[GHC] #10494: Representational equalities over AppTys are not hard failures

[GHC]

#10494: Representational equalities over AppTys are not hard failures
-------------------------------------+-------------------------------------
        Reporter:  goldfire          |                   Owner:  goldfire
            Type:  bug               |                  Status:  merge
        Priority:  normal            |               Milestone:  7.10.3
       Component:  Compiler          |                 Version:
      Resolution:  fixed             |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |               Test Case:
                                     |  typecheck/should_compile/T10494
      Blocked By:                    |                Blocking:
 Related Tickets:                    |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by crockeea):

 I'm about to release a package that triggers the bug, so it currently has
 crufty work-around code that I'd like to kill.

 I have a collection of functions that coerce functions of one signature to
 functions of another signature. Many of the definitions are simply
 `coerce`, but the explicit functions are needed for type inference. Here's
 an example:

 Instead of writing

 {{{
 coerceBasis :: (Coercible (o r) (n m' r))
             => Tagged '(m,m') ([o r]) -> Tagged m [n m' r]
 coerceBasis = coerce
 }}}

 I have to write

 {{{
 coerceBasis :: (a ~ (o r), b ~ (n m' r), Coercible a b)
             => Tagged '(m,m') ([o r]) -> Tagged m [n m' r]
 coerceBasis = coerce
 }}}

 introducing two new intermediate type variables and constraints.

 Is this enough of a problem to make you go through the work of releasing
 7.10.3 just for it? Certainly not. Mostly I was just unsure of the status
 of the patch and didn't want the original bug to be forgotten. That said,
 I *would* like it resolved sooner rather than later. I saw something on
 GHC-Devs about releasing builds, which makes a lot of sense to me.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10494#comment:9>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler