Standalone deriving doesn't do newtype deriving properly

[ryani]

I'd expect the following two declarations to have identical behavior:

newtype S1 x y a = S1 (M S x y a) deriving IxMonad

newtype S2 x y a = S2 (M S x y a)
deriving instance IxMonad S2

Results:

$ ghc-6.11.20081005 -c deriving.hs

deriving.hs:15:0:
    Can't make a derived instance of `IxMonad S2'
      (even with cunning newtype deriving:
       the representation type has wrong kind)
    In the stand-alone deriving instance for `IxMonad S2'

[ajd]

When I comment out the standalone stuff, GHC gives the message

The equation(s) for 'ret' have one argument,
but its type 'M s i i a' has none
In the definition for the method 'ret'
In the instance declaration for IxMonad (M s)

which seems kind of similar...

I'm using 6.8.2 on Windows.

[ryani]

Oops, that was just a typo in my test case.

Change the instance to declare ret as

   ret :: a -> m i i a

It compiles fine with the standalone commented out in GHC6.8.3 but gives the same error with it in.

I'll upload a fixed attachment.

[simonpj]

Excellent report. That's a bug. I'll get to it.

Simon

[simonpj]

Fixed by

Tue Oct 21 07:29:22 PDT 2008  simonpj@microsoft.com
  * Fix Trac #2668, and refactor TcDeriv

Pls merge

[igloo]

Merged