#11520: GHC falls into a hole if given incorrect kind signature -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: bug | Status: new Priority: normal | Milestone: 8.0.1 Component: Compiler (Type | Version: 8.0.1-rc1 checker) | Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Incorrect | Unknown/Multiple warning at compile-time | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by simonpj): The trouble here is this: * The kind error is indeed found and reported (i.e. collected in the bag of errors in the monad) * But we continue anyway into `checkValidInstance` * The latter calls `fvProv` to check for something like the size of instance * Which fails because the unsolved constraint has left a hole in the type Really we should not even attempt `checkValidInstance` if we fail to kind- check the type. We could use a `checkNoErrs` in `TcHsTye.TcHsClsInstType`. But there are many other uses of `solveEqualities` all of which in principle have the same concern. So I thought about adding `checkNoErrs` to `solveEqualities`. But there I see {{{ -- | Type-check a thing that emits only equality constraints, then -- solve those constraints. Emits errors -- but does not fail -- -- if there is trouble. solveEqualities :: TcM a -> TcM a }}} '''Richard''': why did you want the "...but does not fail..." bit? What goes wrong if it fais on error? In effect it always used to! -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11520#comment:4 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler