Ross Paterson writes:

...

Under Hugs, fromInteger (and thus fromIntegral) throws an exception if the argument is outside the representable range of values, while GHC just truncates.

I would say that GHC implements a performance optimization by dropping the check and that the behaviour outside the range of Int should be treated as undefined (which is not the same as 'implementation defined').

If I remember correctly (this point has come up several times in the past), the reasoning is something like this: since Int arithmetic doesn't do any overflow checking, it's inconsistent for fromInteger to do any. Q: should (0xffffffff :: Int32) == (-1 :: Int32) we take the view that it's useful if this holds. So fromInteger on *sized* integral types doesn't do any overflow checking. Now, if a, b :: Int, should (a + b) :: Int == fromInteger (toInteger a + toInteger b) :: Int (eg. take a to be maxBound :: Int, and b to be 1, and this doesn't hold in Hugs). We take the view that this should hold, i.e. it makes no difference if the addition is done with Int or Integer because the final result is modulo 2^n (where n is the number of bits in Int). If you were to do overflow checking on Int addition, then it would also make sense to do overflow checking on fromInteger, and the above property would still hold, but to do one but not the other is inconsistent. Cheers, Simon