Minimal support for 9.10

[Ericson2314]

PR #502 contains some other orthogonal cleanups.

[Ericson2314]

src/Reflex/PerformEvent/Base.hs:76:10: error: [GHC-39999]
Error:     • Could not deduce ‘Monad (HostFrame t)’
        arising from the superclasses of an instance declaration
      from the context: (ReflexHost t, Ref m ~ Ref IO)
        bound by the instance declaration
        at src/Reflex/PerformEvent/Base.hs:76:10-77
      Possible fix:
        If the constraint looks soluble from a superclass of the instance context,
        read 'Undecidable instances and loopy superclasses' in the user manual
    • In the instance declaration for
        ‘PerformEvent t (PerformEventT t m)’
   |
76 | instance (ReflexHost t, Ref m ~ Ref IO) => PerformEvent t (PerformEventT t m) where
   |          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Note that ReflexClass looks like this

class ( Reflex t
      , MonadReflexCreateTrigger t (HostFrame t)
      , MonadSample t (HostFrame t)
      , MonadHold t (HostFrame t)
      , MonadFix (HostFrame t)
      , MonadSubscribeEvent t (HostFrame t)
      ) => ReflexHost t where

I see @ymeister fixed the build in #502, but I don't get why 9.10 is no longer accepting constraints as written.

[ymeister]

src/Reflex/PerformEvent/Base.hs:76:10: error: [GHC-39999]
Error:     • Could not deduce ‘Monad (HostFrame t)’
        arising from the superclasses of an instance declaration
      from the context: (ReflexHost t, Ref m ~ Ref IO)
        bound by the instance declaration
        at src/Reflex/PerformEvent/Base.hs:76:10-77
      Possible fix:
        If the constraint looks soluble from a superclass of the instance context,
        read 'Undecidable instances and loopy superclasses' in the user manual
    • In the instance declaration for
        ‘PerformEvent t (PerformEventT t m)’
   |
76 | instance (ReflexHost t, Ref m ~ Ref IO) => PerformEvent t (PerformEventT t m) where
   |          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Note that ReflexClass looks like this

class ( Reflex t
      , MonadReflexCreateTrigger t (HostFrame t)
      , MonadSample t (HostFrame t)
      , MonadHold t (HostFrame t)
      , MonadFix (HostFrame t)
      , MonadSubscribeEvent t (HostFrame t)
      ) => ReflexHost t where

I see @ymeister fixed the build in #502, but I don't get why 9.10 is no longer accepting constraints as written.

It might be related to this - https://github.com/ghc-proposals/ghc-proposals/blob/master/proposals/0287-simplify-subsumption.rst

Maybe 9.10 made things stricter.

GHC 9.0 made a switch to a type inference scheme for higher rank polymorphism called simplified subsumption. Consquently, expressions like g f in the below, which passed the type checker in the GHC 8 series, stopped passing the type checker in the GHC 9 series.

{-# LANGUAGE RankNTypes #-}

f :: Int -> (forall a. a -> a)
f _ = id

g :: (forall a. Int -> a -> a) -> ()
g _ = ()

main :: IO ()
main = print (g f)

[maralorn]

I don't think anything related to simplified subsumption changed from 9.8 to 9.10. At least the big breaking change was much earlier.

[Ericson2314]

https://gitlab.haskell.org/ghc/ghc/-/issues/25318 I made a GHC issue on this for hopefully better visibility to get to the bottom of it.

[Ericson2314]

See the GHC issue I linked, this typechecker change is intentional

If the constraint looks soluble from a superclass of the instance context,
read 'Undecidable instances and loopy superclasses' in the user manual

and 🤦 that bit in the type error did indeed tell me exactly what to do.