Finalizers etcetera

[Simon Marlow]

> I guess the Hugs documentation should say that using unsafePerformIO
> to call C which calls Haskell should be avoided.

Yes it should, if such behaviour can lead to strange crashes!  Better
still would be to work out exactly what constraints are required on the
use of eval() from primitive operations, and make sure that all the
primitives comply.  Then we can be sure that Haskell finalizers will
work too.

> I think Haskell finalizers are
> different: virtually every finalizer which manipulates Haskell state
> will have race conditions.

This is a separate issue - which I'll wait till I've written the summary
to address.

Cheers,
	Simon