[GHC] #10270: inconsistent semantics of type class instance visibility outside recursive modules

[GHC]

#10270: inconsistent semantics of type class instance visibility outside recursive
modules
-------------------------------------+-------------------------------------
        Reporter:  skilpat           |                   Owner:
            Type:  bug               |                  Status:  new
        Priority:  normal            |               Milestone:
       Component:  Compiler          |                 Version:  7.10.1
      Resolution:                    |                Keywords:
Operating System:  MacOS X           |            Architecture:  x86_64
 Type of failure:  GHC rejects       |  (amd64)
  valid program                      |               Test Case:
      Blocked By:                    |                Blocking:
 Related Tickets:  #9562             |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by ezyang):

 The plan was for orphan instances, to check if any given instance was in
 the set of orphan modules which are transitively imported from this
 module. We did implement it.

 Unfortunately, this doesn't directly help here because the instance in
 A.hs is NOT an orphan, so we just assume that it's in scope. So in some
 sense, this is the same problem that we were grappling with hsigs:
 sometimes, a non-orphan instance should NOT be visible, because it was not
 defined in the hs-boot/hsig file. We ''also'' spent a bit of time thinking
 about how to solve this, but all of our solutions boiled down to some
 variant of "put the set of transitively imported modules (including non-
 orphan modules)" into the interface, with various schemes for optimizing
 this since this results in quadratic growth of interface files.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10270#comment:4>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler