Is this a bug in multithreading code?

[Remi Turk]

On Wed, Apr 23, 2003 at 03:51:25PM +0200, Nick Name wrote:
> Hi all, does someone know why this program works
> 
> ------------------------
> import Concurrent
> 
> loopM = sequence_ . repeat
> 
> main = do
>        v <- newMVar 0
>        forkIO (loopM (modifyMVar_ v (return . (1+))))
> --       forkIO (loopM (modifyMVar_ v (return . (1+))))
>        loopM (readMVar v >>= print)
> -------------------------
> 
> but uncommenting the commented line makes the program fail with "stack 
> overflow" ?
> 
> Thanks for help
> 
> Vincenzo

It definitely seems a bug to me. Though I'm not sure if it's in your
code or in Concurrent ;)

However, the following code _does_ work, which makes me suspect
it has something to do with the locking done inside of an MVar.

import Control.Concurrent

loopM = sequence_ . repeat

main = do
        v <- newMVar 0

        forkIO $ loopM $ do
            modifyMVar_ v $ return . (1+)

        forkIO $ loopM $ do
            modifyMVar_ v $ return . (1+)

        loopM $ do
            readMVar v >>= print
            --withMVar v print

Swapping the last 2 lines (that is, their comments) makes it crash again.

And I guess this actually _is_ the code you should use:
withMVar is a "safe wrapper for operating on the contents of an MVar",
which is exactly what you're doing.

Does anyone else know what exactly is going on?

Happy hacking,
Remi

-- 
Nobody can be exactly like me. Even I have trouble doing it.