"LSP server (ruff) exited with status 2" and "LSP server (ty) exited with status 2" problem

[savagejin-dev]

I'm using the lsp plugin, and both ruff and ty are working fine. However, when the path of the opened file contains non-ASCII characters, error messages such as "LSP server (ruff) exited with status 2" and "LSP server (ty) exited with status 2" appear, and both ruff and ty stop working. This problem only occurs in the windows11 system, and everything works normally in archlinux.

I initially guessed that this might be a system encoding issue, as my Arch Linux is set to UTF-8, while Windows 11 uses CP936. I tried enabling UTF-8 support in Windows 11, but the problem persists. Could any friends provide a solution or insight to resolve this issue? Thank you.

[Konfekt]

Would you mind please sharing your setup? It makes much sense to add both

  g:lspServers->add({
    name: 'ty',
    filetype: ['python'],
    path: 'ty',
    args: ['server'],
    workspaceConfig: { python: { pythonPath: g:GetPythonPath() }},
  })

  g:lspServers->add({
    name: 'ruff',
    filetype: ['python'],
    path: 'ruff',
    args: ['server'],
    workspaceConfig: { python: { pythonPath: g:GetPythonPath() }},
  })

g:LspAddServer(lspServers)

I just wonder which one is taking care of which capabilities now? The complement each other, for example :LspCodeAction is only implemented by ruff, but which one delivers the diagnostics in :LspDiagShow?

[savagejin-dev]

Thank you friend
My setup:
var lspServers = [ { name: 'ruff', filetype: ['python'], path: 'ruff', args: ['server'], rootSearch: ["pyproject.toml", "ruff.toml", ".ruff.toml", ".git"] }, { name: 'ty', filetype: ['python'], path: 'ty', args: ['server'], rootSearch: ["pyproject.toml", "ty.toml", ".ty.toml", ".git"], } ] autocmd User LspSetup g:LspAddServer(lspServers)

Now, I found out that the function LspUriToFile and LspFileToUri in the file autoload/lsp/util.vim may cause the problem.
In line 139-140, the code makes percent encoding. Under normal circumstances, non-ASCII characters such as the Chinese character "查" would be translated to %E6%9F%A5. However, this piece of code does not run correctly. It treats "查" as a single character: the function char2nr converts "查" to the value 26597, while printf ultimately converts "查" to %67E5. I think this is where the problem lies.
Therefore, I have rewritten the code on lines 81 to 82 of the util.vim file as var uri_decoded: string = uri_decode(uri), and the code on line 140 as '=uri_encode(submatch(1))', 'g'). So far, everything is working correctly.
I am not familiar with vim9script, so I am not sure whether such modifications are applicable in all scenarios or if they will cause other new problems.

[Konfekt]

Please share your findings in a PR for expert inspection, @mmrwoods has been recently tackling encoding issues as well

[savagejin-dev]

I found that the behavior of builtin function char2nr has changed in vim9script https://github.com/vim/vim/issues/15324. Maybe this is the reason caurse the problem.