I see some tests are failed. However, I have no idea why scheduling this action would fail so many tests? In my use case, it works well. I think this might because vim.schedule makes the write action async and the tests are assumed it is synchronized?

Can you give me the steps to reproduce the issue? I can't reproduce it so can't really help.

This is a problem that's are really hard to reproduce.

The same config can yield the "E315: ml_get: invalid lnum: 1" on macos, but not on linux. And I cannot create a minimal config either. May need to spend more time on this to find what's triggering the issue.

Another interesting thing is that: :h E315 shows that"

							*E315*  >
  ml_get: invalid lnum: {number}

This is an internal Vim error.  Please try to find out how it can be
reproduced, and submit a |bug-report|.

So this might point to a C source code error.

I was encountering this error on task creation until I set org_agenda_files and org_default_notes_file. I think it was because this plugin was trying to create refile.org in a location it could not. I was also using noice.nvim and, for some reason, it just spawns an empty confirm window when I tried to make a new agenda task. Confirming/closing the empty window showed this error, broke noice to no longer work for the rest of the session, and failed to create the task.

It was not until I set these two variables the error went away.

Just thought I should comment here if someone else finds this error, too!

I don’ t use noice.

Add some context to see if I can help, but the sad thing is that I cannot reproduce the problem…

I encounter this problem when trying to use org-edit-special to edit a SQL code block, and treesitter-sql grammar is installed, and I use efm language server for SQL (to use sql-formatter).

If I am editing another code block (e.g. not SQL), then I do not encounter that error. And if I don’t use efm with SQL, the problem disapperas too. There are too many variables to control and I cannot reproduce.