#8707: Kind inference fails in data instance definition -------------------------------------+------------------------------------ Reporter: goldfire | Owner: jstolarek Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.7 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by jstolarek):
2) The error message is not particularly helpful. GHC 7.6.3 had a more verbose, but more helpful message.
Code responsible for this difference is here [[GhcFile(compiler/types/TypeRep.lhs#L711)]] in the function `pprTcApp`: {{{ | Just dc <- isPromotedDataCon_maybe tc , let dc_tc = dataConTyCon dc , isTupleTyCon dc_tc , let arity = tyConArity dc_tc -- E.g. 3 for (,,) k1 k2 k3 t1 t2 t3 ty_args = drop arity tys -- Drop the kind args , ty_args `lengthIs` arity -- Result is saturated = pprPromotionQuote tc <> (tupleParens (tupleTyConSort dc_tc) $ sep (punctuate comma (map (pp TopPrec) tys))) }}} The only question is how we want the arguments printed? GHC 7.6 prints them like this: {{{ Data constructor `SFalse' returns type `SingDF Bool k ('(,) Bool (k -> *) 'False (Ctor k))' instead of an instance of its parent type `SingDF Bool Bool a' }}} GHC HEAD prints tuple constructors in the standard notation (not prefix as seen in GHC 7.6 output): {{{ Data constructor ‘SFalse’ returns type ‘SingDF Bool k '(Bool, k -> *, 'False, Ctor k)’ instead of an instance of its parent type ‘SingDF Bool Bool a’ }}} But printing kinds as part of a tuple is misleading. Any suggestions how should the output look like? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8707#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler