Do not consult -XGADTs flag when pattern matching on GADTs

See Trac #2004, and Note [Flags and equational constraints] in TcPat.

diff --git a/compiler/typecheck/TcPat.lhs b/compiler/typecheck/TcPat.lhs
index f9e7127..defe6fb 100644 (file)
--- a/compiler/typecheck/TcPat.lhs
+++ b/compiler/typecheck/TcPat.lhs
@@ -867,11 +867,8 @@ refineAlt con pstate ex_tvs [] pat_ty
   = return pstate      -- Common case: no equational constraints
 
 refineAlt con pstate ex_tvs co_vars pat_ty
-  = do { opt_gadt <- doptM Opt_GADTs   -- No type-refinement unless GADTs are on
-       ; if (not opt_gadt) then return pstate
-         else do 
-
-       { checkTc (isRigidTy pat_ty) (nonRigidMatch con)
+  = -- See Note [Flags and equational constraints]
+    do { checkTc (isRigidTy pat_ty) (nonRigidMatch con)
        -- We are matching against a GADT constructor with non-trivial
        -- constraints, but pattern type is wobbly.  For now we fail.
        -- We can make sense of this, however:
@@ -899,9 +896,20 @@ refineAlt con pstate ex_tvs co_vars pat_ty
                                vcat [ ppr con <+> ppr ex_tvs,
                                       ppr [(v, tyVarKind v) | v <- co_vars],
                                       ppr reft]
-       } } }
+       } }
 \end{code}
 
+Note [Flags and equational constraints]
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+If there are equational constraints, we take account of them
+regardless of flag settings; -XGADTs etc applies only to the
+*definition* of a data type.
+
+An alternative would be also to reject a program that *used*
+constructors with equational constraints.  But want we should avoid at
+all costs is simply to *ignore* the constraints, since that gives
+incomprehensible errors (Trac #2004).
+
 
 %************************************************************************
 %*                                                                     *