Workaround for GHC 8.10.5 on macOS

[konn]

As pointed out by @mouse07410 in #1913, we can build HLS with GHC 8.10.5 on macOS just by removing CommandLineTools from the environment.
This PR tries to adopt this workaround in a rather ad-hoc manner.

[konn]

I want to test build action for GHC 8.10.5 + macOS - is there any way to invoke the build workflow manually?

[jneira]

I want to test build action for GHC 8.10.5 + macOS - is there any way to invoke the build workflow manually?

It is triggered on releases so i usually test them in my repo, creating a fake release there and delete it after the test.
(We could change the trigger in the script and restore it after the test but i would recommend test it in your repo, you could link to your succesful build for completeness)

[konn]

Here it goes! 🚀 https://github.com/konn/haskell-language-server/runs/2837481049?check_suite_focus=true

[konn]

Hurray! It at least builds the network with GHC 8.10.5 on macOS 🎉

[Ailrun]

Awesome! Now we can have GHC 8.10.5 on macOS for future versions too :)

[konn]

OK, it finally generated HLS binary for GHC 8.10.5 on macOS 🎉

[konn]

I will re-upload the HLS exes generated by GHA for macOS to haskell/haskell-language-server after the test CI successfully passes.

[jneira]

Amazing work, many thanks, macOS users will be grateful

[konn]

The hack incoporated in this pull request is essentially due to @mouse07410.
Credit goes to them, great discovery!

[konn]

Oops, tests got cancelled by ghc 9 failure -- is this intended behaviour?

[konn]

Ah, I used the wrong commit to build 1.2.0 on my fork repository. This Pull Request itself is correct, but I will re-run the dummy release on my fork to generate macOS binaries with the correct commit.

[konn]

I've just re-uploaded gzip and tarballs, as it seems all the tests pass now.

[jneira]

good to merge then, thanks again