[bug]: There are cases where the error 'Invalid window id 1001' occurs when restoring a saved session.

[tadashi-aikawa]

Description

In certain use cases, such as restoring a session, an error seems to occur when the state of the saved session indicates that a scratch window created by no-neck-pain was visually absent.

From the source code, I confirmed that the issue can be avoided with the following fix.

no-neck-pain.nvim/lua/no-neck-pain/colors.lua

Line 133 in 5d91f31

local bufnr = vim.api.nvim_win_get_buf(win)

    -- my workaround
    if not vim.api.nvim_win_is_valid(win) then
        return
    end

    local bufnr = vim.api.nvim_win_get_buf(win)

However, I am not sure if this is the proper solution. If this approach is appropriate and a PR is needed, please feel free to let me know.

Thank you for maintaining such an excellent plugin!

Neovim version

0.10.x

Steps to reproduce

# Create a sandbox environment
mkdir -p ~/.config/no-neck-pain
cat > ~/.config/no-neck-pain/init.lua << 'EOF'
local lazypath = vim.fn.stdpath("data") .. "/lazy/lazy.nvim"
if not vim.uv.fs_stat(lazypath) then
  local out = vim.fn.system({
    "git",
    "clone",
    "--filter=blob:none",
    "https://github.com/folke/lazy.nvim.git",
    "--branch=stable",
    lazypath,
  })
end
vim.opt.rtp:prepend(lazypath)

require("lazy").setup({
  {
    "shortcuts/no-neck-pain.nvim",
    lazy = false,
    opts = {
      width = 140,
      autocmds = {
        enableOnVimEnter = true,
        enableOnTabEnter = true,
      },
      buffers = {
        colors = { background = "tokyonight-moon" },
      },
    },
  }
})
EOF
alias svim="NVIM_APPNAME=no-neck-pain nvim"

# Install lazy.nvim & no-neck-pain
svim
:q

# Create a sample file
echo "aaa" > aaa.txt

# Open files and make a session
svim
:e aaa.txt
:vnew aaa.txt
:mksession! mk.vim
:qa!

# Restore session
svim
:source mk.vim

Relevant log output

Error executing vim.schedule lua callback: .../nvim/lazy/no-neck-pain.nvim/lua/no-neck-pain/colors.lua:133: Invalid window id: 1001
stack traceback:
	[C]: in function 'nvim_win_get_buf'
	.../nvim/lazy/no-neck-pain.nvim/lua/no-neck-pain/colors.lua:133: in function 'init'
	...hare/nvim/lazy/no-neck-pain.nvim/lua/no-neck-pain/ui.lua:163: in function 'create_side_buffers'
	...re/nvim/lazy/no-neck-pain.nvim/lua/no-neck-pain/main.lua:108: in function 'callback'
	...vim/lazy/no-neck-pain.nvim/lua/no-neck-pain/util/api.lua:115: in function <...vim/lazy/no-neck-pain.nvim/lua/no-neck-pain/util/api.lua:113>

Self-service

• I'd be willing to fix this bug myself.

[shortcuts]

Hey, thanks for using the plugin and reporting the issue!!!

From the source code, I confirmed that the issue can be avoided with the following fix.

Indeed I think most nvim_win_get_buf or similar calls should be preceded by a valid check.

For that specific issue, is it enough to make the plugin work properly after your addition? I'd assume we should create a new window because the follow up logic might try to target a wrong window too

Feel free to go ahead with the implementation! Would be great if you could add a test case to tests/test_colors.lua as well, but if you can't I can add it to your pr

[tadashi-aikawa]

@shortcuts
Thank you for your reply!

For that specific issue, is it enough to make the plugin work properly after your addition? I'd assume we should create a new window because the follow

Unfortunately, I found that the workaround I suggested earlier results in an error under certain conditions.

Upon reviewing the code again, I realized it might be better to address the root cause, which seems to be that the first argument of colors.init is the invalid window id. Specifically, I moved this call inside the preceding if-then block.

no-neck-pain.nvim/lua/no-neck-pain/ui.lua

Line 163 in 5d91f31

colors.init(state.get_side_id(state, side), side)

            if
                wins[side].padding > _G.NoNeckPain.config.minSideBufferWidth
                and not state.is_side_enabled_and_valid(state, side)
            then
                vim.cmd(wins[side].cmd)

                state.set_side_id(state, vim.api.nvim_get_current_win(), side)

                if _G.NoNeckPain.config.buffers.set_names then
                    local exist = vim.fn.bufnr("no-neck-pain-" .. side)

                    if exist ~= -1 then
                        vim.api.nvim_buf_delete(exist, { force = true })
                    end

                    vim.api.nvim_buf_set_name(0, "no-neck-pain-" .. side)
                end

                if _G.NoNeckPain.config.buffers[side].scratchPad.enabled then
                    state.set_scratchPad(state, true)
                    ui.init_scratchPad(side, state.get_side_id(state, side))
                else
                    ui.init_side_options(side, state.get_side_id(state, side))
                end
+
+             colors.init(state.get_side_id(state, side), side)
            end

-           colors.init(state.get_side_id(state, side), side)

If state.is_side_enabled_and_valid is true, colors.init will not be called, but in such cases, it assumes that the highlight settings have already been applied.

After testing for a few minutes, no errors have occurred. I’ll continue using it for a while, and if it works without issues, I’ll create a PR. If you have any concerns or suggestions regarding this approach, please let me know 🙏

[tadashi-aikawa]

@shortcuts
I created a pull request #431 as the development has been going well over the past few days. I appreciate your review!