client_accepted_types_hash = result.get(
"client-accepted-types-hash")
next_expected = result.get("next-expected-sequence")
old_sequence = message_store.get_sequence()
if next_expected is None:
# If the server doesn't specify anything for the next-expected
# value, just assume that it processed all messages that we sent
# fine.
next_expected = message_store.get_sequence()
next_expected += len(payload["messages"])
message_store_state = got_next_expected(message_store, next_expected)
if message_store_state == ANCIENT:
# The server has probably lost some data we sent it. The
# slate has been wiped clean (by got_next_expected), now
# let's fire an event to tell all the plugins that they
# ought to generate new messages so the server gets some
# up-to-date data.
logging.info("Server asked for ancient data: resynchronizing all "
"state with the server.")
self.send({"type": "resynchronize"})
self._reactor.fire("resynchronize-clients")
# Save the exchange token that the server has sent us. We will provide
# it at the next exchange to prove that we're still the same client.
# See also landscape.broker.transport.
message_store.set_exchange_token(result.get("next-exchange-token"))
old_uuid = message_store.get_server_uuid()
new_uuid = result.get("server-uuid")
if new_uuid != old_uuid:
logging.info("Server UUID changed (old=%s, new=%s)."
% (old_uuid, new_uuid))
self._reactor.fire("server-uuid-changed", old_uuid, new_uuid)
message_store.set_server_uuid(new_uuid)
# Extract the server API from the payload. If it's not there it must
# be 3.2, because it's the one that didn't have this field.
server_api = result.get("server-api", "3.2")
if is_version_higher(server_api, message_store.get_server_api()):
# The server can handle a message API that is higher than the one
# we're currently using. If the highest server API is greater than
# our one, so let's use our own, which is the most recent we can
# speak. Otherwise if the highest server API is less than or equal
# than ours, let's use the server one, because is the most recent
# common one.
lowest_server_api = sort_versions([server_api, self._api])[-1]
message_store.set_server_api(lowest_server_api)
message_store.commit()
sequence = message_store.get_server_sequence()
for message in result.get("messages", ()):
self.handle_message(message)
sequence += 1
message_store.set_server_sequence(sequence)
message_store.commit()
if message_store.get_pending_messages(1):
logging.info("Pending messages remain after the last exchange.")
# Either the server asked us for old messages, or we
# otherwise have more messages even after transferring
# what we could.
if next_expected != old_sequence:
self.schedule_exchange(urgent=True)
def register_message(self, type, handler):
"""Register a handler for the given message type.
The C{handler} callable will to be executed when a message of
type C{type} has been received from the server.
Multiple handlers for the same type will be called in the
order they were registered.
"""
self._message_handlers.setdefault(type, []).append(handler)
self._client_accepted_types.add(type)
def handle_message(self, message):
"""
Handle a message received from the server.
Any message handlers registered with L{register_message} will
be called.
"""
if 'operation-id' in message:
# This is a message that requires a response. Store the secure ID
# so we can check for obsolete results later.
self._exchange_store.add_message_context(
message['operation-id'], self._registration_info.secure_id,
message['type'])
self._reactor.fire("message", message)
# This has plan interference! but whatever.
if message["type"] in self._message_handlers:
for handler in self._message_handlers[message["type"]]:
handler(message)
def register_client_accepted_message_type(self, type):
# stringify the type because it's a dbus.String. It should work
# anyway, but this is just for sanity and less confusing logs.
self._client_accepted_types.add(str(type))
def get_client_accepted_message_types(self):
return sorted(self._client_accepted_types)
def get_accepted_types_diff(old_types, new_types):
old_types = set(old_types)
new_types = set(new_types)
added_types = new_types - old_types
stable_types = old_types & new_types
removed_types = old_types - new_types
diff = []
diff.extend(["+%s" % type for type in added_types])
diff.extend(["%s" % type for type in stable_types])
diff.extend(["-%s" % type for type in removed_types])
return " ".join(diff)