Support GHC 9.2

[sjakobi]

No description provided.

[sjakobi]

The build failure with GHC 7.10.3 is due to tsurucapital/vector-th-unbox#7.

I see several options for now:

• Wait until Build failures with GHC < 8 tsurucapital/vector-th-unbox#7 is resolved.
• Drop support for GHC 7.10.
• Migrate the benchmarks from criterion to tasty-bench to avoid the dependency on vector-th-unboxed.

Any preferences?

[chrisdone]

Looks like the first option panned out.

I think migrating to gauge would also be fine, but not a big priority unless someone is especially excited about gauge.

Explicitly dropping support for GHC 7.10 seems unnecessary at this point. It’s just a benchmark failure. That’s why I set it to allow fail, but didn’t explicitly update the base lower bound.

[chrisdone]

Merging the GHC 9 support, thanks!

[sjakobi]

It’s just a benchmark failure. That’s why I set it to allow fail, but didn’t explicitly update the base lower bound.

Oh, I hadn't seen that. With haskell-ci, if you want to preserve a certain setting, it's best to put it into the cabal.haskell-ci config, e.g.

allow-failures: == 7.10.3

Future runs of haskell-ci regenerate will then take this setting into account.

BTW, could you publish the bumped bound on base via a Hackage revision? Thanks!

[chrisdone]

I've published a new revision.

[chrisdone]

Thanks @sjakobi Looks like lucid didn't need any changes to work on GHC 9.2, which is good.