We are given the following API interface in 8th chapter of “Parallel and Concurrent Programming in Haskell” text,

data Async a = Async (MVar (Either SomeException a))

async :: IO a -> IO (Async a)
async action = do
  var <- newEmptyMVar
  forkIO (do r <- try action; putMVar var r)
  return (Async var)

readMVar :: MVar a -> IO a
readMVar m = do
  a <- takeMVar m 
  putMVar m a
  return a

waitCatch :: Async a -> IO (Either SomeException a)
waitCatch (Async var) = readMVar var

wait :: Async a -> IO a
wait a = do
  r <- waitCatch a
  case r of
    Left e -> throwIO e
    Right a -> return a

Now if we define,

waitEither :: Async a -> Async b -> IO (Either a b)
waitEither a b = do
  m <- newEmptyMVar
  forkIO $ do r <- try (fmap Left (wait a)); putMVar m r
  forkIO $ do r <- try (fmap Right (wait b)); putMVar m r
  wait (Async m)

Couldn’t this lead to an issue where the putMVar of line forkIO $ do r <- try (fmap Right (wait b)); putMVar m r executes before the putMVar present inside readMVar as called by implementation of wait? For simplicity, let us assume that there are no asynchronous exceptions to worry about.

Haddock of official implementation readMVar suggest that it will receive the next call of putMVar, do we want this assumption for our example?