#16260: Use of plugins causes -XSafe to fail -------------------------------------+------------------------------------- Reporter: watashi | Owner: watashi Type: bug | Status: closed Priority: normal | Milestone: Component: Compiler | Version: 8.7 Resolution: fixed | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: GHC rejects | Unknown/Multiple valid program | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by watashi): @mpickering Thanks, I didn't notice there was already a similar ticket. `-fno-safe-haskell` would resolve the compilation issue in the ticket, but we still gets inconsistent `TcGblEnv`. For more background, I am trying to use source plugins to implement paralleled haddock, and this will generate wrong "Safe Haskell" information even we don't make any change. As you commented in #15920
I'm not that enthusiastic about a proposal as I don't know that anyone should rely on safe haskell as it's quite a neglected feature. I think I would rather write a proposal about removing safe haskell rather than understand how it should interact more precisely with plugins.
In this case, do you think it makes sense to simply remove calls of `mark_plugin_unsafe` in `renamed/typeCheckResultAction` https://gitlab.haskell.org/ghc/ghc/blob/c2455e64/compiler/typecheck/TcRnDriv... and push the responsibility of calling `recordUnsafeInfer` to the plugin author and document this somewhere? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/16260#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler