endgame
commented
1 month ago Thank you for your question.
Which upper bound is right?
Both. The 2.0 release on Hackage does not compile with GHC 9.8 and has had its base bound revised to reflect this. The code currently on master has been fixed to compile with GHC 9.8 (the meaning of the DuplicateRecordFields extension changed).
Can the two be made to agree?
No, because they're describing different things. One describes the bound on the 2.0 release, and the other describes the bounds at the head of the master branch.
Until a new release is made on Hackage, if you want to use GHC 9.8, you will have to import it from git. The old RC2 release announcement has instructions for how to do this with cabal and stack: https://github.com/brendanhay/amazonka/discussions/716
There is a discrepancy in base upper bound between the v2.0 github code and the v2.0 hackage package description: https://github.com/brendanhay/amazonka/blob/e081929d5a81f30397c9c9a219b154da1a922fb8/lib/amazonka/amazonka.cabal#L75 https://hackage.haskell.org/package/amazonka-2.0/amazonka.cabal
The former has
base >=4.12 && <5while the latter hasbase >=4.12 && <4.19. Which upper bound is right? Can the two be made to agree? I discovered the issue trying to figure out why I can't use amazonka with GHC 9.8. I hope that amazonka is actually usable with GHC 9.8 and the hackage side gets fixed.