[GHC] #10180: Lint check: Empty alternatives imply bottoming scrutinee

[GHC]

#10180: Lint check: Empty alternatives imply bottoming scrutinee
-------------------------------------+-------------------------------------
        Reporter:  nomeata           |                   Owner:  nomeata
            Type:  task              |                  Status:  infoneeded
        Priority:  normal            |               Milestone:
       Component:  Compiler          |                 Version:  7.11
      Resolution:                    |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:
 Type of failure:  None/Unknown      |  Unknown/Multiple
      Blocked By:                    |               Test Case:
 Related Tickets:                    |                Blocking:
                                     |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by nomeata):

 Hi,


 Replying to [comment:3 simonpj]:
 > Very interesting.
 >  * What program gives rise to `case (x::Void) of {}`?

 Precisely that program :-)

 See `Data.Void`:
 {{{
 absurd :: Void -> a
 absurd a = case a of {}
 }}}


 >  * Yes, I think there is a case for making `exprIsBottom` hold for data
 types with no constructors; after all, such an expression is bound to
 diverge.
 >  * A slightly less pervasive change would to to say that an empty `case`
 is ok on a data type that has an empty set of constructors.  Less
 pervasive in the sense that it affects this Lint check only, whereas the
 `exprIsBottom` fix would have broader implications: good ones, I think,
 but also rare.

 I’ll do that first, and afterwards look into moving the no-constructor-
 test into `exprIsBottom`.

 Replying to [comment:4 simonpj]:
 > Replying to [comment:2 nomeata]:
 > > So there is a lambda under the case, but it is a type lambda. Bogus or
 not?
 >
 > Ha!  There's a missing case in `exprIsBottom`:
 > {{{
 > exprIsBottom (Lam v e) | isTyVar v = exprIsBottom e
 > }}}
 > Well spotted.

 I’ll add that and see if the linter is happy then.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10180#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler