[GHC] #13750: GHC produces incorrect coercions in hairy code

[GHC]

#13750: GHC produces incorrect coercions in hairy code
-------------------------------------+-------------------------------------
        Reporter:  dfeuer            |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  highest           |            Milestone:  8.2.2
       Component:  Compiler          |              Version:  8.2.1-rc2
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  Runtime crash     |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #13429            |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Well this is embarrassing.  I did take a quick look.  The test program is
 littered with `unsafeCoerce` which makes it harder to determine where
 things actually go wrong.

 I still think the "decideQuantification" patch is fine; I think it is just
 exposing bugs elsewhere.

 I discovered that (like #13429), adding `-fno-specialise` (to switch off
 the specialiser) makes it work again.  So that may be a workaround for
 Andres.

 I'm not against reverting the "decideQauntification" patch as an expedient
 way to make the release work; but that would re-introduce the bugs that
 the patch cured.

 I'm on holiday now for a week, I'm afraid.  It'd be great if someone was
 able to characterise more precisely what is going wrong.  At the moment I
 have no clue.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13750#comment:6>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler