#9279: Local wrapper function remains in final program; result = extra closure allocation -------------------------------------+------------------------------------- Reporter: simonmar | Owner: simonpj Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.2 Resolution: | Keywords: LateLamLift Operating System: Unknown/Multiple | Architecture: Type of failure: Runtime | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by sgraf): `Note [Absent errors]` reads {{{ Note: I did try the experiment of using an error thunk for unlifted things too, relying on the simplifier to drop it as dead code. But this is fragile - It fails when profiling is on, which disables various optimisations - It fails when reboxing happens. E.g. data T = MkT Int Int# f p@(MkT a _) = ...g p.... where g is /lazy/ in 'p', but only uses the first component. Then 'f' is /strict/ in 'p', and only uses the first component. So we only pass that component to the worker for 'f', which reconstructs 'p' to pass it to 'g'. Alas we can't say ...f (MkT a (absentError Int# "blah"))... bacause `MkT` is strict in its Int# argument, so we get an absentError exception when we shouldn't. Very annoying! So absentError is only used for lifted types. }}} But I suppose having just use `NULL` for the other unlifted+boxed cases is OK. We've been doing it for `Addr#` for a long time. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/9279#comment:17 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler