On Fri, May 8, 2015 at 7:40 PM, Donn Cave wrote:

But fatal if compilation is conditional on something that affects the ability to type check, am I right? Such as different compilers or versions of same compiler.

Not per the abstract (paper itself seems to be paywalled). They had an earlier work with that issue, the linked one is about how to be robust in the face of such conditionals. -- brandon s allbery kf8nh sine nomine associates allbery.b@gmail.com ballbery@sinenomine.net unix, openafs, kerberos, infrastructure, xmonad http://sinenomine.net

Hi Wren,

Incidentally, if we really want to pursue the "get rid of CPP by building it into the GHC distro"...

In recent years there've been a number of papers on "variational lambda-calculi"[1] which essentially serve to embed flag-based preprocessor conditionals directly into the language itself. One major benefit of this approach is that the compiler can then typecheck *all* variations of the code, rather than only checking whichever particular variation we happen to be compiling at the time. This is extremely useful for avoiding bitrot in the preprocessor conditionals.

...If we were to try and obviate the dependency on CPP, variational typing seems like a far more solid approach than simply reinventing the preprocessing wheel yet again. (The downside, of course, is making the Haskell spec significantly more complex.)

I think even more beneficial than type checking all cases is the easier support for any Haskell tooling operating with the Haskell source if all cases are part of the AST. Greetings, Daniel