[GHC] #7633: Checkable "minimal complete definitions"

[GHC]

#7633: Checkable "minimal complete definitions"
---------------------------------+------------------------------------------
    Reporter:  shachaf           |       Owner:                  
        Type:  feature request   |      Status:  patch           
    Priority:  normal            |   Milestone:  7.8.1           
   Component:  Compiler          |     Version:  7.6.1           
    Keywords:                    |          Os:  Unknown/Multiple
Architecture:  Unknown/Multiple  |     Failure:  None/Unknown    
  Difficulty:  Unknown           |    Testcase:                  
   Blockedby:                    |    Blocking:                  
     Related:  6028              |  
---------------------------------+------------------------------------------

Comment(by simonpj):

 twanl: your points are valid, but I still vote for putting it in `Sig`.
 The `SPECIALISE instance` pragmas for an instance decl are in `Sig`, for
 example (`SpecInstSig` constructor).  And `GenericSig` is valid only in a
 class decl, and nowhere else.  It's true that, as a result, the `Sig` data
 type isn't as precise as it could be, but lumping them together reduces
 the number of fields, arguments, and plumbing.  A judgement call, I agree,
 but doing it this way would be consistent with what we have.

 Once we get all the way to a `Class`, then indeed we need a special
 purpose field.  I'm only talking about `HsSyn`.

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/7633#comment:21>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler