"Plugin telescope-nvim not found", even when I can do :packadd telescope-nvim

[viperML]

Hello!

I'm currently rewriting all of nixpkgs' wrappers around neovim, so I just started to try lz.n...

The issue is simple: after configuring lz with the following:

  {
    "telescope-nvim",
    cmd = "Telescope",
    after = function()
      vim.notify("Telescope loaded")
    end,
  },

And running :Telescope, I will get an error saying Plugin telescope-nvim not found, even when I can do :packadd telescope-nvim.

Interestingly, removing the cmd will work fine!

You can run my neovim wrapper with: nix build github:viperML/dotfiles?ref=lz#neovim, the relevant files are here:

• Wrapper: (you might want to inspect the resulting nvim binary if needed) https://github.com/viperML/dotfiles/blob/lz/modules/wrapper-manager/neovim/default.nix
• Init: https://github.com/viperML/dotfiles/blob/lz/modules/wrapper-manager/neovim/viper-init-plugin/lua/viper/init.lua#L18

Thank you for your time

[mrcjkb]

Hey 👋

The plugin's name is "telescope.nvim", not telescope-nvim (in nixpkgs, the pname is "telescope.nvim", which is the plugin's name that should be passed to packadd).

I have no idea why packadd telescope-nvim is working for you.

[viperML]

I'm not using nixpkgs' vimPlugins, but rather packaging them by myself, and I'm using telescope-nvim as the name of the directory in opt:

$ nix build "github:viperML/dotfiles?ref=lz#neovim.packDir"

$ file ./result/pack/viper-pack/opt/telescope-nvim
./result/pack/viper-pack/opt/telescope-nvim: symbolic link to /nix/store/hl0nv3rg24vy88fn2hawc99s14w0xd7b-telescope-nvim-0.1.8

I don't know know if there can be problems by not matching the dir name to upstream's repository, but as I said, :packadd telescope-nvim works for this setup.

[mrcjkb]

The Plugin <plugin> not found error indicates that the plugin may have been loaded, but not properly unregistered from the cmd handler.

I noticed you have require("lzn-auto-require.loader").register_loader() in your init.lua. Does it work if you comment that line?

[viperML]

I'm afraid it's the same...

[mrcjkb]

🤔 where is load being called a second time in your config?
Calling load { } (with an empty list) is definitely undefined behaviour.
It would probably be better if lz.n threw an error in that case.

[viperML]

I was preparing to set-up a lazy plugin (premature optimization eh?), and these were the two calls https://github.com/viperML/dotfiles/blob/cf358a0b91d4cafcc9bcf7e0dd63ae7f8eca8f11/modules/wrapper-manager/neovim/default.nix#L102-L106 and https://github.com/viperML/dotfiles/blob/cf358a0b91d4cafcc9bcf7e0dd63ae7f8eca8f11/modules/wrapper-manager/neovim/viper-init-plugin/lua/viper/init.lua#L15-L32 . I guess I can merge them without any hassle.

Calling load { } (with an empty list) is definitely undefined behaviour. It would probably be better if lz.n threw an error in that case.

So the problem is calling load a second time, or calling load on the empty list?

In any case, I agree that throwing an error and exiting is the better UX.

[mrcjkb]

So the problem is calling load a second time, or calling load on the empty list?

I'd have to check, but I think it was intended to either be called once on a list of specs or multiple times on a single spec.
I didn't think someone would want to do both, but maybe I can make that possible.

[mrcjkb]

This should be fixed when #56, which removes the lz.n.state module as a redundant source of truth, is merged.

[viperML]

Thank you for your time