#8310: Can we change the semantics of `Trustworthy`? -------------------------------------+------------------------------------ Reporter: ekmett | Owner: dterei Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.6.3 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 ekmett): My understanding was that if I import a module that is Trustworthy, my module could still infer as Safe-Inferred. Perhaps that was a naïve interpretation. I'll need to go back through and re-read the specification, and understand the safe import machinery better. Right now my main problem is that no matter what I do, I've just not proven smart enough to make the most specific annotation that works for all of my dependencies without unnecessarily enlarging the trusted code base. One possible middle ground that doesn't muddle those semantics is that we could issue a warning when you needlessly mark a module as Trustworthy, when it would otherwise infer as Safe-Inferred. Then I'd at least get some kind of feedback during the development process and from my end-users, and cleaning up warnings would catalyze users into writing patches. This would still result in me having to maintain some painfully complicated logic to get the right safety properties, but it would at least tell me pointwise when I was in a situation where I'd gotten it wrong. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8310#comment:4 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler