[GHC] #8503: New GeneralizedNewtypeDeriving check still isn't permissive enough

[GHC]

#8503: New GeneralizedNewtypeDeriving check still isn't permissive enough
-------------------------------------+------------------------------------
        Reporter:  goldfire          |            Owner:  goldfire
            Type:  bug               |           Status:  new
        Priority:  normal            |        Milestone:
       Component:  Compiler          |          Version:  7.7
      Resolution:                    |         Keywords:
Operating System:  Unknown/Multiple  |     Architecture:  Unknown/Multiple
 Type of failure:  None/Unknown      |       Difficulty:  Unknown
       Test Case:                    |       Blocked By:
        Blocking:                    |  Related Tickets:  #8541
-------------------------------------+------------------------------------

Comment (by nomeata):

 I cherry-picked Richard’s implementation of GND via coerce
 ([https://github.com/goldfirere/ghc/commit/c76fed9a9c6bffc5b0858d39589dabfe93970c3e
 c76fed]) onto wip/T8503 and GHC compiles (after `make clean`) just fine –
 which is great.

 There are test suite failures, though A lot are performance degrations
 (some, if they are about the performance of GHC, are to be expected, the
 Coercible machinery with subsequent simplification is not the fastest
 code). The others are:

 {{{
 Unexpected failures:
    codeGen/should_run            cgrun068 [exit code non-0] (normal)
    deriving/should_compile       T2856 [exit code non-0] (normal)
    deriving/should_compile       drv015 [exit code non-0] (normal)
    deriving/should_fail          T1496 [stderr mismatch] (normal)
    deriving/should_fail          T2721 [stderr mismatch] (normal)
    deriving/should_fail          T4846 [stderr mismatch] (normal)
    deriving/should_fail          T7148 [stderr mismatch] (normal)
    deriving/should_fail          T7148a [stderr mismatch] (normal)
    deriving/should_run           drvrun019 [exit code non-0] (normal)
    gadt                          CasePrune [stderr mismatch] (normal)
    gadt                          gadt6 [exit code non-0] (normal)
    ghci.debugger/scripts         print018 [bad stdout] (ghci)
    indexed-types/should_compile  DerivingNewType [exit code non-0]
 (normal)
    indexed-types/should_compile  IndTypesPerf [bad exit code] (normal)
    indexed-types/should_compile  T2850 [exit code non-0] (normal)
    indexed-types/should_compile  T3423 [exit code non-0] (normal)
    indexed-types/should_compile  T4185 [exit code non-0] (normal)
    indexed-types/should_compile  T5955 [bad exit code] (normal)
    indexed-types/should_compile  T7474 [exit code non-0] (normal)
    lib/integer                   integerConstantFolding [bad stderr]
 (normal)
    polykinds                     T7332 [exit code non-0] (normal)
    roles/should_fail             Roles10 [stderr mismatch] (normal)
    simplCore/should_compile      T4203 [exit code non-0] (normal)
    typecheck/should_compile      T3955 [exit code non-0] (normal)
    typecheck/should_fail         T5246 [stderr mismatch] (normal)
 }}}

 Of these, there are a few where we get unresolved Coercible instances due
 to roles (and many of them might be genuine and can probably be marked as
 expected to fail), but there is also a number of panic and failed
 assertions.

 wip/T8503 itself validates, so these were either introduced or triggered
 by implementing GND via coerce. Probably triggered, given that the GND
 code now generates mostly straight-forward code.

 I’ll start looking into some of them tomorrow, but Richard, if you feel
 like helping debugging them, that would be appreciated. Some of them
 appear in code that I do know know much about, such as module interfaces.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8503#comment:30>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler