[GHC] #7411: Exceptions are optimized away in certain situations

[GHC]

#7411: Exceptions are optimized away in certain situations
-------------------------------------+-------------------------------------
        Reporter:  SimonHengel       |                Owner:  tdammers
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:  8.6.1
       Component:  Compiler          |              Version:  7.6.1
      Resolution:                    |             Keywords:  seq, deepseq,
                                     |  evaluate, exceptions
Operating System:  Linux             |         Architecture:  x86_64
                                     |  (amd64)
 Type of failure:  Incorrect result  |            Test Case:
  at runtime                         |  simplCore/should_fail/T7411
      Blocked By:                    |             Blocking:
 Related Tickets:  #5129             |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by dfeuer):

 I think the `-fno-state-hack` is almost certainly a clue, although I'm not
 exactly sure how it ends up relating in that specific example. What I
 ''suspect'' is happening (without actually tracing simplification) is
 roughly this:

 {{{#!hs
 evaluate (undefined `seq` return ())
 -- ===> inline
 IO $ \s -> seq# (undefined `seq` IO (\s' -> (# s', () #))) s
 -- ===> simplify the coercions for my sanity
 IO $ \s -> seq# (IO $ undefined `seq` \s' -> (# s', () #)) s
 -- ===> This is where I suspect things go sideways. I believe we generally
 --      assume that it's okay to eta-expand IO things.
 IO $ \s -> seq# (IO $ \s'' -> (undefined `seq` (\s' -> (# s', () #))) s'')
 s
 }}}

 We've effectively turned a bottom with an `IO` type into a non-bottom `IO`
 value that only ''returns'' a bottom. Then the `seq#` ends up going away:

 {{{#!hs
 -- ===> beta reduce past seq
 IO $ \s -> seq# (IO $ \s'' -> undefined `seq` (# s'', () #)) s
 -- ===> PrelRules.seqRule says we can eliminate seq# for WHNF things
 IO $ \s -> (# s, IO $ \s'' -> undefined `seq` (# s'', () #))
 }}}

 I imagine the behavior is a bit fragile because it depends on GHC not
 recognizing that the undefined value is in fact undefined, or at least not
 recognizing it too early. I suspect the solution is likely to make
 `-fpedantic-bottoms` go a little further than it currently does, although
 I don't know enough to say just what.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/7411#comment:25>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler