Default normal mode CursorLine highlight not used

[budswa]

Hi. When starting Neovim my normal CursorLine highlights are used but after going into insert or visual mode and returning into normal mode or if I press escape after startup, staying in normal mode, the CursorLine highlight isn't properly applied.

On startup:

After entering insert/visual mode and returning to normal mode or pressing escape:

The result of me running :hi CursorLine shows me that its guibg value is Grey40, rather than the expected hex value. The highlight is defined by my colorscheme. I believe what I'm seeing is some sort of fallback highlight, which isn't based on the CursorLine highlight.

Thanks

[fitrh]

It seems that nvim_get_hl_id_by_name gives strange results when trying to extract from a linked hl group, in this case, your theme links CursorLine to CursorColumn.

@mvllow any reason why you use nvim_get_hl_id_by_name instead of nvim_get_hl_by_name ?

@budswa BTW, before I make a PR, you can try the following patch to fix this behavior or you can define CursorLine instead of linking it

diff --git a/lua/modes/util.lua b/lua/modes/util.lua
index bc657c8..43c8714 100644
--- a/lua/modes/util.lua
+++ b/lua/modes/util.lua
@@ -47,6 +47,11 @@ function util.hl(group, color)
 end

 function util.get_fg_from_hl(hl_name, fallback)
+   local ok, hl = pcall(vim.api.nvim_get_hl_by_name, hl_name, true)
+   if ok and hl.background then
+       return ('#%06x'):format(hl.background)
+   end
+
    local id = vim.api.nvim_get_hl_id_by_name(hl_name)
    if not id then
        return fallback
@@ -61,6 +66,11 @@ function util.get_fg_from_hl(hl_name, fallback)
 end

 function util.get_bg_from_hl(hl_name, fallback)
+   local ok, hl = pcall(vim.api.nvim_get_hl_by_name, hl_name, true)
+   if ok and hl.background then
+       return ('#%06x'):format(hl.background)
+   end
+
    local id = vim.api.nvim_get_hl_id_by_name(hl_name)
    if not id then
        return fallback

[mvllow]

@mvllow any reason why you use nvim_get_hl_id_by_name instead of nvim_get_hl_by_name ?

Didn't know the latter existed. Thank you for debugging this issue 🥰