#8569: ASSERT in testcase type-rep, only in some ways: -------------------------------------+------------------------------------ Reporter: nomeata | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Test Suite | Version: 7.7 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by nomeata): Tracing this further I find this equation: {{{ lubUse :: UseDmd -> UseDmd -> UseDmd [..] lubUse (UProd ux1) (UProd ux2) | length ux1 == length ux2 = UProd $ zipWith lubMaybeUsed ux1 ux2 | otherwise = Used lubUse (UProd {}) (UCall {}) = Used -- lubUse (UProd {}) Used = Used lubUse (UProd ux) Used = UProd (map (`lubMaybeUsed` useTop) ux) lubUse Used (UProd ux) = UProd (map (`lubMaybeUsed` useTop) ux) [...] lubUse Used _ = Used -- Note [Used should win] }}} Note that `UProd` lub’ed with `Used` yields `UProd`. This contradicts the note `Used should win`. Time for some git archeology... hmm, seems to be like this since dawn of time`^W`the demand analyzer. And as long as no GADTs are around, it probably does not cause problems. A simple fix would be to change that equation, but surely there are reasons for it... -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8569#comment:6 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler