reactor.callWhenRunning should cause a DirtyReactorError in trial, if the reactor isn't started during the test.

	@tomprince reported
Trac ID	trac#5792
Type	defect
Created	2012-07-18 19:04:16Z

The following code reports an error in test_2 when the error is actually in test_1

from twisted.internet import defer, task
from twisted.internet import reactor
from twisted.trial import unittest

class Test(unittest.TestCase):
    def test_1(self):
        def cb():
            task.LoopingCall(lambda: None).start(1)
        reactor.callWhenRunning(cb)
    def test_2(self):
        d = defer.Deferred()
        reactor.callLater(0.1, d.callback, None)
        return d

trial output:

test
  Test
    test_1 ...                                                             [OK]
    test_2 ...                                                          [ERROR]

===============================================================================
[ERROR]
Traceback (most recent call last):
Failure: twisted.trial.util.DirtyReactorAggregateError: Reactor was unclean.
DelayedCalls: (set twisted.internet.base.DelayedCall.debug = True to debug)
<DelayedCall 0x2b4efc8 [0.898566961288s] called=0 cancelled=0 LoopingCall<1>(<lambda>, *(), **{})()>

test.Test.test_2
-------------------------------------------------------------------------------
Ran 2 tests in 0.107s

FAILED (errors=1, successes=1

trac-id__5792 5792
type__defect defect
reporter__tom_prince tom.prince
priority__normal normal
milestone__ 
branch__ 
branch_author__ 
status__new new
resolution__None None
component__trial trial
keywords__ 
time__1342638256000000 1342638256000000
changetime__1342638256000000 1342638256000000
version__None None
owner__ 
cc__jml

Reproduce the two-test example with Twisted Trial, then inspect reactor.callWhenRunning and Trial's DirtyReactor checks. Trace when the LoopingCall created by test_1 is detected and attributed. Done means an unstarted reactor callback causes DirtyReactorError to identify test_1 rather than test_2.