Buggy interaction of `g:vimtex_quickfix_autoclose_after_keystrokes` with command-line window.

[Aster89]

Description

When g:vimtex_quickfix_autoclose_after_keystrokes > 0, say n, moving the cursor n times in the command-line window causes an error.

Steps to reproduce

1. Ensure let g:vimtex_quickfix_autoclose_after_keystrokes > 0, say n
2. Open the quickfix window (even without compiling, technically, but you do need a log to make it open).
3. Open the command line window (via q:, q/ or whatever)
4. Move the cursor n times, and you'll see an error
5. Dismiss the error and move the cursor again n times, and you'll see the error again.

Expected behavior

VimTeX's quickfix closing normally upon n cursor movements.

Actual behavior

Every

Error detected while processing CursorMoved Autocommands for "*"..function <SNR>56_qf_autoclose_check:
line   19:
E11: Invalid in command-line window; :q<CR> closes the window:     cclose

Do you use a latexmkrc file?

Yes, but I'm pretty sure it's irrelevant

VimtexInfo

System info:
  OS: Arch Linux
  Vim version: VIM 9.0 (1-1302)
  Has clientserver: true
  Servername: VIM2

VimTeX project: main
  base: main.tex
  root: /home/enrico/deleteme
  tex: /home/enrico/deleteme/main.tex
  main parser: current file verified
  document class: book
  compiler: latexmk
    engine: -pdf
    options:
      -verbose
      -file-line-error
      -synctex=1
      -interaction=nonstopmode
    callback: 1
    continuous: 1
    executable: latexmk
  viewer: General
  qf method: LaTeX logfile

[lervag]

The issue with the :VimtexInfo was a peculiarity (#2638), but I don't really think it was a "real" issue. That is, doing q::VimtexInfo should never be part of a real workflow. However, I can see how this issue is real and I think it should be possible to adjust the function that causes this issue to not count movements within the command-line window. I'll look into it when I get the time!

Btw, :VimtexInfo still works as expected and the point of adding the info is to give me context such as OS, Vim version, etc. I've added the info from #2638 now as I think it must be safe to assume you are on the same system still. :)

[Aster89]

That is, doing q::VimtexInfo should never be part of a real workflow.

Why is that? I guess that claim is specific to VimtexInfo or to a category of : commands that have something in common, in which case, what is this thing in common?

Because you certainly can't refer to any q::any-command, because a very legit case of it is when some-command is the substitution command, e.g. one can do q:, then maybe a few ks to go to some line in the history, and then :s/pattern/subst/ to edit it.

[lervag]

I guess that claim is specific to VimtexInfo or to a category of : commands that have something in common, in which case, what is this thing in common?

Yes, I mean that :VimtexInfo specifically after q: is unnecessary. I regularly use the command-line window myself, and it is clearly a useful feature. I can see how substitutes with :s can be useful there, although I usually only use the command-line window to navigate history and make fast adjustments with normal-mode editing.

If you want to execute :VimtexInfo, you either use q: to find it in history and execute it, or you just execute it normally without the command-line window.

The "common thing" here is explained in :help E11:

							*E11* *E1188*
The command-line window is not a normal window.  It is not possible to move to
another window or edit another buffer.  All commands that would do this are
disabled in the command-line window.  Of course it _is_ possible to execute
any command that you entered in the command-line window.  Other text edits are
discarded when closing the window.

[lervag]

There, I think this is resolved now.