summary refs log tree commit diff
diff options
context:
space:
mode:
-rw-r--r--INSTALL.md12
-rw-r--r--changelog.d/6322.misc1
-rw-r--r--changelog.d/6362.misc1
-rw-r--r--changelog.d/6388.doc1
-rw-r--r--changelog.d/6390.doc1
-rw-r--r--changelog.d/6392.misc1
-rw-r--r--docker/README.md12
-rw-r--r--docs/user_directory.md3
-rw-r--r--synapse/app/federation_sender.py2
-rw-r--r--synapse/appservice/api.py2
-rw-r--r--synapse/config/appservice.py2
-rw-r--r--synapse/config/room_directory.py2
-rw-r--r--synapse/config/server.py6
-rw-r--r--synapse/federation/persistence.py4
-rw-r--r--synapse/federation/sender/__init__.py2
-rw-r--r--synapse/federation/sender/transaction_manager.py4
-rw-r--r--synapse/handlers/directory.py2
-rw-r--r--synapse/http/servlet.py2
-rw-r--r--synapse/logging/_structured.py14
-rw-r--r--synapse/logging/_terse_json.py106
-rw-r--r--synapse/push/httppusher.py5
-rw-r--r--synapse/push/mailer.py4
-rw-r--r--synapse/rest/media/v1/preview_url_resource.py2
-rw-r--r--synapse/server_notices/consent_server_notices.py2
-rw-r--r--synapse/storage/_base.py2
-rw-r--r--synapse/storage/data_stores/main/deviceinbox.py2
-rw-r--r--synapse/storage/data_stores/main/end_to_end_keys.py6
-rw-r--r--synapse/storage/data_stores/main/events.py8
-rw-r--r--synapse/storage/data_stores/main/filtering.py2
-rw-r--r--synapse/storage/data_stores/main/media_repository.py6
-rw-r--r--synapse/storage/data_stores/main/registration.py4
-rw-r--r--synapse/storage/data_stores/main/stream.py2
-rw-r--r--synapse/storage/data_stores/main/tags.py4
-rw-r--r--synapse/storage/prepare_database.py2
-rw-r--r--synapse/streams/config.py9
-rw-r--r--tests/rest/client/v1/test_rooms.py74
-rw-r--r--tests/server.py2
37 files changed, 236 insertions, 80 deletions
diff --git a/INSTALL.md b/INSTALL.md
index 29e0abafd3..9b7360f0ef 100644
--- a/INSTALL.md
+++ b/INSTALL.md
@@ -133,9 +133,9 @@ sudo yum install libtiff-devel libjpeg-devel libzip-devel freetype-devel \
 sudo yum groupinstall "Development Tools"
 ```
 
-#### Mac OS X
+#### macOS
 
-Installing prerequisites on Mac OS X:
+Installing prerequisites on macOS:
 
 ```
 xcode-select --install
@@ -144,6 +144,14 @@ sudo pip install virtualenv
 brew install pkg-config libffi
 ```
 
+On macOS Catalina (10.15) you may need to explicitly install OpenSSL
+via brew and inform `pip` about it so that `psycopg2` builds:
+
+```
+brew install openssl@1.1
+export LDFLAGS=-L/usr/local/Cellar/openssl\@1.1/1.1.1d/lib/
+```
+
 #### OpenSUSE
 
 Installing prerequisites on openSUSE:
diff --git a/changelog.d/6322.misc b/changelog.d/6322.misc
new file mode 100644
index 0000000000..70ef36ca80
--- /dev/null
+++ b/changelog.d/6322.misc
@@ -0,0 +1 @@
+Improve the performance of outputting structured logging.
diff --git a/changelog.d/6362.misc b/changelog.d/6362.misc
new file mode 100644
index 0000000000..b79a5bea99
--- /dev/null
+++ b/changelog.d/6362.misc
@@ -0,0 +1 @@
+Clean up some unnecessary quotation marks around the codebase.
\ No newline at end of file
diff --git a/changelog.d/6388.doc b/changelog.d/6388.doc
new file mode 100644
index 0000000000..c777cb6b8f
--- /dev/null
+++ b/changelog.d/6388.doc
@@ -0,0 +1 @@
+Fix link in the user directory documentation.
diff --git a/changelog.d/6390.doc b/changelog.d/6390.doc
new file mode 100644
index 0000000000..093411bec1
--- /dev/null
+++ b/changelog.d/6390.doc
@@ -0,0 +1 @@
+Add build instructions to the docker readme.
\ No newline at end of file
diff --git a/changelog.d/6392.misc b/changelog.d/6392.misc
new file mode 100644
index 0000000000..a00257944f
--- /dev/null
+++ b/changelog.d/6392.misc
@@ -0,0 +1 @@
+Add a test scenario to make sure room history purges don't break `/messages` in the future.
diff --git a/docker/README.md b/docker/README.md
index 24dfa77dcc..9f112a01d0 100644
--- a/docker/README.md
+++ b/docker/README.md
@@ -130,3 +130,15 @@ docker run -it --rm \
 This will generate the same configuration file as the legacy mode used, but
 will store it in `/data/homeserver.yaml` instead of a temporary location. You
 can then use it as shown above at [Running synapse](#running-synapse).
+
+## Building the image
+
+If you need to build the image from a Synapse checkout, use the following `docker
+ build` command from the repo's root:
+ 
+```
+docker build -t matrixdotorg/synapse -f docker/Dockerfile .
+```
+
+You can choose to build a different docker image by changing the value of the `-f` flag to
+point to another Dockerfile.
diff --git a/docs/user_directory.md b/docs/user_directory.md
index e64aa453cc..37dc71e751 100644
--- a/docs/user_directory.md
+++ b/docs/user_directory.md
@@ -7,7 +7,6 @@ who are present in a publicly viewable room present on the server.
 
 The directory info is stored in various tables, which can (typically after
 DB corruption) get stale or out of sync.  If this happens, for now the
-solution to fix it is to execute the SQL here
-https://github.com/matrix-org/synapse/blob/master/synapse/storage/schema/delta/53/user_dir_populate.sql
+solution to fix it is to execute the SQL [here](../synapse/storage/data_stores/main/schema/delta/53/user_dir_populate.sql)
 and then restart synapse. This should then start a background task to
 flush the current tables and regenerate the directory.
diff --git a/synapse/app/federation_sender.py b/synapse/app/federation_sender.py
index 139221ad34..448e45e00f 100644
--- a/synapse/app/federation_sender.py
+++ b/synapse/app/federation_sender.py
@@ -69,7 +69,7 @@ class FederationSenderSlaveStore(
         self.federation_out_pos_startup = self._get_federation_out_pos(db_conn)
 
     def _get_federation_out_pos(self, db_conn):
-        sql = "SELECT stream_id FROM federation_stream_position" " WHERE type = ?"
+        sql = "SELECT stream_id FROM federation_stream_position WHERE type = ?"
         sql = self.database_engine.convert_param_style(sql)
 
         txn = db_conn.cursor()
diff --git a/synapse/appservice/api.py b/synapse/appservice/api.py
index 3e25bf5747..57174da021 100644
--- a/synapse/appservice/api.py
+++ b/synapse/appservice/api.py
@@ -185,7 +185,7 @@ class ApplicationServiceApi(SimpleHttpClient):
 
                 if not _is_valid_3pe_metadata(info):
                     logger.warning(
-                        "query_3pe_protocol to %s did not return a" " valid result", uri
+                        "query_3pe_protocol to %s did not return a valid result", uri
                     )
                     return None
 
diff --git a/synapse/config/appservice.py b/synapse/config/appservice.py
index e77d3387ff..ca43e96bd1 100644
--- a/synapse/config/appservice.py
+++ b/synapse/config/appservice.py
@@ -134,7 +134,7 @@ def _load_appservice(hostname, as_info, config_filename):
             for regex_obj in as_info["namespaces"][ns]:
                 if not isinstance(regex_obj, dict):
                     raise ValueError(
-                        "Expected namespace entry in %s to be an object," " but got %s",
+                        "Expected namespace entry in %s to be an object, but got %s",
                         ns,
                         regex_obj,
                     )
diff --git a/synapse/config/room_directory.py b/synapse/config/room_directory.py
index 7c9f05bde4..7ac7699676 100644
--- a/synapse/config/room_directory.py
+++ b/synapse/config/room_directory.py
@@ -170,7 +170,7 @@ class _RoomDirectoryRule(object):
             self.action = action
         else:
             raise ConfigError(
-                "%s rules can only have action of 'allow'" " or 'deny'" % (option_name,)
+                "%s rules can only have action of 'allow' or 'deny'" % (option_name,)
             )
 
         self._alias_matches_all = alias == "*"
diff --git a/synapse/config/server.py b/synapse/config/server.py
index 00d01c43af..11336d7549 100644
--- a/synapse/config/server.py
+++ b/synapse/config/server.py
@@ -223,7 +223,7 @@ class ServerConfig(Config):
             self.federation_ip_range_blacklist.update(["0.0.0.0", "::"])
         except Exception as e:
             raise ConfigError(
-                "Invalid range(s) provided in " "federation_ip_range_blacklist: %s" % e
+                "Invalid range(s) provided in federation_ip_range_blacklist: %s" % e
             )
 
         if self.public_baseurl is not None:
@@ -787,14 +787,14 @@ class ServerConfig(Config):
             "--print-pidfile",
             action="store_true",
             default=None,
-            help="Print the path to the pidfile just" " before daemonizing",
+            help="Print the path to the pidfile just before daemonizing",
         )
         server_group.add_argument(
             "--manhole",
             metavar="PORT",
             dest="manhole",
             type=int,
-            help="Turn on the twisted telnet manhole" " service on the given port.",
+            help="Turn on the twisted telnet manhole service on the given port.",
         )
 
 
diff --git a/synapse/federation/persistence.py b/synapse/federation/persistence.py
index 44edcabed4..d68b4bd670 100644
--- a/synapse/federation/persistence.py
+++ b/synapse/federation/persistence.py
@@ -44,7 +44,7 @@ class TransactionActions(object):
             response code and response body.
         """
         if not transaction.transaction_id:
-            raise RuntimeError("Cannot persist a transaction with no " "transaction_id")
+            raise RuntimeError("Cannot persist a transaction with no transaction_id")
 
         return self.store.get_received_txn_response(transaction.transaction_id, origin)
 
@@ -56,7 +56,7 @@ class TransactionActions(object):
             Deferred
         """
         if not transaction.transaction_id:
-            raise RuntimeError("Cannot persist a transaction with no " "transaction_id")
+            raise RuntimeError("Cannot persist a transaction with no transaction_id")
 
         return self.store.set_received_txn_response(
             transaction.transaction_id, origin, code, response
diff --git a/synapse/federation/sender/__init__.py b/synapse/federation/sender/__init__.py
index 2b2ee8612a..4ebb0e8bc0 100644
--- a/synapse/federation/sender/__init__.py
+++ b/synapse/federation/sender/__init__.py
@@ -49,7 +49,7 @@ sent_pdus_destination_dist_count = Counter(
 
 sent_pdus_destination_dist_total = Counter(
     "synapse_federation_client_sent_pdu_destinations:total",
-    "" "Total number of PDUs queued for sending across all destinations",
+    "Total number of PDUs queued for sending across all destinations",
 )
 
 
diff --git a/synapse/federation/sender/transaction_manager.py b/synapse/federation/sender/transaction_manager.py
index 67b3e1ab6e..5fed626d5b 100644
--- a/synapse/federation/sender/transaction_manager.py
+++ b/synapse/federation/sender/transaction_manager.py
@@ -84,7 +84,7 @@ class TransactionManager(object):
             txn_id = str(self._next_txn_id)
 
             logger.debug(
-                "TX [%s] {%s} Attempting new transaction" " (pdus: %d, edus: %d)",
+                "TX [%s] {%s} Attempting new transaction (pdus: %d, edus: %d)",
                 destination,
                 txn_id,
                 len(pdus),
@@ -103,7 +103,7 @@ class TransactionManager(object):
             self._next_txn_id += 1
 
             logger.info(
-                "TX [%s] {%s} Sending transaction [%s]," " (PDUs: %d, EDUs: %d)",
+                "TX [%s] {%s} Sending transaction [%s], (PDUs: %d, EDUs: %d)",
                 destination,
                 txn_id,
                 transaction.transaction_id,
diff --git a/synapse/handlers/directory.py b/synapse/handlers/directory.py
index 69051101a6..a07d2f1a17 100644
--- a/synapse/handlers/directory.py
+++ b/synapse/handlers/directory.py
@@ -119,7 +119,7 @@ class DirectoryHandler(BaseHandler):
             if not service.is_interested_in_alias(room_alias.to_string()):
                 raise SynapseError(
                     400,
-                    "This application service has not reserved" " this kind of alias.",
+                    "This application service has not reserved this kind of alias.",
                     errcode=Codes.EXCLUSIVE,
                 )
         else:
diff --git a/synapse/http/servlet.py b/synapse/http/servlet.py
index e9a5e46ced..13fcb408a6 100644
--- a/synapse/http/servlet.py
+++ b/synapse/http/servlet.py
@@ -96,7 +96,7 @@ def parse_boolean_from_args(args, name, default=None, required=False):
             return {b"true": True, b"false": False}[args[name][0]]
         except Exception:
             message = (
-                "Boolean query parameter %r must be one of" " ['true', 'false']"
+                "Boolean query parameter %r must be one of ['true', 'false']"
             ) % (name,)
             raise SynapseError(400, message)
     else:
diff --git a/synapse/logging/_structured.py b/synapse/logging/_structured.py
index 334ddaf39a..ffa7b20ca8 100644
--- a/synapse/logging/_structured.py
+++ b/synapse/logging/_structured.py
@@ -261,6 +261,18 @@ def parse_drain_configs(
             )
 
 
+class StoppableLogPublisher(LogPublisher):
+    """
+    A log publisher that can tell its observers to shut down any external
+    communications.
+    """
+
+    def stop(self):
+        for obs in self._observers:
+            if hasattr(obs, "stop"):
+                obs.stop()
+
+
 def setup_structured_logging(
     hs,
     config,
@@ -336,7 +348,7 @@ def setup_structured_logging(
             # We should never get here, but, just in case, throw an error.
             raise ConfigError("%s drain type cannot be configured" % (observer.type,))
 
-    publisher = LogPublisher(*observers)
+    publisher = StoppableLogPublisher(*observers)
     log_filter = LogLevelFilterPredicate()
 
     for namespace, namespace_config in log_config.get(
diff --git a/synapse/logging/_terse_json.py b/synapse/logging/_terse_json.py
index 76ce7d8808..05fc64f409 100644
--- a/synapse/logging/_terse_json.py
+++ b/synapse/logging/_terse_json.py
@@ -17,25 +17,29 @@
 Log formatters that output terse JSON.
 """
 
+import json
 import sys
+import traceback
 from collections import deque
 from ipaddress import IPv4Address, IPv6Address, ip_address
 from math import floor
-from typing import IO
+from typing import IO, Optional
 
 import attr
-from simplejson import dumps
 from zope.interface import implementer
 
 from twisted.application.internet import ClientService
+from twisted.internet.defer import Deferred
 from twisted.internet.endpoints import (
     HostnameEndpoint,
     TCP4ClientEndpoint,
     TCP6ClientEndpoint,
 )
+from twisted.internet.interfaces import IPushProducer, ITransport
 from twisted.internet.protocol import Factory, Protocol
 from twisted.logger import FileLogObserver, ILogObserver, Logger
-from twisted.python.failure import Failure
+
+_encoder = json.JSONEncoder(ensure_ascii=False, separators=(",", ":"))
 
 
 def flatten_event(event: dict, metadata: dict, include_time: bool = False):
@@ -141,12 +145,50 @@ def TerseJSONToConsoleLogObserver(outFile: IO[str], metadata: dict) -> FileLogOb
 
     def formatEvent(_event: dict) -> str:
         flattened = flatten_event(_event, metadata)
-        return dumps(flattened, ensure_ascii=False, separators=(",", ":")) + "\n"
+        return _encoder.encode(flattened) + "\n"
 
     return FileLogObserver(outFile, formatEvent)
 
 
 @attr.s
+@implementer(IPushProducer)
+class LogProducer(object):
+    """
+    An IPushProducer that writes logs from its buffer to its transport when it
+    is resumed.
+
+    Args:
+        buffer: Log buffer to read logs from.
+        transport: Transport to write to.
+    """
+
+    transport = attr.ib(type=ITransport)
+    _buffer = attr.ib(type=deque)
+    _paused = attr.ib(default=False, type=bool, init=False)
+
+    def pauseProducing(self):
+        self._paused = True
+
+    def stopProducing(self):
+        self._paused = True
+        self._buffer = None
+
+    def resumeProducing(self):
+        self._paused = False
+
+        while self._paused is False and (self._buffer and self.transport.connected):
+            try:
+                event = self._buffer.popleft()
+                self.transport.write(_encoder.encode(event).encode("utf8"))
+                self.transport.write(b"\n")
+            except Exception:
+                # Something has gone wrong writing to the transport -- log it
+                # and break out of the while.
+                traceback.print_exc(file=sys.__stderr__)
+                break
+
+
+@attr.s
 @implementer(ILogObserver)
 class TerseJSONToTCPLogObserver(object):
     """
@@ -165,8 +207,9 @@ class TerseJSONToTCPLogObserver(object):
     metadata = attr.ib(type=dict)
     maximum_buffer = attr.ib(type=int)
     _buffer = attr.ib(default=attr.Factory(deque), type=deque)
-    _writer = attr.ib(default=None)
+    _connection_waiter = attr.ib(default=None, type=Optional[Deferred])
     _logger = attr.ib(default=attr.Factory(Logger))
+    _producer = attr.ib(default=None, type=Optional[LogProducer])
 
     def start(self) -> None:
 
@@ -187,38 +230,43 @@ class TerseJSONToTCPLogObserver(object):
         factory = Factory.forProtocol(Protocol)
         self._service = ClientService(endpoint, factory, clock=self.hs.get_reactor())
         self._service.startService()
+        self._connect()
 
-    def _write_loop(self) -> None:
+    def stop(self):
+        self._service.stopService()
+
+    def _connect(self) -> None:
         """
-        Implement the write loop.
+        Triggers an attempt to connect then write to the remote if not already writing.
         """
-        if self._writer:
+        if self._connection_waiter:
             return
 
-        self._writer = self._service.whenConnected()
+        self._connection_waiter = self._service.whenConnected(failAfterFailures=1)
+
+        @self._connection_waiter.addErrback
+        def fail(r):
+            r.printTraceback(file=sys.__stderr__)
+            self._connection_waiter = None
+            self._connect()
 
-        @self._writer.addBoth
+        @self._connection_waiter.addCallback
         def writer(r):
-            if isinstance(r, Failure):
-                r.printTraceback(file=sys.__stderr__)
-                self._writer = None
-                self.hs.get_reactor().callLater(1, self._write_loop)
+            # We have a connection. If we already have a producer, and its
+            # transport is the same, just trigger a resumeProducing.
+            if self._producer and r.transport is self._producer.transport:
+                self._producer.resumeProducing()
                 return
 
-            try:
-                for event in self._buffer:
-                    r.transport.write(
-                        dumps(event, ensure_ascii=False, separators=(",", ":")).encode(
-                            "utf8"
-                        )
-                    )
-                    r.transport.write(b"\n")
-                self._buffer.clear()
-            except Exception as e:
-                sys.__stderr__.write("Failed writing out logs with %s\n" % (str(e),))
-
-            self._writer = False
-            self.hs.get_reactor().callLater(1, self._write_loop)
+            # If the producer is still producing, stop it.
+            if self._producer:
+                self._producer.stopProducing()
+
+            # Make a new producer and start it.
+            self._producer = LogProducer(buffer=self._buffer, transport=r.transport)
+            r.transport.registerProducer(self._producer, True)
+            self._producer.resumeProducing()
+            self._connection_waiter = None
 
     def _handle_pressure(self) -> None:
         """
@@ -277,4 +325,4 @@ class TerseJSONToTCPLogObserver(object):
             self._logger.failure("Failed clearing backpressure")
 
         # Try and write immediately.
-        self._write_loop()
+        self._connect()
diff --git a/synapse/push/httppusher.py b/synapse/push/httppusher.py
index e994037be6..d0879b0490 100644
--- a/synapse/push/httppusher.py
+++ b/synapse/push/httppusher.py
@@ -246,7 +246,7 @@ class HttpPusher(object):
                     # fixed, we don't suddenly deliver a load
                     # of old notifications.
                     logger.warning(
-                        "Giving up on a notification to user %s, " "pushkey %s",
+                        "Giving up on a notification to user %s, pushkey %s",
                         self.user_id,
                         self.pushkey,
                     )
@@ -299,8 +299,7 @@ class HttpPusher(object):
                     # for sanity, we only remove the pushkey if it
                     # was the one we actually sent...
                     logger.warning(
-                        ("Ignoring rejected pushkey %s because we" " didn't send it"),
-                        pk,
+                        ("Ignoring rejected pushkey %s because we didn't send it"), pk,
                     )
                 else:
                     logger.info("Pushkey %s was rejected: removing", pk)
diff --git a/synapse/push/mailer.py b/synapse/push/mailer.py
index 1d15a06a58..b13b646bfd 100644
--- a/synapse/push/mailer.py
+++ b/synapse/push/mailer.py
@@ -43,7 +43,7 @@ logger = logging.getLogger(__name__)
 
 
 MESSAGE_FROM_PERSON_IN_ROOM = (
-    "You have a message on %(app)s from %(person)s " "in the %(room)s room..."
+    "You have a message on %(app)s from %(person)s in the %(room)s room..."
 )
 MESSAGE_FROM_PERSON = "You have a message on %(app)s from %(person)s..."
 MESSAGES_FROM_PERSON = "You have messages on %(app)s from %(person)s..."
@@ -55,7 +55,7 @@ MESSAGES_FROM_PERSON_AND_OTHERS = (
     "You have messages on %(app)s from %(person)s and others..."
 )
 INVITE_FROM_PERSON_TO_ROOM = (
-    "%(person)s has invited you to join the " "%(room)s room on %(app)s..."
+    "%(person)s has invited you to join the %(room)s room on %(app)s..."
 )
 INVITE_FROM_PERSON = "%(person)s has invited you to chat on %(app)s..."
 
diff --git a/synapse/rest/media/v1/preview_url_resource.py b/synapse/rest/media/v1/preview_url_resource.py
index 15c15a12f5..a23d6f5c75 100644
--- a/synapse/rest/media/v1/preview_url_resource.py
+++ b/synapse/rest/media/v1/preview_url_resource.py
@@ -122,7 +122,7 @@ class PreviewUrlResource(DirectServeResource):
                 pattern = entry[attrib]
                 value = getattr(url_tuple, attrib)
                 logger.debug(
-                    "Matching attrib '%s' with value '%s' against" " pattern '%s'",
+                    "Matching attrib '%s' with value '%s' against pattern '%s'",
                     attrib,
                     value,
                     pattern,
diff --git a/synapse/server_notices/consent_server_notices.py b/synapse/server_notices/consent_server_notices.py
index 415e9c17d8..5736c56032 100644
--- a/synapse/server_notices/consent_server_notices.py
+++ b/synapse/server_notices/consent_server_notices.py
@@ -54,7 +54,7 @@ class ConsentServerNotices(object):
                 )
             if "body" not in self._server_notice_content:
                 raise ConfigError(
-                    "user_consent server_notice_consent must contain a 'body' " "key."
+                    "user_consent server_notice_consent must contain a 'body' key."
                 )
 
             self._consent_uri_builder = ConsentURIBuilder(hs.config)
diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py
index ab596fa68d..6b8a9cd89a 100644
--- a/synapse/storage/_base.py
+++ b/synapse/storage/_base.py
@@ -851,7 +851,7 @@ class SQLBaseStore(object):
             allvalues.update(values)
             latter = "UPDATE SET " + ", ".join(k + "=EXCLUDED." + k for k in values)
 
-        sql = ("INSERT INTO %s (%s) VALUES (%s) " "ON CONFLICT (%s) DO %s") % (
+        sql = ("INSERT INTO %s (%s) VALUES (%s) ON CONFLICT (%s) DO %s") % (
             table,
             ", ".join(k for k in allvalues),
             ", ".join("?" for _ in allvalues),
diff --git a/synapse/storage/data_stores/main/deviceinbox.py b/synapse/storage/data_stores/main/deviceinbox.py
index 96cd0fb77a..a23744f11c 100644
--- a/synapse/storage/data_stores/main/deviceinbox.py
+++ b/synapse/storage/data_stores/main/deviceinbox.py
@@ -380,7 +380,7 @@ class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore)
             devices = list(messages_by_device.keys())
             if len(devices) == 1 and devices[0] == "*":
                 # Handle wildcard device_ids.
-                sql = "SELECT device_id FROM devices" " WHERE user_id = ?"
+                sql = "SELECT device_id FROM devices WHERE user_id = ?"
                 txn.execute(sql, (user_id,))
                 message_json = json.dumps(messages_by_device["*"])
                 for row in txn:
diff --git a/synapse/storage/data_stores/main/end_to_end_keys.py b/synapse/storage/data_stores/main/end_to_end_keys.py
index 073412a78d..d8ad59ad93 100644
--- a/synapse/storage/data_stores/main/end_to_end_keys.py
+++ b/synapse/storage/data_stores/main/end_to_end_keys.py
@@ -138,9 +138,9 @@ class EndToEndKeyWorkerStore(SQLBaseStore):
                 result.setdefault(user_id, {})[device_id] = None
 
         # get signatures on the device
-        signature_sql = (
-            "SELECT * " "  FROM e2e_cross_signing_signatures " " WHERE %s"
-        ) % (" OR ".join("(" + q + ")" for q in signature_query_clauses))
+        signature_sql = ("SELECT *  FROM e2e_cross_signing_signatures WHERE %s") % (
+            " OR ".join("(" + q + ")" for q in signature_query_clauses)
+        )
 
         txn.execute(signature_sql, signature_query_params)
         rows = self.cursor_to_dict(txn)
diff --git a/synapse/storage/data_stores/main/events.py b/synapse/storage/data_stores/main/events.py
index 878f7568a6..627c0b67f1 100644
--- a/synapse/storage/data_stores/main/events.py
+++ b/synapse/storage/data_stores/main/events.py
@@ -713,9 +713,7 @@ class EventsStore(
 
                 metadata_json = encode_json(event.internal_metadata.get_dict())
 
-                sql = (
-                    "UPDATE event_json SET internal_metadata = ?" " WHERE event_id = ?"
-                )
+                sql = "UPDATE event_json SET internal_metadata = ? WHERE event_id = ?"
                 txn.execute(sql, (metadata_json, event.event_id))
 
                 # Add an entry to the ex_outlier_stream table to replicate the
@@ -732,7 +730,7 @@ class EventsStore(
                     },
                 )
 
-                sql = "UPDATE events SET outlier = ?" " WHERE event_id = ?"
+                sql = "UPDATE events SET outlier = ? WHERE event_id = ?"
                 txn.execute(sql, (False, event.event_id))
 
                 # Update the event_backward_extremities table now that this
@@ -1479,7 +1477,7 @@ class EventsStore(
 
         # We do joins against events_to_purge for e.g. calculating state
         # groups to purge, etc., so lets make an index.
-        txn.execute("CREATE INDEX events_to_purge_id" " ON events_to_purge(event_id)")
+        txn.execute("CREATE INDEX events_to_purge_id ON events_to_purge(event_id)")
 
         txn.execute("SELECT event_id, should_delete FROM events_to_purge")
         event_rows = txn.fetchall()
diff --git a/synapse/storage/data_stores/main/filtering.py b/synapse/storage/data_stores/main/filtering.py
index a2a2a67927..f05ace299a 100644
--- a/synapse/storage/data_stores/main/filtering.py
+++ b/synapse/storage/data_stores/main/filtering.py
@@ -55,7 +55,7 @@ class FilteringStore(SQLBaseStore):
             if filter_id_response is not None:
                 return filter_id_response[0]
 
-            sql = "SELECT MAX(filter_id) FROM user_filters " "WHERE user_id = ?"
+            sql = "SELECT MAX(filter_id) FROM user_filters WHERE user_id = ?"
             txn.execute(sql, (user_localpart,))
             max_id = txn.fetchone()[0]
             if max_id is None:
diff --git a/synapse/storage/data_stores/main/media_repository.py b/synapse/storage/data_stores/main/media_repository.py
index 84b5f3ad5e..0f2887bdce 100644
--- a/synapse/storage/data_stores/main/media_repository.py
+++ b/synapse/storage/data_stores/main/media_repository.py
@@ -337,7 +337,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
         if len(media_ids) == 0:
             return
 
-        sql = "DELETE FROM local_media_repository_url_cache" " WHERE media_id = ?"
+        sql = "DELETE FROM local_media_repository_url_cache WHERE media_id = ?"
 
         def _delete_url_cache_txn(txn):
             txn.executemany(sql, [(media_id,) for media_id in media_ids])
@@ -365,11 +365,11 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
             return
 
         def _delete_url_cache_media_txn(txn):
-            sql = "DELETE FROM local_media_repository" " WHERE media_id = ?"
+            sql = "DELETE FROM local_media_repository WHERE media_id = ?"
 
             txn.executemany(sql, [(media_id,) for media_id in media_ids])
 
-            sql = "DELETE FROM local_media_repository_thumbnails" " WHERE media_id = ?"
+            sql = "DELETE FROM local_media_repository_thumbnails WHERE media_id = ?"
 
             txn.executemany(sql, [(media_id,) for media_id in media_ids])
 
diff --git a/synapse/storage/data_stores/main/registration.py b/synapse/storage/data_stores/main/registration.py
index ee1b2b2bbf..6a594c160c 100644
--- a/synapse/storage/data_stores/main/registration.py
+++ b/synapse/storage/data_stores/main/registration.py
@@ -377,9 +377,7 @@ class RegistrationWorkerStore(SQLBaseStore):
         """
 
         def f(txn):
-            sql = (
-                "SELECT name, password_hash FROM users" " WHERE lower(name) = lower(?)"
-            )
+            sql = "SELECT name, password_hash FROM users WHERE lower(name) = lower(?)"
             txn.execute(sql, (user_id,))
             return dict(txn)
 
diff --git a/synapse/storage/data_stores/main/stream.py b/synapse/storage/data_stores/main/stream.py
index 8780fdd989..9ae4a913a1 100644
--- a/synapse/storage/data_stores/main/stream.py
+++ b/synapse/storage/data_stores/main/stream.py
@@ -616,7 +616,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore):
 
     def _get_max_topological_txn(self, txn, room_id):
         txn.execute(
-            "SELECT MAX(topological_ordering) FROM events" " WHERE room_id = ?",
+            "SELECT MAX(topological_ordering) FROM events WHERE room_id = ?",
             (room_id,),
         )
 
diff --git a/synapse/storage/data_stores/main/tags.py b/synapse/storage/data_stores/main/tags.py
index 10d1887f75..aa24339717 100644
--- a/synapse/storage/data_stores/main/tags.py
+++ b/synapse/storage/data_stores/main/tags.py
@@ -83,9 +83,7 @@ class TagsWorkerStore(AccountDataWorkerStore):
         )
 
         def get_tag_content(txn, tag_ids):
-            sql = (
-                "SELECT tag, content" " FROM room_tags" " WHERE user_id=? AND room_id=?"
-            )
+            sql = "SELECT tag, content FROM room_tags WHERE user_id=? AND room_id=?"
             results = []
             for stream_id, user_id, room_id in tag_ids:
                 txn.execute(sql, (user_id, room_id))
diff --git a/synapse/storage/prepare_database.py b/synapse/storage/prepare_database.py
index 2e7753820e..731e1c9d9c 100644
--- a/synapse/storage/prepare_database.py
+++ b/synapse/storage/prepare_database.py
@@ -447,7 +447,7 @@ def _apply_module_schema_files(cur, database_engine, modname, names_and_streams)
         # Mark as done.
         cur.execute(
             database_engine.convert_param_style(
-                "INSERT INTO applied_module_schemas (module_name, file)" " VALUES (?,?)"
+                "INSERT INTO applied_module_schemas (module_name, file) VALUES (?,?)"
             ),
             (modname, name),
         )
diff --git a/synapse/streams/config.py b/synapse/streams/config.py
index 02994ab2a5..cd56cd91ed 100644
--- a/synapse/streams/config.py
+++ b/synapse/streams/config.py
@@ -88,9 +88,12 @@ class PaginationConfig(object):
             raise SynapseError(400, "Invalid request.")
 
     def __repr__(self):
-        return (
-            "PaginationConfig(from_tok=%r, to_tok=%r," " direction=%r, limit=%r)"
-        ) % (self.from_token, self.to_token, self.direction, self.limit)
+        return ("PaginationConfig(from_tok=%r, to_tok=%r, direction=%r, limit=%r)") % (
+            self.from_token,
+            self.to_token,
+            self.direction,
+            self.limit,
+        )
 
     def get_source_config(self, source_name):
         keyname = "%s_key" % source_name
diff --git a/tests/rest/client/v1/test_rooms.py b/tests/rest/client/v1/test_rooms.py
index 5e38fd6ced..e84e578f99 100644
--- a/tests/rest/client/v1/test_rooms.py
+++ b/tests/rest/client/v1/test_rooms.py
@@ -25,7 +25,9 @@ from twisted.internet import defer
 
 import synapse.rest.admin
 from synapse.api.constants import EventContentFields, EventTypes, Membership
+from synapse.handlers.pagination import PurgeStatus
 from synapse.rest.client.v1 import login, profile, room
+from synapse.util.stringutils import random_string
 
 from tests import unittest
 
@@ -910,6 +912,78 @@ class RoomMessageListTestCase(RoomBase):
 
         return channel.json_body["chunk"]
 
+    def test_room_messages_purge(self):
+        store = self.hs.get_datastore()
+        pagination_handler = self.hs.get_pagination_handler()
+
+        # Send a first message in the room, which will be removed by the purge.
+        first_event_id = self.helper.send(self.room_id, "message 1")["event_id"]
+        first_token = self.get_success(
+            store.get_topological_token_for_event(first_event_id)
+        )
+
+        # Send a second message in the room, which won't be removed, and which we'll
+        # use as the marker to purge events before.
+        second_event_id = self.helper.send(self.room_id, "message 2")["event_id"]
+        second_token = self.get_success(
+            store.get_topological_token_for_event(second_event_id)
+        )
+
+        # Send a third event in the room to ensure we don't fall under any edge case
+        # due to our marker being the latest forward extremity in the room.
+        self.helper.send(self.room_id, "message 3")
+
+        # Check that we get the first and second message when querying /messages.
+        request, channel = self.make_request(
+            "GET",
+            "/rooms/%s/messages?access_token=x&from=%s&dir=b&filter=%s"
+            % (self.room_id, second_token, json.dumps({"types": [EventTypes.Message]})),
+        )
+        self.render(request)
+        self.assertEqual(channel.code, 200, channel.json_body)
+
+        chunk = channel.json_body["chunk"]
+        self.assertEqual(len(chunk), 2, [event["content"] for event in chunk])
+
+        # Purge every event before the second event.
+        purge_id = random_string(16)
+        pagination_handler._purges_by_id[purge_id] = PurgeStatus()
+        self.get_success(
+            pagination_handler._purge_history(
+                purge_id=purge_id,
+                room_id=self.room_id,
+                token=second_token,
+                delete_local_events=True,
+            )
+        )
+
+        # Check that we only get the second message through /message now that the first
+        # has been purged.
+        request, channel = self.make_request(
+            "GET",
+            "/rooms/%s/messages?access_token=x&from=%s&dir=b&filter=%s"
+            % (self.room_id, second_token, json.dumps({"types": [EventTypes.Message]})),
+        )
+        self.render(request)
+        self.assertEqual(channel.code, 200, channel.json_body)
+
+        chunk = channel.json_body["chunk"]
+        self.assertEqual(len(chunk), 1, [event["content"] for event in chunk])
+
+        # Check that we get no event, but also no error, when querying /messages with
+        # the token that was pointing at the first event, because we don't have it
+        # anymore.
+        request, channel = self.make_request(
+            "GET",
+            "/rooms/%s/messages?access_token=x&from=%s&dir=b&filter=%s"
+            % (self.room_id, first_token, json.dumps({"types": [EventTypes.Message]})),
+        )
+        self.render(request)
+        self.assertEqual(channel.code, 200, channel.json_body)
+
+        chunk = channel.json_body["chunk"]
+        self.assertEqual(len(chunk), 0, [event["content"] for event in chunk])
+
 
 class RoomSearchTestCase(unittest.HomeserverTestCase):
     servlets = [
diff --git a/tests/server.py b/tests/server.py
index f878aeaada..2b7cf4242e 100644
--- a/tests/server.py
+++ b/tests/server.py
@@ -379,6 +379,7 @@ class FakeTransport(object):
 
     disconnecting = False
     disconnected = False
+    connected = True
     buffer = attr.ib(default=b"")
     producer = attr.ib(default=None)
     autoflush = attr.ib(default=True)
@@ -402,6 +403,7 @@ class FakeTransport(object):
                     "FakeTransport: Delaying disconnect until buffer is flushed"
                 )
             else:
+                self.connected = False
                 self.disconnected = True
 
     def abortConnection(self):