Re: [Haskell-cafe] Compiler's bane

Here is a *much* bigger problem: How do you check that an interpretter is correct?? You can't very easily write a QuickCheck property that will generate every possible valid expression and then check that the output of the interpretter is formally equivilent. The QuickCheck property would be a second copy of the interpretter, which proves nothing! Any hints? Just how *do* you check something large like this? I wonder - how do the GHC developers check that GHC works properly? (I guess by compiling stuff and running it... It's a bit harder to check that a lambda interpretter is working right.)