Import Multiplicity from GHC.Types

[tomsmeding]

GHC 9.2 did not yet export Multiplicity from GHC.Exts. The rest of the repository imports Multiplicity from ghc-prim:GHC.Types, from which the type has always been exported, but this one occurrence in Control.Functor.Linear.Internal.Reader didn't.

[aspiwack]

Does linear-base even compile with GHC 9.2 (other than this infelicity?). I confess that I'm rather a burn-bridges kind of person, and I tend to drop support for old versions as soon as they become inconvenient.

I agree that it's better to be uniform, but on the other hand, GHC.Types is is ghc-prim and GHC.Exts is in base. So it's generally preferable to import from the latter (maybe we can even drop our direct dependency on ghc-prim, that could be nice; even if for a library like linear-base, it's not particularly surprising to depend on ghc-prim).

On the other other hand, I assume that this pull request comes from a need. Could you elaborate the need a little?

[tomsmeding]

For me personally, it doesn't matter if you fix the import or increase your lower bound on base. :)

I operate the playground and had to rebuild the package set for GHC 9.2; my script happily told me that cabal was happy with a package set that included linear-base, but the build then proceeded to fail. Seeing it was linear-base I decided to spend a few minutes seeing if I could fix it, and because it was just an import I opened a PR.

I appreciate your arguments, but cabal test succeeds, and if really all that is needed is importing some things from ghc-prim to support 9.2, I'd suggest potentially keeping a few additional users happy.

[aspiwack]

This is fair enough. I'm really surprised that we don't have more incompatibilities with GHC 8.2, though.