[GHC] #10845: Incorrect behavior when let binding implicit CallStack object

[GHC]

#10845: Incorrect behavior when let binding implicit CallStack object
-------------------------------------+-------------------------------------
        Reporter:  nitromaster101    |                Owner:  gridaphobe
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  7.11
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #10846            |  Differential Rev(s):
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Ah, yes, you are correct.   (You can have a nested implication with
 skolems but no givens; that's important.  Here there are no skolems
 either, and it'd take a little longer to explain why.)

 Rather than conditionally suppressing the no-given solver for `CallStack`,
 a better way is probably this.

  * Remove the auto-solve for `CallStack` (with no givens) altogether.

  * Instead, look at `TcSimplify.simplifyInfer`.  It figures out what to
 quantify over.  At top level (look at `rhs_tclvl`) we don't want to
 quantify over `CallStack` constraints; instead we want to discharge
 (solve) them.  After getting the constraints back from `approximateWC`, we
 can simply solve any un-solved `CallStack` constraints.

 Does that help?  I should have though of this first.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10845#comment:10>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler