GitLab

This erroneous program is rejected with a rather strange error message:

{-# LANGUAGE ScopedTypeVariables #-}
module Bug where

instance forall c. c

$ /opt/ghc/8.6.3/bin/ghc Bug.hs
[1 of 1] Compiling Bug              ( Bug.hs, Bug.o )

Bug.hs:4:20: error: Not in scope: type variable ‘c’
  |
4 | instance forall c. c
  |                    ^

This error message is hogwash, since c is absolutely in scope. The reason GHC believes that c is out of scope is because before GHC renamed the type signature in the instance, it performs a pass over all top-level binders in RnNames.getLocalNonValBinders to obtain their Names. In particular, this is the culprit:

    new_assoc overload_ok (L _ (ClsInstD _ (ClsInstDecl { cid_poly_ty = inst_ty
                                                      , cid_datafam_insts = adts })))
      | Just (L loc cls_rdr) <- getLHsInstDeclClass_maybe inst_ty
      = do { cls_nm <- setSrcSpan loc $ lookupGlobalOccRn cls_rdr
           ; (avails, fldss)
                    <- mapAndUnzipM (new_loc_di overload_ok (Just cls_nm)) adts
           ; return (avails, concat fldss) }
      | otherwise
      = return ([], [])    -- Do not crash on ill-formed instances
                           -- Eg   instance !Show Int   Trac #3811c

Notice the use of lookupGlobalOccRn there. Since this code just tunnels into the instance type signature (using getLHsInstDeclClass_maybe) without binding any type variables, this calls lookupGlobalOcc on an unbound type variable c. Eek.

I believe the fix would be to use lookupGlobalOccRn_maybe instead and simply default to return ([], []) in the event that lookupGlobalOccRn_maybe returns Nothing.

Trac metadata

Trac field	Value
Version	8.6.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture