GADT forall does not round-trip on ghc-8.6

[lspitzner]

data StupidExistential a where
  StupidExistential :: forall a . StupidExistential a

turns into

data StupidExistential a where
  StupidExistential :: a StupidExistential a

(and interestingly into this in the next step, although that really does not matter:)

data StupidExistential a where
  StupidExistential ::a StupidExistential a

This is a bug in ghc-exactprint, confirmed on ghc-exactprint-5.8 and ghc-exactprint-0.6, confirmed on ghc-8.6.5. Does not occur on earlier ghc versions.

Will open an issue on ghc-exactprint. See alanz/ghc-exactprint#79.

Might want a work-around in the meantime.

[tfausak]

The space thing was reported earlier in #217.

[lspitzner]

(This is a bit worse though, because removing forall can break the code.)