Re: [GHC] #8425: ghc-7.6.3: crossmodule inline leads to buggy code (-O2)

10 Oct 2013

      #8425: ghc-7.6.3: crossmodule inline leads to buggy code (-O2)
-------------------------------------+----------------------------------
        Reporter:  slyfox            |            Owner:
            Type:  bug               |           Status:  new
        Priority:  normal            |        Milestone:
       Component:  Compiler          |          Version:  7.6.3
      Resolution:                    |         Keywords:
Operating System:  Unknown/Multiple  |     Architecture:  x86_64 (amd64)
 Type of failure:  Runtime crash     |       Difficulty:  Unknown
       Test Case:                    |       Blocked By:
        Blocking:                    |  Related Tickets:
-------------------------------------+----------------------------------

Comment (by slyfox):

 My theory is the bug is induced by '''{-# INLINABLE insert #-}''' and
 '''{-# INLINABLE lookup #-}'''
 pragmas.

 It makes Main module use '''BuggyOpt.$sinsert''' specialisation (with it's
 environment taken into account).
 It's seen nicely when we build Main in both modes:

 good (all ok: Base.lookup):
 {{{
 ==================== Occurrence analysis ====================
 [snip instances]
 Main.mkLin
   :: Arr.Array GHC.Types.Int
      -> Base.Map (Arr.Array GHC.Types.Int) GHC.Types.Int
 [LclIdX,
  Arity=1,
  Unf=Unf{Src=<vanilla>, TopLvl=True, Arity=1, Value=True,
          ConLike=True, WorkFree=True, Expandable=True,
          Guidance=IF_ARGS [0] 110 0}]
 Main.mkLin =
   \ (mseqs_amq :: Arr.Array GHC.Types.Int) ->
     case Base.lookup
            @ (Arr.Array GHC.Types.Int)
            @ (GHC.Prim.Any *)
            (Arr.$fEqArray @ GHC.Types.Int)
            mseqs_amq
            (Base.empty @ (Arr.Array GHC.Types.Int) @ (GHC.Prim.Any *))
     of _ { __DEFAULT ->
     Base.insert
       @ (Arr.Array GHC.Types.Int)
       @ GHC.Types.Int
       $dOrd_apD
       mseqs_amq
       (GHC.Types.I# 1)
       (Base.empty @ (Arr.Array GHC.Types.Int) @ GHC.Types.Int)
     }
 }}}

 bad (note BuggyOpt.$sinsert on an early phase):
 {{{
 ==================== Occurrence analysis ====================
 Main.mkLin
   :: Arr.Array GHC.Types.Int
      -> Base.Map (Arr.Array GHC.Types.Int) GHC.Types.Int
 [LclIdX,
  Arity=1,
  Unf=Unf{Src=<vanilla>, TopLvl=True, Arity=1, Value=True,
          ConLike=True, WorkFree=True, Expandable=True,
          Guidance=IF_ARGS [0] 100 0}]
 Main.mkLin =
   \ (mseqs_amy :: Arr.Array GHC.Types.Int) ->
     case Base.lookup
            @ (Arr.Array GHC.Types.Int)
            @ (GHC.Prim.Any *)
            (Arr.$fEqArray @ GHC.Types.Int)
            mseqs_amy
            (Base.empty @ (Arr.Array GHC.Types.Int) @ (GHC.Prim.Any *))
     of _ { __DEFAULT ->
     BuggyOpt.$sinsert
       @ GHC.Types.Int
       mseqs_amy
       (GHC.Types.I# 1)
       (Base.empty @ (Arr.Array GHC.Types.Int) @ GHC.Types.Int)
     }
 }}}

 Does it look suspicious?

--
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8425#comment:5
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler