Order of arguments to test functions is awkward

[barrucadu]

dejafu :: Show a => (forall t. ConcST t a) -> (String, Predicate a) -> IO Bool
dejafus :: Show a => (forall t. ConcST t a) -> [(String, Predicate a)] -> IO Bool
dejafuWay :: Show a => Way -> MemType -> (forall t. ConcST t a) -> (String, Predicate a) -> IO Bool
dejafusWay :: Show a => Way -> MemType -> (forall t. ConcST t a) -> [(String, Predicate a)] -> IO Bool
dejafuDiscard :: Show a => (Either Failure a -> Maybe Discard) -> Way -> MemType -> (forall t. ConcST t a) -> (String, Predicate a) -> IO Bool
-- etc

Currently, the action to test comes somewhere in the middle of the arguments to the etst functions. This pattern is followed in the hunit-dejafu and tasty-dejafu packages. But as I've found in dejafu-tests, it's kind of a pain, and I define wrapper functions so I can do this:

dejafu/dejafu-tests/Cases/MultiThreaded.hs

Lines 24 to 39 in 31224ba

	threadingTests :: [Test]
	threadingTests = toTestList
	[ djfuT "Fork reports the thread ID of the child" (gives' [True]) $ do
	var <- newEmptyMVar
	tid <- fork $ myThreadId >>= putMVar var
	(tid ==) <$> readMVar var
	, djfuT "Different threads have different thread IDs" (gives' [True]) $ do
	tid <- spawn myThreadId
	(/=) <$> myThreadId <*> readMVar tid
	, djfuT "A thread doesn't wait for its children before terminating" (gives' [Nothing, Just ()]) $ do
	x <- newCRef Nothing
	_ <- fork . writeCRef x $ Just ()
	readCRef x
	]

The action to test should go last.

[barrucadu]

Implemented in next-supermajor branch.