Re: [Haskell-cafe] Printing of asynchronous exceptions to stderr

On 10/11/2010 14:32, Bas van Dijk wrote:

On Wed, Nov 10, 2010 at 2:39 PM, Mitar wrote:

...

...

Strange. It would help if you could show more of of your code.

I am attaching a sample program which shows this. I am using 6.12.3 on both Linux and Mac OS X. And I run this program with runhaskell Test.hs. Without "throwIO ThreadKilled" it outputs:

Test.hs: MyTerminateException MVar was successfully taken

With "throwIO ThreadKilled" is as expected, just:

MVar was successfully taken

So MVar is filled. What means that thread gets exception after that. But there is nothing after that. ;-) (At least nothing visible.)

This is really interesting. Presumably what happens is that an exception is indeed thrown and then raised in the thread after the final action. Now if you synchronously throw an exception at the end it looks like it's raised before the asynchronous exception is raised.

Hopefully one of the GHC devs (probably Simon Marlow) can confirm this behavior and shed some more light on it.

I think it's behaving as expected - there's a short window during which exceptions are unblocked and a second exception can be thrown. The program has let run = doSomething `catches` [ Handler (\(_ :: MyTerminateException) -> return ()), Handler (\(e :: SomeException) -> putStrLn $ "Exception: " ++ show e) ] `finally` (putMVar terminated ()) nid <- forkIO run The first MyTerminateException gets handled by the first exception handler. This handler returns, and at that point exceptions are unblocked again, so the second MyTerminateException can be thrown. The putMVar gets to run, and then the exception is re-thrown by finally, and caught and printed by the outer exception handler. The right way to fix it is like this: let run = unblock doSomething `catches` [ Handler (\(_ :: MyTerminateException) -> return ()), Handler (\(e :: SomeException) -> putStrLn $ "Exception: " ++ show e) ] `finally` (putMVar terminated ()) nid <- block $ forkIO run and the same will be true in GHC 7.0, except you'll need to use mask instead of block/unblock. Cheers, Simon