> The current behaviour of finalizers in GHC is a bit strange:
>
> -- Strange.hs --------------------------------------------------------
> import Foreign
> import System.Environment
>
> useMem :: IO ()
> useMem = if length (fib 27) == 0 then return () else return ()
> where fib :: Int -> [Int]
> fib 0 = [1]
> fib 1 = [1]
> fib n = fib (n-1) ++ fib (n-2)
>
> main :: IO ()
> main = do
> args <- getArgs
> ptr <- malloc
> newForeignPtr (ptr :: Ptr Int32) (putStrLn "Finalising...")
> case args of
> ["1"] -> return ()
> ["2"] -> putStrLn "Created"
> ["3"] -> do useMem ; putStrLn "Created"
> _ -> putStrLn "unknown test"
> ----------------------------------------------------------------------
> panne@jeanluc:~> ghc -O -Wall -o Strange Strange.hs
> panne@jeanluc:~> ./Strange 1
> Finalising...
> panne@jeanluc:~> ./Strange 2
> Created
> panne@jeanluc:~> ./Strange 3
> Finalising...
> Created
> ----------------------------------------------------------------------
>
> The first and third runs are fine, but why is the finalizer
> not invoked on the second run?
I think what's going on here is that the finalizer for stdout is running
before your finalizer, and since stdout has already been closed it can't
print out the message :-)
In the first test, stdout hasn't been created at all (it isn't created
until the first use), and in the third test your finalizer gets a chance
to run before the end of the program. You could also do 'performGC;
yield' to get the same effect.
> The thing I wanted to check in the first place was: Are
> foreign objects
> really kept alive if they can only be reached via a callback made with
> 'foreign import "wrapper"'? They definitely should be, but I couldn't
> verify this, given the current state of affairs... :-(
Yes, they should be. The wrapper closure is registered as a StablePtr.
Cheers,
Simon
_______________________________________________
Cvs-ghc mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/cvs-ghc
