[GHC] #11371: Bogus in-scope set in substitutions

[GHC]

#11371: Bogus in-scope set in substitutions
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:  niteria
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:  8.2.1
       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:  #11360            |  Differential Rev(s):  phab:D1792,
       Wiki Page:                    |  phab:D1801, phab:D1802
-------------------------------------+-------------------------------------

Comment (by simonpj):

 There's a lot going on here.  Concerning comment:39, here is what is
 happening:

 * `CoreSyn.applyTypeToArg` deliberately calls `piResultTy` with a
 possibly-bottom second argument. It'll be bottom when the type is `t1 ->
 t2` and the arg is some non-`Type` expression.

 * `piResultTy` calls `piResultTys`

 * When `piResultTys` finishes it calls `substTy`

 * Which looks at the substitution

 * Which is bottom because you built it strictly from its in-scope set

 * Which was bottom because the second arg to `piResultTy` was bottom

 Sigh.  We could improve matters, but that is at least why you are getting
 the panic.

 So go back to being lazy.  The exponential `typeKind` stuff seems more
 likely. Can you just count how many calls to `piResultTy` and/or
 `typeKind` are being made?   And if large, then on what shaped types?

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11371#comment:41>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler