Support for GHC-9.12.x?

[mouse07410]

Is your enhancement request related to a problem? Please describe.

GHC-9.12.1 has been around for a while, and 9.12.2 is on the horizon. It would be nice to have HLS supporting these.

Describe the solution you'd like

Added support for GHC-9.12.x. Current master doesn't seem to work because of (at least?) two dependent packages failing to compile under 9.12.1:

• ghc-exactprint-1.10.0.0
• ormolu-0.7.7.0

Describe alternatives you've considered

There's no alternative, except for downgrading GHC to 9.10.1. Which is not that attractive for those living on the "bleeding edge". ;-)

[eldritch-cookie]

For ghc-exactprint a version is compatible with only one major version of GHC so 1.10.0.0 only works with GHC 9.10.x
but the solution is simple just upgrade to 1.12.0.0,
however given that it is a major version upgrade it is highly likely there will be breakages that need to be fixed.
Regarding ormolu there is a newer version that has a built with that claims compability with GHC 9.12.1 which i am inclined to believe as the latest fourmolu can in fact be compiled on GHC 9.12.1

[fendor]

There will not be binaries for 9.12.1, as the GHC 9.12.1 release is broken.

However, we are actively working on 9.12.2 support as soon as it releases.

[sohang3112]

There will not be binaries for 9.12.1, as the GHC 9.12.1 release is broken.

@fendor Can you please clarify, how is GHC 9.12.1 release "broken"?

[jhrcek]

In general you can find the list of issues via GHC status page.
GHC 9.12.1 in particular has some "incorrect runtime result" issues: https://gitlab.haskell.org/ghc/ghc/-/milestones/405#tab-issues

[fendor]

@sohang3112 I am referring to the issue https://gitlab.haskell.org/ghc/ghc/-/issues/25653. There was also this announcement: https://discourse.haskell.org/t/psa-correctness-issue-in-ghc-9-12/11204

[Jashweii]

GHC 9.12.2 has been released and added to ghcup
https://discourse.haskell.org/t/ghc-9-12-2-is-now-available/11639

[soulomoon]

I’m personally very excited about this release! I can imagine that semantic tokens will be much faster now that version 9.12.2 is equipped with additional entity information for each token. We no longer need to ad-hoc search for them through various built artifacts.

Preparing to update 9.12.2.

1. Preparing to update to 9.12.2, since ghcide-test now lift up to haskell-language-server.cabal independently. It is best to move it to a stand-along dir. Maybe also drop older ghc support too.
2. update ghcide.
3. build ghcide-test to test ghcide.
4. update indivisual extensions.

[larskuhtz]

There will not be binaries for 9.12.1, as the GHC 9.12.1 release is broken.

However, we are actively working on 9.12.2 support as soon as it releases.

I don't think that it is the job of a language server to decide what version of a compiler users should use.

There are valid reasons to use experimental or even broken versions of GHC -- for testing, early integration, or development of language tools and infrastructure.

There are programming languages for which language servers are available even for pre-releases of compilers. I think, the Haskell ecosystem would benefit if projects that form part of the basic language infrastructure would be available as soon as possible for new (or even upcoming) versions of GHC.

[fendor]

We generally try to do exactly that, to have the core of HLS running the moment a GHC version is released.
However, in this particular case, there had been a real chance that the release CI wouldn't even have succeeded, which would have been rather awkward.

Otherwise, we agree with you, and we try to have support for new GHC ASAP.