>> We should fix this for Emacs-30.
>> [ And then add some backward compatibility hack (presumably using
>> an advice for when AUCTeX is used on Emacs<30). ]
> I attach my tentative proposal for Emacs<30, using an advice.
So, IIUC, you're counting on Someone™ to do the work on the Emacs side 🙂
> +(defcustom TeX-hack-former-dir-local-variables nil
> + "If non-nil, apply dir local vars under former mode names.
> +For example, directory local variables prepared for `latex-mode'
> +will be available in `LaTeX-mode' buffers.
> +It is highly discouraged to enable this option. Instead, please
> +rewrite mode names in .dir-locals.el whenever possible."
I hope we can arrange the code in Emacs-30 so that no changes are needed
in `.dir-locals.el`, in which case the advice should be activated by
default and we likely shouldn't plead with our users to rewrite their
`.dir-locals.el`.
> + (setq orig-former-parent (get former 'derived-mode-parent))
> + (put former 'derived-mode-parent former-parent)
> + ;; Very ugly hack to deceive `derived-mode-p'. Suppose
> + ;; - aaa-mode is an alias of bbb-mode
> + ;; - major-mode value is aaa-mode
> + ;; Then (derived-mode-p 'aaa-mode) returns nil on Emacs 28 and
> + ;; later, which makes `dir-locals-collect-variables' ignore a
> + ;; directory local variable entry for aaa-mode and look for an
> + ;; entry for bbb-mode instead. Hence we temporally make
> + ;; aaa-mode an alias of aaa-mode itself. :-(
> + (setq orig-former (symbol-function former))
> + (fset former former)
Hmm... this does look awful 🙁
Some questions/comments about the above comment:
- Indeed, I see that (derived-mode-p 'aaa-mode) returns nil in that
case, which is a bug. Do we have a bug-number for that?
Amusingly, it returns `aaa-mode` if `aaa-mode` is undefined :-)
- Is the problem new in Emacs-28 (if so, I guess it came about
in the rewrite which introduced `provided-mode-derived-p`).
- How exactly is it related to the issue of activating `latex-mode`
dir-local settings in `LaTeX-mode`?
I really should look at how best to solve the problem in Emacs-30, it
might give me some ideas about how to write a better workaround for
Emacs<30.
Do you think it would be OK for AUCTeX to declare that `latex-mode` is
*another* parent of `LaTeX-mode`? The idea is not to allow multiple
inheritance in `define-derived-mode`, but to allow it in `derived-mode-p`.
Stefan
