hie-compat: Add basic support for ghc 9.0.1

[anka-213]

I made a three way merge between the original in GHC-8.10, GHC-9.0 and the custom version for GHC-8.10 from this package, in order to figure out what should be changed.

Perhaps we don't actually need the full fork of HieAst.hs for ghc-9.0, since the difference from the original is so small, so we could just reexport the original and replace mkHieFile with the custom version.

(Tiny step towards #297)

[jneira]

Many thanks for working on ghc-9.0.1 support 🙏

[anka-213]

All of these changes are in the new PR: #1649, so I'll just close this.

[wz1000]

Does this need to be duplicated? I think it should be possible to just reexport GHC.Iface.Ext.Binary

[anka-213]

Not sure. I just copied whatever was in the ghc-810 versions and updated it to the ghc-901 version.

[anka-213]

Is there a reason why the circleci tests are not running? Is it waiting for something else first?

[Ailrun]

There's nothing on the queue (even this one), so it's an actual problem. I don't know whether it's a CircleCI bug or our setting's. One possible thing to try is add one commit (or rebase this PR) so that CI check is reactivated.

[jneira]

The circleci job for ead87d8 finished succesfully but the status was not changed: https://app.circleci.com/pipelines/github/haskell/haskell-language-server/3004/workflows/999e4192-a140-4e0f-a99f-ffeea46d5778
I've reran it, lets see what happens: https://app.circleci.com/pipelines/github/haskell/haskell-language-server/3004/workflows/ecd57e2f-3184-4032-87cf-63320e5854b6

[Ailrun]

It looks like the linked CI run is for ead87d8 and the problematic one is for 47a7b44, so unfortunately I don't think rerun will help...

[jneira]

Mmm you are right, the merge was clean and maybe circleci has some trouble with that this time

[anka-213]

Is it possible to manually trigger circleci? Or should I rebase, so it can have a new chance to trigger?

[anka-213]

I'll do that and then it should hopefully work as soon as someone re-approves the changes

[anka-213]

Maybe it didn't think it was worth re-running the CI because the merge only changed the README?

It seems to run the circleci again now.

[jneira]

Maybe it didn't think it was worth re-running the CI because the merge only changed the README?

It seems to run the circleci again now.

Unfortunately circleci dont support skip builds by file globs so not sure what happened (otoh checks would be not required in that case)
Afaik it does not support neither new manual builds chosing a commit of a branch out of the box. 😐