Don't reuse coercion holes as covars when deferring errorswip/covar-holes

Previously TcErrors would use the variable associated with a coercion
hole as a normal coercion variable when deferring type errors. This
resulted in a Core Lint error now since we distinguish coercion holes
from "vanilla" covars via their IdDetails.

1 files changed, 7 insertions, 1 deletions

diff --git a/compiler/typecheck/TcErrors.hs b/compiler/typecheck/TcErrors.hs
index 02b888703d..c0343bfa15 100644
--- a/compiler/typecheck/TcErrors.hs
+++ b/compiler/typecheck/TcErrors.hs
@@ -898,7 +898,9 @@ addDeferredBinding ctxt err ct
              -> addTcEvBind ev_binds_var $ mkWantedEvBind evar err_tm
            HoleDest hole
              -> do { -- See Note [Deferred errors for coercion holes]
-                     let co_var = coHoleCoVar hole
+                     uniq <- newUnique
+                   ; let co_var = mkLocalCoVar (mkSystemVarName uniq (fsLit "deferred_co"))
+                                               (varType (coHoleCoVar hole))
                    ; addTcEvBind ev_binds_var $ mkWantedEvBind co_var err_tm
                    ; fillCoercionHole hole (mkTcCoVarCo co_var) }}
 
@@ -1029,6 +1031,10 @@ from that EvVar, filling the hole with that coercion. Because coercions'
 types are unlifted, the error is guaranteed to be hit before we get to the
 coercion.
 
+We make a new variable for the coercion so that its IdDetails say that it's
+a CoVar, not a CoercionHole. (This step is very roughly like skolemisation
+of TcTyVars.)
+
 Note [Do not report derived but soluble errors]
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 The wc_simples include Derived constraints that have not been solved,