"lazy" is not optimized away

Consider:

import GHC.Magic  
  
{-# INLINE f #-}  
f x = True  
  
g = lazy f False

Two things should happen: (1) f should not be inlined, but (2) lazy should not be present in final STG (having been eliminated in core prep). I thought this was briefly working on HEAD, but apparently it never worked at all:

ezyang@sabre:~$ ghc-8.0 -c test.hs -ddump-stg -fforce-recomp -O2

==================== STG syntax: ====================
Test.f [InlPrag=INLINE (sat-args=1)]
  :: forall t_axB. t_axB -> GHC.Types.Bool
[GblId,
 Arity=1,
 Caf=NoCafRefs,
 Str=DmdType <L,A>,
 Unf=OtherCon []] =
    \r srt:SRT:[] [eta_sHD] GHC.Types.True [];

Test.g :: GHC.Types.Bool
[GblId, Str=DmdType] =
    \u srt:SRT:[] [] GHC.Magic.lazy Test.f GHC.Types.False;

Clearly it has not been eliminated.

Trac metadata

Trac field	Value
Version	8.1
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture