Byron Cook <[EMAIL PROTECTED]> writes:
> What is your opinion, should this typecheck?
>
> import LazyST
> f = runST ( return 5 )
>
> I do.
> Hugs doesnt:
> *** expression : runST (return 5)
> *** term : return 5
> *** type : Num _3 => ST b _3
> *** does not match : ST c _3
> *** because : types do not match
>
> Is it a bug?
Given that the Haskell report doesn't runST, Hugs is pretty much free to
do what it wants. I believe this is the expected behaviour. That is,
it's what the original runST paper described and therefore what Mark Jones
implemented.
In this case, it's easy enough to workaround - though I can imagine
cases where that wouldn't be true.
Alastair
ps Is there any particular reason not to use the January 1998 release?
pps The story is probably completely different if you use the new Hugs
typechecker in Hugs 1.3c since the new typechecker is supposed to be
more compatible with GHC. (Of course, Hugs 1.3c doesn't include
modules and other features added here at Yale.)
