[GHC] #11276: GHC hangs/takes an exponential amount of time with simple program

[GHC]

#11276: GHC hangs/takes an exponential amount of time with simple program
-------------------------------------+-------------------------------------
        Reporter:  mpickering        |                Owner:  gkaracha
            Type:  bug               |               Status:  new
        Priority:  highest           |            Milestone:  8.0.1
       Component:  Compiler          |              Version:  7.10.3
      Resolution:                    |             Keywords:
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 gkaracha):

 Ha, finally! It took more than I expected but I found the source: I do not
 understand why but
 GHC wraps all patterns in `CoPat`s:
 {{{#!hs
 Just (EventBeginDocument |> <Event>_R)
 Just (EventEndDocument   |> <Event>_R)
 ...
 }}}
 Hence, the following match of `translatePat` (in `deSugar/Check.hs`) is
 triggered:
 {{{#!hs
   CoPat wrapper p ty -> do
     ps      <- translatePat p
     (xp,xe) <- mkPmId2FormsSM ty
     let g = mkGuard ps (HsWrap wrapper (unLoc xe))
     return [xp,g]
 }}}
 This means that, e.g. for the first clause, instead of the *expected*:
 {{{
 Just EventBeginDocument
 }}}
 The following is generated:
 {{{
 Just (d2JK (EventBeginDocument <- d2JK))
 }}}
 And like this, we end up with a bunch of guards which make the checker
 explode.

 The reason I
 had it translated like this is that `CoPat`s are used for data families,
 where we have the
 original and a representation type constructor. Dropping the `wrapper`
 would give a type
 error (while it shouldn't) because the two constructors are different so
 the guaird translation
 allowed me to match against `d2JK` using the source type (original tyCon)
 and then match the
 guard `(EventBeginDocument <- d2JK)` using the internal type
 (representation tyCon) and avoid
 the mix-up.

 It will need some tinkering but I think I can find a workaround.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11276#comment:9>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler