[Haskell-cafe] Cleaning up threads

[Bas van Dijk]

Don't forget to block asynchronous exception _before_ you fork in:

>        tid <- forkIO (someWorkToDo `finally` putMVar mv ())

Otherwise an asynchronous exception might be thrown to the thread
_before_ the 'putMVar mv ()' exception handler is installed leaving
your main thread in a dead-lock!

You can use the threads library which correctly abstracts over this pattern:

http://hackage.haskell.org/package/threads

Regards,

Bas

On Wed, Sep 15, 2010 at 2:23 AM, Gregory Collins
<greg at gregorycollins.net> wrote:
> Mitar <mmitar at gmail.com> writes:
>
>> Hi!
>>
>> On Tue, Sep 14, 2010 at 11:46 PM, Bas van Dijk <v.dijk.bas at gmail.com> wrote:
>>> Note that killing the main thread will also kill all other threads. See:
>>
>> Yes. But how does those other threads have time to cleanup is my question.
>
> What we do in Snap is this: the master thread has a catch handler which
> catches the AsyncException generated by the call to killThread. When we
> get this, we instruct any service loop threads to exit, and they all
> wait for service threads to terminate (currently by sleep-polling a
> connections table, which I should probably fix...). Then the master
> thread exits by just returning.
>
> Note that I think the "main thread being killed kills all threads" issue
> can be circumvented by using a little gadget like this:
>
> ------------------------------------------------------------------------
>    someWorkToDo :: IO ()
>    someWorkToDo = someStuff `catch` cleanupHandler
>
>    main :: IO ()
>    main = do
>        mv  <- newEmptyMVar
>        tid <- forkIO (someWorkToDo `finally` putMVar mv ())
>
>        -- wait on thread to finish; any exception here is probably an
>        -- AsyncException, so kill the someWorkToDo master thread
>        -- yourself and wait on the mvar again
>
>        takeMVar mv `catch` \(e::SomeException) -> do
>            killThread tid
>            takeMVar mv
> ------------------------------------------------------------------------
>
> At least, this is what we do in our webserver, and it seems to work
> fine -- users complain about the delay involved in our slow cleanup
> handler when they ctrl-c the server. :)
>
> G
> --
> Gregory Collins <greg at gregorycollins.net>
>