问题
(Title was: "How to write a unit test for a DBUS service written in Python?")
I've started to write a DBUS service using dbus-python, but I'm having trouble writing a test case for it.
Here is an example of the test I am trying to create. Notice that I have put a GLib event loop in the setUp(), this is where the problem hits:
import unittest
import gobject
import dbus
import dbus.service
import dbus.glib
class MyDBUSService(dbus.service.Object):
def __init__(self):
bus_name = dbus.service.BusName('test.helloservice', bus = dbus.SessionBus())
dbus.service.Object.__init__(self, bus_name, '/test/helloservice')
@dbus.service.method('test.helloservice')
def hello(self):
return "Hello World!"
class BaseTestCase(unittest.TestCase):
def setUp(self):
myservice = MyDBUSService()
loop = gobject.MainLoop()
loop.run()
# === Test blocks here ===
def testHelloService(self):
bus = dbus.SessionBus()
helloservice = bus.get_object('test.helloservice', '/test/helloservice')
hello = helloservice.get_dbus_method('hello', 'test.helloservice')
assert hello() == "Hello World!"
if __name__ == '__main__':
unittest.main()
My problem is that the DBUS implementation requires you to start an event loop so that it can start dispatching events. The common approach is to use GLib's gobject.MainLoop().start() (although I'm not married to this approach, if someone has a better suggestion). If you don't start an event loop, the service still blocks, and you also cannot query it.
If I start my service in the test, the event loop blocks the test from completing. I know the service is working because I can query the service externally using the qdbus tool, but I can't automate this inside the test that starts it.
I'm considering doing some kind of process forking inside the test to handle this, but I was hoping someone might have a neater solution, or at least a good starting place for how I would write a test like this.
回答1:
With some help from Ali A's post, I have managed to solve my problem. The blocking event loop needed to be launched into a separate process, so that it can listen for events without blocking the test.
Please be aware my question title contained some incorrect terminology, I was trying to write a functional test, as opposed to a unit test. I was aware of the distinction, but didn't realise my mistake until later.
I've adjusted the example in my question. It loosely resembles the "test_pidavim.py" example, but uses an import for "dbus.glib" to handle the glib loop dependencies instead of coding in all the DBusGMainLoop stuff:
import unittest
import os
import sys
import subprocess
import time
import dbus
import dbus.service
import dbus.glib
import gobject
class MyDBUSService(dbus.service.Object):
def __init__(self):
bus_name = dbus.service.BusName('test.helloservice', bus = dbus.SessionBus())
dbus.service.Object.__init__(self, bus_name, '/test/helloservice')
def listen(self):
loop = gobject.MainLoop()
loop.run()
@dbus.service.method('test.helloservice')
def hello(self):
return "Hello World!"
class BaseTestCase(unittest.TestCase):
def setUp(self):
env = os.environ.copy()
self.p = subprocess.Popen(['python', './dbus_practice.py', 'server'], env=env)
# Wait for the service to become available
time.sleep(1)
assert self.p.stdout == None
assert self.p.stderr == None
def testHelloService(self):
bus = dbus.SessionBus()
helloservice = bus.get_object('test.helloservice', '/test/helloservice')
hello = helloservice.get_dbus_method('hello', 'test.helloservice')
assert hello() == "Hello World!"
def tearDown(self):
# terminate() not supported in Python 2.5
#self.p.terminate()
os.kill(self.p.pid, 15)
if __name__ == '__main__':
arg = ""
if len(sys.argv) > 1:
arg = sys.argv[1]
if arg == "server":
myservice = MyDBUSService()
myservice.listen()
else:
unittest.main()
回答2:
Simple solution: don't unit test through dbus.
Instead write your unit tests to call your methods directly. That fits in more naturally with the nature of unit tests.
You might also want some automated integration tests, that check running through dbus, but they don't need to be so complete, nor run in isolation. You can have setup that starts a real instance of your server, in a separate process.
回答3:
I might be a bit out of my league here, since I don't know python and only somewhat understand what this magical "dbus" is, but if I understand correctly, it requires you to create a rather unusual testing environment with runloops, extended setup/teardown, and so on.
The answer to your problem is to use mocking. Create an abstract class which defines your interface, and then build an object from that to use in your actual code. For the purposes of testing, you build a mock object communicates through that same interface, but has behavior which you would define for the purposes of testing. You can use this approach to "simulate" the dbus object running through an event loop, doing some work, etc., and then simply concentrate on testing how your class ought to react to the result of the "work" done by that object.
回答4:
You just need to make sure you are handling your main loop properly.
def refresh_ui():
while gtk.events_pending():
gtk.main_iteration_do(False)
This will run the gtk main loop until it has finished processing everything, rather than just run it and block.
For a complete example of it in practise, unit testing a dbus interface, go here: http://pida.co.uk/trac/browser/pida/editors/vim/test_pidavim.py
回答5:
You could also start the mainloop in a separate thread very simply inside your setUp method.
Something like this:
import threading
class BaseTestCase(unittest.TestCase):
def setUp(self):
myservice = MyDBUSService()
self.loop = gobject.MainLoop()
threading.Thread(name='glib mainloop', target=self.loop.run)
def tearDown(self):
self.loop.quit()
回答6:
Check out python-dbusmock library.
It hides the ugly subprocess logic behind your eyes, so you don't have to worry about it in your tests.
来源：https://stackoverflow.com/questions/510821/how-to-write-a-functional-test-for-a-dbus-service-written-in-python