BIND 10 trac1927, updated. c8b7d82c25958cdae8bb02e3a93b2b8d6dd078d1 [1927] Test msgq normal operations
BIND 10 source code commits
bind10-changes at lists.isc.org
Tue Mar 12 11:20:53 UTC 2013
The branch, trac1927 has been updated
via c8b7d82c25958cdae8bb02e3a93b2b8d6dd078d1 (commit)
from 5e66a7a9ecddb8e00f5ed44fa22a94d02a177e8d (commit)
Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.
- Log -----------------------------------------------------------------
commit c8b7d82c25958cdae8bb02e3a93b2b8d6dd078d1
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date: Tue Mar 12 12:19:26 2013 +0100
[1927] Test msgq normal operations
The success cases of sending, receiving, subscribing, etc.
-----------------------------------------------------------------------
Summary of changes:
src/bin/msgq/msgq.py.in | 1 +
src/bin/msgq/tests/msgq_run_test.py | 160 ++++++++++++++++++++++++++++++++++-
2 files changed, 157 insertions(+), 4 deletions(-)
-----------------------------------------------------------------------
diff --git a/src/bin/msgq/msgq.py.in b/src/bin/msgq/msgq.py.in
index 6435c92..f1f20eb 100755
--- a/src/bin/msgq/msgq.py.in
+++ b/src/bin/msgq/msgq.py.in
@@ -42,6 +42,7 @@ import isc.cc
isc.util.process.rename()
isc.log.init("b10-msgq", buffer=True)
+isc.log.resetUnitTestRootLogger()
# Logger that is used in the actual msgq handling - startup, shutdown and the
# poller thread.
logger = isc.log.Logger("msgq")
diff --git a/src/bin/msgq/tests/msgq_run_test.py b/src/bin/msgq/tests/msgq_run_test.py
index 8323ae7..2721daf 100644
--- a/src/bin/msgq/tests/msgq_run_test.py
+++ b/src/bin/msgq/tests/msgq_run_test.py
@@ -40,6 +40,8 @@ import subprocess
import time
import isc.log
+import isc.cc.session
+from isc.cc.proto_defs import *
SOCKET_PATH = os.path.abspath(os.environ['BIND10_TEST_SOCKET_FILE'])
MSGQ_PATH = os.environ['B10_FROM_BUILD'] + '/src/bin/msgq/run_msgq.sh'
@@ -53,6 +55,7 @@ class MsgqRunTest(unittest.TestCase):
and wait for it to start.
"""
self.__msgq = None
+ self.__opened_connections = []
# A precondition check
self.assertFalse(os.path.exists(SOCKET_PATH))
signal.alarm(TIMEOUT)
@@ -71,6 +74,15 @@ class MsgqRunTest(unittest.TestCase):
while not os.path.exists(SOCKET_PATH):
# Just a short wait, so we don't hog CPU, but don't wait too long
time.sleep(0.01)
+ # Some testing data
+ self.__no_recpt = {"result": [-1, "No such recipient"]}
+
+ def __message(self, data):
+ """
+ Provide some testing message. The data will be included in it, so
+ several different messages can be created.
+ """
+ return {"Message": "Text", "Data": data}
def tearDown(self):
"""
@@ -93,6 +105,9 @@ class MsgqRunTest(unittest.TestCase):
Kill msgq (if running) and restore original signal handlers.
"""
# Remove the socket (as we kill, msgq might not clean up)
+ for conn in self.__opened_connections:
+ conn.close()
+ self.__opened_connections = []
if self.__msgq:
self.__msgq.kill()
self.__msgq = None
@@ -103,13 +118,150 @@ class MsgqRunTest(unittest.TestCase):
# Cancel timeout (so someone else is not hit by it)
signal.alarm(0)
- def test_empty_run(self):
+ def __get_connection(self):
+ """
+ Create a connection to the daemon and make sure it is properly closed
+ at the end of the test.
+ """
+ connection = isc.cc.session.Session(SOCKET_PATH)
+ self.__opened_connections.append(connection)
+ return connection
+
+ def test_send_direct(self):
+ """
+ Connect twice to msgq, send a message from one to another using direct
+ l-name and see it comes.
+ """
+ # Create the connections
+ conn1 = self.__get_connection()
+ conn2 = self.__get_connection()
+ # Send the message
+ lname1 = conn1.lname
+ conn2.group_sendmsg(self.__message(1), "*", to=lname1)
+ # Receive the message and see it contains correct data
+ (msg, env) = conn1.group_recvmsg(nonblock=False)
+ self.assertEqual(self.__message(1), msg)
+ # We don't check there are no extra headers, just that none are missing
+ # or wrong.
+ self.assertEqual(lname1, env[CC_HEADER_TO])
+ self.assertEqual(conn2.lname, env[CC_HEADER_FROM])
+ self.assertEqual("*", env[CC_HEADER_GROUP])
+ self.assertEqual(CC_INSTANCE_WILDCARD, env[CC_HEADER_INSTANCE])
+ self.assertEqual(CC_COMMAND_SEND, env[CC_HEADER_TYPE])
+ self.assertFalse(env[CC_HEADER_WANT_ANSWER])
+
+ def __barrier(self, connections):
+ """
+ Make sure all previous commands on all supplied connections are
+ processed, by sending a ping and waiting for an answer.
+ """
+ for c in connections:
+ c.sendmsg({"type": "ping"})
+ for c in connections:
+ pong = c.recvmsg(nonblock=False)
+ self.assertEqual(({"type": "pong"}, None), pong)
+
+ def test_send_group(self):
+ """
+ Create several connections. First, try to send a message to a (empty)
+ group and see an error is bounced back. Then subscribe the others
+ to the group and send it again. Send to a different group and see it
+ bounced back. Unsubscribe and see it is bounced again.
+
+ Then the other connections answer (after unsubscribing, strange, but
+ legal). See both answers come.
+
+ Then, look there are no more waiting messages.
+ """
+ conn_a = self.__get_connection()
+ conn_b = []
+ for i in range(0, 10):
+ conn_b.append(self.__get_connection())
+ # Send a message to empty group and get an error answer
+ seq = conn_a.group_sendmsg(self.__message(1), "group",
+ want_answer=True)
+ (msg, env) = conn_a.group_recvmsg(nonblock=False, seq=seq)
+ self.assertEqual(self.__no_recpt, msg)
+ self.assertEqual(conn_a.lname, env[CC_HEADER_TO])
+ # Subscribe the two connections
+ for c in conn_b:
+ c.group_subscribe("group")
+ # The subscribe doesn't wait for answer, so make sure it is
+ # all processed before continuing.
+ self.__barrier(conn_b)
+ # Send a message to the group (this time not empty)
+ seq = conn_a.group_sendmsg(self.__message(2), "group",
+ want_answer=True)
+ envs = []
+ for c in conn_b:
+ (msg, env) = c.group_recvmsg(nonblock=False)
+ self.assertEqual(self.__message(2), msg)
+ self.assertEqual(conn_a.lname, env[CC_HEADER_FROM])
+ # The daemon does not mangle the headers. Is it OK?
+ self.assertEqual(CC_TO_WILDCARD, env[CC_HEADER_TO])
+ self.assertEqual("group", env[CC_HEADER_GROUP])
+ self.assertEqual(CC_INSTANCE_WILDCARD, env[CC_HEADER_INSTANCE])
+ self.assertEqual(CC_COMMAND_SEND, env[CC_HEADER_TYPE])
+ self.assertTrue(env[CC_HEADER_WANT_ANSWER])
+ envs.append(env)
+ # Send to non-existing group
+ seq_ne = conn_a.group_sendmsg(self.__message(3), "no-group",
+ want_answer=True)
+ (msg, env) = conn_a.group_recvmsg(nonblock=False, seq=seq_ne)
+ self.assertEqual(self.__no_recpt, msg)
+ self.assertEqual(conn_a.lname, env[CC_HEADER_TO])
+ # Unsubscribe the connections
+ for c in conn_b:
+ c.group_unsubscribe("group")
+ # Synchronize the unsubscriptions
+ self.__barrier(conn_b)
+ seq_ne = conn_a.group_sendmsg(self.__message(4), "group",
+ want_answer=True)
+ (msg, env) = conn_a.group_recvmsg(nonblock=False, seq=seq_ne)
+ self.assertEqual(self.__no_recpt, msg)
+ self.assertEqual(conn_a.lname, env[CC_HEADER_TO])
+ # Send answers for the original message that was delivered
+ lnames = set()
+ for (c, env) in zip(conn_b, envs):
+ c.group_reply(env, self.__message("Reply"))
+ lnames.add(c.lname)
+ # Check the both answers come
+ while lnames:
+ # While there are still connections we didn't get the answer from
+ # (the order is not guaranteed, therefore the juggling with set)
+ (msg, env) = conn_a.group_recvmsg(nonblock=False, seq=seq)
+ self.assertEqual(self.__message("Reply"), msg)
+ lname = env[CC_HEADER_FROM]
+ self.assertTrue(lname in lnames)
+ lnames.remove(lname)
+
+ # Wait a short time, so messages sent where they don't belong get some
+ # time to arrive. They could still be late, so the test could still
+ # pass on a slow machine, but an unreliable test is better than none
+ # at all. Any idea how to do it in a better way?
+ time.sleep(0.1)
+ for c in conn_b:
+ self.assertEqual((None, None), c.group_recvmsg())
+ self.assertEqual((None, None), conn_a.group_recvmsg())
+
+ def test_conn_disconn(self):
"""
- Temporary empty test, to see if we can start and stop msgq.
- Testing the test harness.
+ Keep connecting and disconnecting, checking we can still send
+ and receive messages.
"""
- pass
+ conn = self.__get_connection()
+ conn.group_subscribe("group")
+ for i in range(0, 50):
+ new = self.__get_connection()
+ new.group_subscribe("group")
+ self.__barrier([conn, new])
+ new.group_sendmsg(self.__message(i), "group")
+ (msg, env) = conn.group_recvmsg(nonblock=False)
+ self.assertEqual(self.__message(i), msg)
+ conn.close()
+ conn = new
if __name__ == '__main__':
+ isc.log.init("msgq-tests")
isc.log.resetUnitTestRootLogger()
unittest.main()
More information about the bind10-changes
mailing list