[GHC] #5664: Missing checks for FFI declaration types

[GHC]

#5664: Missing checks for FFI declaration types
---------------------------------+------------------------------------------
    Reporter:  simonmar          |       Owner:  pcapriotti                 
        Type:  bug               |      Status:  patch                      
    Priority:  normal            |   Milestone:  7.6.1                      
   Component:  Compiler (FFI)    |     Version:  7.2.1                      
    Keywords:                    |          Os:  Unknown/Multiple           
Architecture:  Unknown/Multiple  |     Failure:  GHC accepts invalid program
  Difficulty:  Unknown           |    Testcase:                             
   Blockedby:                    |    Blocking:                             
     Related:                    |  
---------------------------------+------------------------------------------

Comment(by simonpj):

 Thanks.  Looks right to me, but it took me a while to be sure.
 Suggestions for comments to make it easier:

  * Needs a `Note [blah]` with an example in `isFFIDynTy`.

  * Comment on `isFFIDynTy` mentions `Addr` but code does not?

  * The comment on `isFFIDynTy` should say what the two argument types are,
 linked to your example.  It took me a while to understand what the
 `curried_res_ty` was doing, until I realised that `arg_tys` was all the
 args ''except'' the `FunPtr`.

  * The comment says that the check is modulo newtypes, but
 `splitTyConApp_maybe` does not look through newtypes, so I thought it was
 wrong.  But it's ok because we have previously called `normaliseFfiType`.
 Worth a comment to point this out, as is done in `TcForeign` line 366, for
 example.

 Please push, and add a test.  Thanks!

 Simon

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5664#comment:6>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler