#10755: Add `MonadPlus IO` and `Alternative IO` instances -------------------------------------+------------------------------------- Reporter: hvr | Owner: Type: task | Status: patch Priority: normal | Milestone: 7.12.1 Component: Core Libraries | Version: Resolution: | Keywords: report- | impact Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: Phab:D1148 -------------------------------------+------------------------------------- Comment (by hvr): Replying to [comment:5 bgamari]:
+1 to putting these in `base-orphans`. It's unfortunate that people will need to update their dependency lists, but it seems like the right thing to do.
Adding those to `base-orphans` is not the problem (and I'm +1 on that as well). The problem is with already released packages. Specifically, consider the following case: - A package `foo-0.1` relies on `transformer >= 0.4` to supply the `MonadPlus IO` instance. - Now a new `transformers-0.5` is released which lacks the instance. - So now `foo-0.1` breaks when using `base-4.8`+`transformers-0.5`. - `base-4.9` would however have the instance again and `foo-0.1` happens to compile again with `base-4.9`+`transformers-0.5` Now in order to fixup this breakage would need to add upper bounds to the `foo-0.1`, but the problem is we can either - add a `transformers < 0.5` bound to `foo-0.1`, which would block out the sensible `base-4.9`+`transformers-0.5` combination, or - add a `base < 4.9` bound to `foo-0.1`, which block `base-4.9`+`transformers-0.5` as well. What we would need to be able to do is to retrofit the constraint "`base
= 4.9 || transformers < 0.5`", but this requires cabal conditionals and automatic cabal flags, which can't be edited into a package retroactively (for good reason).
If this specific constellation affects a very small amount of packages*versions, we can handle it by a combination of cabal-edits and interacting w/ the authors to upload point releases. Otherwise we're faced with some kind of amputation-dilemma (not sure if there's a better term for it: i.e. when we need to amputate healthy install-plan-limbs in order to save the Hackage-patient...). -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10755#comment:6 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler