On Thu, Feb 26, 2015 at 9:14 PM, Roman Cheplyaka <roma@ro-che.info> wrote:

I don't think frozen-base by itself is enough to achieve the goal of
compiling n-year-old code without modifications.

I second Joachim's concerns about stability. While not completely au fait with the technicalities of frozen-base, it does seem to me that something more is needed.

Just a couple of examples that jump to mind:

- the recent change to require extensions for inferred types

I did a search on ghc tickets and while a few seems to fit the bill, I don't know what this refers to.

- AMP (e.g. your code had a monad defined without an applicative instance)

As a first pass, punt on this?

-- Kim-Ee