Update for GHC 9.8

[amesgen]

Otherwise, one might get errors like

Error: cabal: Could not resolve dependencies:
[__0] trying: ormolu-0.7.2.0 (user goal)
[__1] trying: th-env-0.1.1 (dependency of ormolu)
[__2] trying: template-haskell-2.21.0.0/installed-54dc (dependency of th-env)
[__3] trying: pretty-1.1.3.6/installed-8ddc (dependency of template-haskell)
[__4] trying: megaparsec-9.5.0 (dependency of ormolu)
[__5] next goal: Cabal-syntax (dependency of ormolu)
[__5] rejecting: Cabal-syntax-3.10.2.0/installed-6d41 (conflict: megaparsec =>
bytestring>=0.2 && <0.12, Cabal-syntax => bytestring==0.12.0.2/installed-66e0)
[__5] rejecting: Cabal-syntax-3.10.1.0 (conflict: pretty =>
deepseq==1.5.0.0/installed-2833, Cabal-syntax => deepseq>=1.3.0.1 && <1.5)
[__5] rejecting: Cabal-syntax-3.8.1.0 (conflict: ormolu => Cabal-syntax>=3.10
&& <3.11)
[__5] skipping: Cabal-syntax-3.6.0.0 (has the same characteristics that caused
the previous version to fail: excluded by constraint '>=3.10 && <3.11' from
'ormolu')
[__5] fail (backjumping, conflict set: Cabal-syntax, megaparsec, ormolu,
pretty)

as bytestring-0.12 (shipped with GHC 9.8) is non-reinstallable when using TH.

• Allowing containers-0.7 is not strictly necessary, but it compiles fine.

• Due to this change in base-4.19

  Added {-# WARNING in "x-partial" #-} to Data.List.{head,tail}.

  there are various new warnings with GHC 9.8. We fix these by enabling -Wno-x-partial.

[andreasabel]

Thanks. LGTM, with a question about NE.fromList.

[andreasabel]

Isn't a use of NE.fromList as bad as using head? Both will throw the exception in the library, not in the application code.
If all that head is bad excitement should lead to improvement, then the new code should be fromMaybe undefined $ NE.nonEmpty $ take 1 r which will raise the error (via undefined) here, not in the library.

[amesgen]

Both head and NE.fromList use HasCallStack, so they will at least point to the code here. But yeah, I don't think there is any big advantage of using NE.fromList rather than head other than to avoid the warning (in case that is even desired).

[andreasabel]

I published a revision of the bytestring bound to hackage: https://hackage.haskell.org/package/megaparsec-9.5.0/revisions/
This was missing for GHC 9.8.1, which ships bytestring-0.12.

[mrkkrp]

Thanks!

[mrkkrp]

I will make a revision on Hackage.