[GHC] #9718: Avoid TidyPgm predicting what CorePrep will do

[GHC]

#9718: Avoid TidyPgm predicting what CorePrep will do
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:  (none)
            Type:  task              |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  7.8.3
      Resolution:                    |             Keywords:  CodeGen, CAFs
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Re "Finally I see" Simon M pointed out the that the code generator has a
 special case for indirection bindings `f = g`:
 {{{
 cgTopRhsClosure dflags rec id ccs upd_flag args body =
    ...
   where
   -- special case for a indirection (f = g).  We create an IND_STATIC
   -- closure pointing directly to the indirectee.  This is exactly
   -- what the CAF will eventually evaluate to anyway, we're just
   -- shortcutting the whole process, and generating a lot less code
   -- (#7308)
   --
   -- Note: we omit the optimisation when this binding is part of a
   -- recursive group, because the optimisation would inhibit the black
   -- hole detection from working in that case.  Test
   -- concurrent/should_run/4030 fails, for instance.
   --
   gen_code dflags _ closure_label
     | StgApp f [] <- body, null args, isNonRec rec
     = ...
 }}}
 We decided that:
 * The code generator makes an IND_STATIC for these
 * So `f` itself is not a CAF
 * But it might still be CAFFY if `g` is.

 Hence the CAF analysis could (and probably should) take advantage of this
 special case (alas, still a form of prediction!) provided it points to a
 specific Note about it in the code generator.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9718#comment:30>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler