[GHC] #10009: type inference regression when faking injective type families

[GHC]

#10009: type inference regression when faking injective type families
-------------------------------------+-------------------------------------
        Reporter:  aavogt            |                   Owner:
            Type:  bug               |                  Status:  new
        Priority:  high              |               Milestone:  7.10.2
       Component:  Compiler (Type    |                 Version:  7.10.1-rc1
  checker)                           |                Keywords:
      Resolution:                    |            Architecture:
Operating System:  Unknown/Multiple  |  Unknown/Multiple
 Type of failure:  GHC rejects       |               Test Case:
  valid program                      |                Blocking:
      Blocked By:                    |  Differential Revisions:
 Related Tickets:  #10226            |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Thanks for re-highlighting this.  I've just been looking at this again, as
 it happens.

 The sad story is that it's an unexpected consequence of some refactoring I
 did to the type-family solver.   I can't undo that on the branch; it's
 quite pervasive.  But it showed up a shortcoming in the way we do type
 inference generally. That is, I exposed an existing bug which happened not
 to show up before.  Fixing that bug is quite non-trivial.

 I do intend to fix it, probably in the next couple of weeks.  Let's see
 how invasive the fix is.

 Meanwhile you say "it is affecting code in the wild".  Can you be more
 specific?

 Simon

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10009#comment:21>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler