Re: fail on regression test

[Arash Esbati]

Hi Keita,

Ikumi Keita <ikumi@ikumi.que.jp> writes:

> Since the recent update of style/siunitx.el, AUCTeX regression test
> began to fail with `LaTeX-count-ampersands-inserted-in-tabular' test in
> tests/latex/latex-test.el: [...]
> Could you take care of it?

Thanks for catching this.  I pushed a change (5f99bc2929) which should
fix the issue as you have suggested.

> And I'm wondering whether this update to siunitx.el is a bit too
> agressive. Although I haven't looked into the detail, this update drops
> support for siunitx.sty before v3, doesn't it?

Yes, it does.  The only thing I left over is fontification for macros
which are now deprecated.

> If so, AUCTeX users who don't update TeX Live installation frequently
> are left behind with incomplete support from the latest AUCTeX. I
> think siunitx.el should support both v2 and v3 of siunitx.sty, if
> possible. Is that difficult?

Yes, that will be the result.  As a general approach, I tend to stick
with the current TeXlive release and what packages support in the
current release.  Backward compatibility in styles isn't at top of my
list.

But reg. supporting older version: I don't think it is difficult, we
need a bare siunitx.el which provides common functions and package
options, and then check for

    \usepackage{siunitx}[=v2]

In the style hook, you would then say:

    (TeX-add-style-hook
     "siunitx"
     (lambda ()
       (if ('given-option' "=v2" or "=2021-04-09")
           (TeX-run-style-hooks "siunitx-v2")
         (TeX-run-style-hooks "siunitx-v3"))))

I have to check if the current parser for \usepackage also looks after
the optional argument after the package name.

As always, I would just wait and see if users complain about the current
state, and then think about a change.  But I'm not against a change a
priori if somebody wants to do it.

Does the above make sense?

Best, Arash