#12522: GHC 8.0.1 hangs, looping forever in type-checker -------------------------------------+------------------------------------- Reporter: clinton | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Compile-time | Unknown/Multiple crash | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by goldfire): Replying to [comment:11 simonpj]:
Question: why do we use pre-unification when matching the RHS in the injectivity-improvement step. That's what is causing the trouble.
This all looks like reasonable behavior to me on a line-by-line basis. Obviously a loop is bad, but each step along the way is OK. We use pre-unification here because we want to match the pattern `Maybe (TF (x, a))` against `Maybe Char`. And indeed we should, because injectivity tells us that the equation `TF (D1 x, a) = Maybe (TF (x, a))` must be the one used to produce `Maybe Char`. (In fact, the solver should also generate `[W] TF (x, a) ~ Char`. Or maybe it should be Derived. Does it?) So I like the use of pre-unification here. What I find strange is that we're using the model when triggering improvement as we're looking at a Wanted. Doesn't the model (that is, all the Derived constraints) live off in its own world with minimal interaction with Wanteds? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/12522#comment:12 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler