[GHC] #8950: Typeable instances for promoted lists and tuples

[GHC]

#8950: Typeable instances for promoted lists and tuples
-------------------------------------+------------------------------------
        Reporter:  kosmikus          |            Owner:
            Type:  bug               |           Status:  new
        Priority:  normal            |        Milestone:
       Component:  libraries/base    |          Version:  7.8.1-rc2
      Resolution:                    |         Keywords:
Operating System:  Unknown/Multiple  |     Architecture:  Unknown/Multiple
 Type of failure:  None/Unknown      |       Difficulty:  Unknown
       Test Case:                    |       Blocked By:
        Blocking:                    |  Related Tickets:
-------------------------------------+------------------------------------

Comment (by goldfire):

 What's the desired behavior here? There are several cases to consider:
 1. No `AutoDeriveTypeable`, no `DataKinds`, `deriving` clause on a
 promotable datatype
 2. No `AutoDeriveTypeable`, no `DataKinds`, standalone deriving on a
 promotable datatype
 3. No `AutoDeriveTypeable`, yes `DataKinds`, `deriving` clause on a
 promotable dataype
 4. No `AutoDeriveTypeable`, yes `DataKinds`, standalone deriving on a
 promotable datatype
 5. Yes `AutoDeriveTypeable`, no `DataKinds`, no explicit `deriving`
 6. Yes `AutoDeriveTypeable`, yes `DataKinds`, no explicit `deriving`
 7. Yes `AutoDeriveTypeable`, no `DataKinds`, explicit `deriving` somewhere
 8. Yes `AutoDeriveTypeable`, yes `DataKinds`, explicit `deriving`
 somewhere

 Currently, 1, 2, and 5 do their thing for the unpromoted datatype. 3, 4,
 and 6 work for both the unpromoted and the promoted datatypes. Curiously,
 7 & 8 don't work! That is, if you say

 {{{
 {-# LANGUAGE AutoDeriveTypeable #-}

 data Foo = Bar
   deriving (Eq)
 }}}

 you '''don't''' get `Typeable` instances. This seems to be a bug. It seems
 to be trying to prevent duplicate instances, but any `deriving` clause
 (even standalone) triggers the filter. (See around line 550 of !TcDeriv.)

 I propose: all 8 derive for both the promoted and unpromoted datatype. I
 can't see a downside to too many `Typeable` instances.

 Thoughts?

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8950#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler