Because GHC does not respect let-should-not-be-generalized in the
#16146: Trivial partial type signature kills type inference in the presence of GADTs -------------------------------------+------------------------------------- Reporter: goldfire | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.6.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by simonpj): '''Let-must-not-be-generalised''' behaviour presence of partial type signatures, this is rejected again. It's hard to see how GHC ''could'' respect let-should-not-be-generalised. Consider {{{ f :: forall a. _ -> a f = e }}} How could we check this signature without generalising the type of `e`? I suppose there could be a special case if there is no user-written `forall`. So `f :: forall. blah` means "please generalise` and `f :: blah` means "don't generalise". But that's in conflict with some of the choices under "forall-or-nothing" above. Regardless, it'd be good to document the behaviour in the manual. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/16146#comment:12 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler