[GHC] #16296: OccurAnal should not break the linter assumptions

[GHC]

#16296: OccurAnal should not break the linter assumptions
-------------------------------------+-------------------------------------
        Reporter:  aspiwack          |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.6.3
      Resolution:                    |             Keywords:  Simplifier
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):

 I wrote a quick fix, which has now landed as
 https://gitlab.haskell.org/ghc/ghc/commit/0eb7cf03da3783ca887d5de44d312cf6f3a4113c]

 But it's not the Right Solution; it's messy, and it loses some useful
 optimisations, namely
 {{{
 case Imported.x of
   True -> .....(case Imported.x of { True -> e1; False -> e2 })......
   False -> ...
 }}}
 We'd like the inner case to disappear, but now it won't.

 I think the right solution is as outlined in the Description.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/16296#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler