No skolem info: b_azg[sk]

I was testing this code, which is from our ICFP paper on Coercible:

{-# LANGUAGE TypeFamilies, GeneralizedNewtypeDeriving, MultiParamTypeClasses, FlexibleInstances, AllowAmbiguousTypes #-}

newtype Id1 a = MkId1 a
newtype Id2 a = MkId2 (Id1 a) deriving (UnsafeCast b)

type family Discern a b
type instance Discern (Id1 a) b = a
type instance Discern (Id2 a) b = b

class UnsafeCast to from where
  unsafe :: from -> Discern from to

instance UnsafeCast b (Id1 a) where
  unsafe (MkId1 x) = x

unsafeCoerce :: a -> b
unsafeCoerce x = unsafe (MkId2 (MkId1 x))

without AllowAmbiguousTypes I get

UnsafeCast.hs:11:3: error:
    Couldn't match type ‘Discern from to0’ with ‘Discern from to’
    NB: ‘Discern’ is a type function, and may not be injective
    The type variable ‘to0’ is ambiguous
    Expected type: from -> Discern from to
      Actual type: from -> Discern from to0
    In the ambiguity check for the type signature for ‘unsafe’:
      unsafe :: forall to from.
                UnsafeCast to from =>
                from -> Discern from to
    To defer the ambiguity check to use sites, enable AllowAmbiguousTypes
    When checking the class method:
      unsafe :: forall to from.
                UnsafeCast to from =>
                from -> Discern from to
    In the class declaration for ‘UnsafeCast’

(is that a bug? I feel like it could be, but I’m intimidated by the error message).

So I put in the suggested pragma, and now I get

UnsafeCast.hs:4:41: error:ghc-stage2: panic! (the 'impossible' happened)
  (GHC version 7.11.20151111 for x86_64-unknown-linux):
	No skolem info: b_azg[sk]

Please report this as a GHC bug:  http://www.haskell.org/ghc/reportabug

This is GHC-almost-HEAD (changeset:2f6e87/ghc). I’ll start a rebuild with head and see what has changed.

Trac metadata

Trac field	Value
Version	7.11
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

changed weight to 5

added Tbug Trac import labels

Other tickets with this error message: #10946 (closed), #10045 (closed)

Trac metadata

Trac field	Value
Related	- → #10045 (closed), #10946 (closed)

Still in todays’s HEAD.

mentioned in commit 1a8b752d

I'm on this.

Done. I put the wrong ticket number in the commit, but this is it

commit 02c1c5735aff0cce2b04a6b3e4732d62bb0a4f3c
Author: Simon Peyton Jones <simonpj@microsoft.com>
Date:   Wed Jan 6 17:22:02 2016 +0000

    Use an Implication in 'deriving' error
    
    Trac #11437 showed that erroneous constraints from a 'deriving'
    clause need to be wrapped in an Implication to properly scope
    their skolems.
    
    The main change is in TcDeriv.simplifyDeriv; the call to
    buildImplicationFor is new.


>---------------------------------------------------------------

02c1c5735aff0cce2b04a6b3e4732d62bb0a4f3c
 compiler/typecheck/TcDeriv.hs                      | 23 +++++++++++++++------
 compiler/typecheck/TcInstDcls.hs                   |  2 +-
 compiler/typecheck/TcPatSyn.hs                     |  2 +-
 compiler/typecheck/TcRnTypes.hs                    | 11 +++++++++-
 .../tests/deriving/should_compile/T10561.stderr    |  8 +++++---
 testsuite/tests/deriving/should_fail/T7148.stderr  | 24 +++++++++++++---------
 .../tests/typecheck/should_fail/T11347.stderr      | 13 ++++++++++--

Trac metadata

Trac field	Value
Test case	- → typecheck/should_fail/T11347

added 1 deleted label

changed milestone to %8.0.1

closed

Merged as 6c956b6d.

Trac metadata

Trac field	Value
Resolution	Unresolved → ResolvedFixed

removed 1 deleted label

added Pnormal label