Position of functions in `completion-at-point-functions'

[Arash Esbati]

Hi all,

Emacs 30.0.50 has this in text-mode.el (line 155; line break added for
better legibility):

  (when (eq text-mode-ispell-word-completion 'completion-at-point)
    (add-hook 'completion-at-point-functions
              #'ispell-completion-at-point 10 t))

Since LaTeX-mode now derives from text-mode, the value of
`completion-at-point-functions' is this in a .tex file:

  (TeX--completion-at-point t ispell-completion-at-point
                            LaTeX--arguments-completion-at-point)

The problem is that `ispell-completion-at-point' kicks in also when
point is inside an argument and one actually expects key=val completion
for instance.  Since we require Emacs 27 now, we can cure this by
changing our addition in latex.el from:

  (add-hook 'completion-at-point-functions
            #'LaTeX--arguments-completion-at-point t t)

to something like:

  (add-hook 'completion-at-point-functions
            #'LaTeX--arguments-completion-at-point 5 t)

because docstring of `add-hook' says:

  For backward compatibility reasons, a symbol other than nil is
  interpreted as a DEPTH of 90.

Before making this change to 5 (which is an arbitrary choice): Is there
a kind of convention where major/minor modes should put their completion
functions?  Any pointer is appreciated.

Best, Arash