Krzysztof Gogolewski

This is the result of `make accept` for this test: https://github.com/haskell/haddock/commit/21e156a76c1dfad008546688a8f9f3a803c063ba AFAICT the changes are benign but I didn't look thoroughly.

I'm currently busy but will go back to it.

At the moment, CI on the `ghc-9.0` branch in GHC gitlab is broken, I will get back to this once it's fixed. https://gitlab.haskell.org/ghc/ghc/-/merge_requests/6336#note_370676

Since 9.4 was released and this MR is only about a different expected output, I'm closing as abandoned/no longer relevant.

A related example, involving a qualified name, is available at https://gitlab.haskell.org/ghc/ghc/-/issues/20945. ``` data Foo a = Foo a deriving (Functor) instance Applicative (Foo) where pure = Foo ``` gives ```...

Earlier discussion: https://gitlab.haskell.org/ghc/ghc/-/issues/17816#note_256167

Yes - `error-message-index` is used to describe existing errors, while this one is for improving them.

It depends on how smart we want Lint to be. More complex properties such as associativity could be just axioms and Core would have coercions witnessing them.

Small failing testcase, with `-O -dcore-lint`: ``` module Base where data Map = Bin Int !Map | Tip go :: Map -> Map go Tip = Bin 0 Tip go...

I've marked the comments about CoVar as "resolved" since they'll be handled at https://gitlab.haskell.org/ghc/ghc/merge_requests/2268.