#15584: nonVoid is too conservative w.r.t. strict argument types -------------------------------------+------------------------------------- Reporter: RyanGlScott | Owner: (none) Type: bug | Status: patch Priority: normal | Milestone: 8.8.1 Component: Compiler | Version: 8.5 Resolution: | Keywords: | PatternMatchWarnings Operating System: Unknown/Multiple | Architecture: Type of failure: Incorrect | Unknown/Multiple error/warning at compile-time | Test Case: Blocked By: | Blocking: Related Tickets: #15305 | Differential Rev(s): Phab:D5116 Wiki Page: | -------------------------------------+------------------------------------- Comment (by RyanGlScott): Replying to [comment:2 dfeuer]:
Users need to be able to predict, reasonably reliably, how far they should go in explaining totality to the compiler.
I agree! This proposed change would make it easier for users' to predict when a function is total. The rule is this: if it is possible to pass a well typed argument to a function //that doesn't bottom when demanded//, then that function must match that argument in order to be considered total. (The part in italics is what is new in GHC 8.8 after #15305, and further improved upon in this ticket.)
We certainly wouldn't want GHC to complain about a redundant pattern match in
{{{#!hs f :: S -> a f (MkS (MkV v)) = absurd v }}}
because that's perfectly reasonable code.
I would disagree about this being perfectly reasonable code—there's no reason to ever match on `MkS` (or `MkV`), since that is guaranteed to bottom when demanded. In other words, it's another form of unreachable code, since calling `f (MkS (MkV ⊥))` will bottom out before `MkV` (or `MkS`) are ever reached. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/15584#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler