Add hls-1.6.0.0

[jneira]

No description provided.

[jneira]

@hasufell ci is happy 😄

[hasufell]

[ Info  ] verifying digest of: haskell-language-server-Linux-alpine-x86_64-1.6.0.0.tar.xz
ghcup-gen: Codec.Compression.Lzma.decompressWith: decoding error LzmaRetFormatError

[jneira]

is that a local check? i cant find that in the ci logs, it seems ci only checks own ghcup tarballs

[hasufell]

See https://github.com/haskell/ghcup-metadata#adding-a-new-ghc-version step 7

cabal run ghcup-gen -- check-tarballs -f ghcup-0.0.6.yaml -u 'hls-1\.6\.0\.0'

[jneira]

thanks! i inferred it from the ci invocation and it is running now

does it take too much time to check it in ci?

[hasufell]

does it take too much time to check it in ci?

How would you run it?

[jneira]

hmm adding a step running cabal run ghcup-gen -- check-tarballs -f ghcup-0.0.6.yaml -u 'hls-1\.6\.0\.0'?

[hasufell]

hmm adding a step running cabal run ghcup-gen -- check-tarballs -f ghcup-0.0.6.yaml -u 'hls-1\.6\.0\.0'?

How will that work for the next PR adding a new HLS version or a different GHC version?

[jneira]

ok reproduced the check error locally but i still am not sure what is going on, i can extract the files succesfully. The unique thing i did locally was rename the file, so ci produced the tarball

[hasufell]

ok reproduced the check error locally but i still am not sure what is going on, i can extract the files succesfully. The unique thing i did locally was rename the file, so ci produced the tarball

We use lzma to decompress:

https://gitlab.haskell.org/haskell/ghcup-hs/-/blob/655ee432f8330d426199ad6ece190fc194f860eb/lib/GHCup/Utils.hs#L716-719

https://hackage.haskell.org/package/lzma-static

It appears to have problem with some feature of the archive, I guess.

[jneira]

The encoding for haskell-language-server-macOS-aarch64-1.6.0.0.tar.xz for example is LZMA2:23 CRC64, for alpine is LZMA:23
So each tarball is generated in the system where hls is being built and the version of the tool can affect the final result 😟

[jneira]

I am thinking in removing the alpine tarballs, like 1.5.1. At the end those binaries will suffer the same template haskell bug we hit in hls , no?

[jneira]

In fact all tarballs are LZMA2:23 CRC64 but the alpine one

[hasufell]

If the current linux binaries are dynamic, this will not work.