[GHC] #15886: Spurious warning about incomplete pattern with PatternSynonyms

[GHC]

#15886: Spurious warning about incomplete pattern with PatternSynonyms
-------------------------------------+-------------------------------------
        Reporter:  selinger          |                Owner:  (none)
            Type:  bug               |               Status:  closed
        Priority:  normal            |            Milestone:  8.6.3
       Component:  Compiler          |              Version:  8.6.1
      Resolution:  invalid           |             Keywords:
                                     |  PatternSynonyms,
                                     |  PatternMatchWarnings
Operating System:  Linux             |         Architecture:  x86_64
 Type of failure:  Incorrect         |  (amd64)
  error/warning at compile-time      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by RyanGlScott):

 The rationale for requiring `COMPLETE` signatures to coverage-check
 pattern synonyms can be found in
 [https://ghc.haskell.org/trac/ghc/wiki/PatternSynonyms/CompleteSigs?version=15
 the corresponding GHC wiki page]:

 > The exhaustiveness checker currently chokes on pattern synonyms.
 > They are marked as always fallible patterns which means that we must
 also always include a catch-all case in order to avoid a warning.
 >
 > {{{#!hs
 > data A = A
 >
 > pattern :: A
 > pattern P = A
 >
 > foo :: A -> A
 > foo P = A
 > }}}
 >
 > leads to the warning that pattern matches for `foo` are non-exhaustive.
 >
 > {{{
 > simpletest.hs:7:1: warning: [-Wincomplete-patterns]
 >     Pattern match(es) are non-exhaustive
 >     In an equation for ‘foo’: Patterns not matched: _
 > }}}
 >
 > Inspecting the definition of `P` we can see that the matches for `foo`
 are indeed exhaustive as `A` is a unary data type but the pattern match
 checker does not make use of this information.
 >
 > And neither should it! Pattern synonyms are a means of *abstraction*, if
 the exhaustiveness checker could look through a definition then the
 implementation of `P` would leak into error messages.
 > We want users to be able to replace bona-fide data constructors with
 pattern synonyms without consumers noticing.
 > To that end, we allow users to specify a complete set of pattern
 synonyms in order to sate the pattern match checker. If a complete pragma
 is not provided then we keep the same behaviour as in previous releases.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15886#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler