Mutually recursive equality constraints

[MartijnVanSteenbergen]

Given this piece of code:

{-# LANGUAGE TypeFamilies #-}

class C a where
  type T a :: *

f1 :: T a ~ () => a
f1 = f2

f2 :: T a ~ () => a
f2 = f1

GHC complains:

    Couldn't match expected type `T a ~ ()'
           against inferred type `T a1 ~ ()'
    When matching the contexts of the signatures for
      f1 :: forall a. (T a ~ ()) => a
      f2 :: forall a. (T a ~ ()) => a
    The signature contexts in a mutually recursive group should all be identical
    When generalising the type(s) for f1, f2

Is this a bug? Enabling RelaxedPolyRec? fixes the problem. Should TypeFamilies?—just like GADTs—imply RelaxedPolyRec??

[simonpj]

I agree:

Tue Jan 13 16:27:16 GMT 2009  simonpj@microsoft.com
  * Make -XTypeFamilies imply -XRelaxedPolyRec (Trac #2944)

It's moot whether we should merge this, but on balance, yes.

Simon

[igloo]

Merged

[MartijnVanSteenbergen]

Thanks. =)