When the marks are deleted, <Tab> will jump to the deleted mark.

[vinoff]

Say you are on a .py file and you trigger a 'For' snippet. You immediately deleted the snippet and then you press Tab. You will be jumped to the deleted mark for the For snippet. This, of course, when you have tab setup to jump marks.

As far as I remember, before the "breaking changes", this did not happen.

[hrsh7th]

Please minimal repro.

[vinoff]

repro? you mean config? Here is my cmp config:

local has_words_before = function()
  local line, col = unpack(vim.api.nvim_win_get_cursor(0))
  return col ~= 0 and vim.api.nvim_buf_get_lines(0, line - 1, line, true)[1]:sub(col, col):match("%s") == nil
end

local luasnip = require("luasnip")
local cmp = require("cmp")

cmp.setup({

  -- ... Your other configuration ...
    snippet = {
        expand = function(args)
            require("luasnip").lsp_expand(args.body)
        end,
    },

  mapping = {

    -- ... Your other mappings ...

    ["<Tab>"] = cmp.mapping(function(fallback)
      if luasnip.expand_or_jumpable() then
        luasnip.expand_or_jump()
      -- elseif vim.fn.pumvisible() == 1 then
      --   feedkey("<C-n>")
      elseif cmp.visible() then
        cmp.select_next_item()
      elseif has_words_before() then
        cmp.complete()
      else
        fallback()
      end
    end, { "i", "s" }),

    ["<S-Tab>"] = cmp.mapping(function(fallback)
      if cmp.visible() then
        cmp.select_prev_item()
      elseif luasnip.jumpable(-1) then
        luasnip.jump(-1)
      else
        fallback()
      end
    end, { "i", "s" }),

    -- ... Your other mappings ...
  },

  sources = {
    { name = "luasnip" },
    { name = "nvim_lsp" },
    { name = "buffer" },
    { name = "path" },
  },
  -- ... Your other configuration ...
})

If "repro" means reproduction, here it is.

1. Open .py file, say test.py;
2. Type "for" and press tab to expand snippet;
3. After snippet is expanded, select the whole snippet (using visual mode) and delete it;
4. Go into insert mode and press tab;
5. Cursor jumps into the next (invisible) mark, that was deleted previously;

[hrsh7th]

Your <Tab> mapping checks luasnip.expand_or_jumpable() so it seems LuaSnip problem to me.