[GHC] #14895: STG CSE makes dead binders undead

[GHC]

#14895: STG CSE makes dead binders undead
-------------------------------------+-------------------------------------
        Reporter:  hsyl20            |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.2
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Debugging         |  Unknown/Multiple
  information is incorrect           |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Delicate!

 Can you add a Note in `StgCSE` to explain?

 Actually it'd be more correct to do this in `StgCSE`

 * Zap the dead-binder flag on most case-binders (explaining why)
 * But not the ones for primop applications (because they aren't CSE'd, I
 think you are saying).

 That way they never lie.

 We'd still need a note to say that dead-ness is now a bit pessimistic; but
 the code gen doesn't use it (except in a narrow case) so the pessimism
 doesn't matter.

 Not hard to do.  Thanks

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14895#comment:4>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler