[GHC] #12549: Panic on ":t datatypeName"

[GHC]

#12549: Panic on ":t datatypeName"
---------------------------------+--------------------------------------
        Reporter:  johnleo       |                Owner:  johnleo
            Type:  bug           |               Status:  new
        Priority:  normal        |            Milestone:
       Component:  Compiler      |              Version:  8.1
      Resolution:                |             Keywords:
Operating System:  MacOS X       |         Architecture:  x86_64 (amd64)
 Type of failure:  None/Unknown  |            Test Case:
      Blocked By:                |             Blocking:
 Related Tickets:                |  Differential Rev(s):
       Wiki Page:                |
---------------------------------+--------------------------------------

Comment (by goldfire):

 Your analysis is spot on. Thank you!

 The problem is (as you may have suspected) that comment on
 `newMetaTyVars`. It's just plain wrong.

 The problem is that we must always obey '''The substitution invariant'''
 from the eponymous Note in !TyCoRep:

 {{{
 When calling (substTy subst ty) it should be the case that
 the in-scope set in the substitution is a superset of both:

   * The free vars of the range of the substitution
   * The free vars of ty minus the domain of the substitution
 }}}

 `new_meta_tv_x` takes care to make sure, as it's building the
 substitution, that the first point is obeyed. But no one is handling the
 second point!j

 Perhaps it was envisioned that `nwMetaTyVars` would be called only on a
 '''closed''' sequence of tyvars. But the function doesn't say that, and
 evidently it's not true. (I doubt it ever was!) So it seems the answer is
 to add an `InScopeSet` parameter to `newMetaTyVars` that should contain
 the free variables of the type being instantiated.

 Does that help?

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12549#comment:17>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler