GitLab

In fixing the ambiguity check I came across this tricky program

data T a b where
  MkT :: (a~Maybe b) => a -> Maybe b -> T a b

f :: T a a -> Bool
f (MkT x y) = [x,y] `seq` True

We get an error message

Frozen2.hs:12:18:
    Could not deduce (a ~ Maybe a)
    from the context (a ~ Maybe a)
      bound by a pattern with constructor
                 MkT :: forall a b. a ~ Maybe b => a -> Maybe b -> T a b,
               in an equation for `f'
      at Frozen2.hs:12:4-10
      `a' is a rigid type variable bound by
          the type signature for f :: T a a -> Bool at Frozen2.hs:11:6
    Relevant bindings include
      f :: T a a -> Bool (bound at Frozen2.hs:12:1)
      x :: a (bound at Frozen2.hs:12:8)
      y :: Maybe a (bound at Frozen2.hs:12:10)
    In the expression: y
    In the first argument of `seq', namely `[x, y]'
    In the expression: [x, y] `seq` True

This error message is nonsense! It arises becuase the "insolubles" get both a "given" insoluble (a~T a) and a "wanted" insoluble with the same type.

This can also arise, rather more easily, with the new ambiguity check, via an inaccessible context

foo :: forall a. (a ~ T a) => a -> a 

It's a bit obscure, but it needs fixing.

Trac metadata

Trac field	Value
Version	7.6.1
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture