[GHC] #11163: New exhaustiveness checker breaks T5642

[GHC]

#11163: New exhaustiveness checker breaks T5642
-------------------------------------+-------------------------------------
        Reporter:  bgamari           |                Owner:  gkaracha
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:  8.0.1
       Component:  Compiler          |              Version:  7.10.2
      Resolution:                    |             Keywords:  pattern
                                     |  checker, PatternMatchWarnings
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Compile-time      |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by gkaracha):

 After merging Phab:D1795, I checked again the situation with T5642:

 With the old pattern match checker the deviation (for bytes allocated) for
 #5642 was 196.8%. With
 the current HEAD it is 222.3%, yet this increase is not only due to the
 new checker. Disabling
 coverage checking for the test gives deviation 212.0%, which means that
 only 10.3% is due to the
 new checker.

 Concerning this 10.3%, the module uses the generics library and function
 `to :: Rep a x -> a` has
 100 clauses, all with nested patterns. Most importantly, `Rep` is a type
 family, which means that
 the new checker goes the extra mile to ensure they are all well-typed.
 Hence, this 10.3% is quite
 reasonable (the old pattern match checker was --in most cases-- faster
 because it did not take into
 account types).

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11163#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler