bug#67158: [PATCH] Repair tab-always-indent

[Eli Zaretskii]

> From: Aymeric Agon-Rambosson <aymeric.agon@yandex.com>
> Date: Tue, 14 Nov 2023 00:43:17 +0100
> 
> tab-first-completion does not work correctly at the moment, and 
> indent-for-tab-command must be modified in several ways to take 
> its meaning into account correctly :

Could you please post a recipe which reproduces the problems you
describe?

> Since syntax-after returns a list and not an integer, the forms 
> like (eql 2 syn) will always return nil. This was introduced by 
> commit c7234011518. We partially revert that commit, although it 
> would have been possible to solve this issue by wrapping 
> (syntax-after (point)) with syntax-class like so :
> 
> diff --git a/lisp/indent.el b/lisp/indent.el
> index 89de0a1d7d1..e5f2acdd33b 100644
> --- a/lisp/indent.el
> +++ b/lisp/indent.el
> @@ -171,7 +171,7 @@ prefix argument is ignored."
>      (let ((old-tick (buffer-chars-modified-tick))
>            (old-point (point))
>         (old-indent (current-indentation))
> -          (syn (syntax-after (point))))
> +          (syn (syntax-class (syntax-after (point)))))
>  
>        ;; Indent the line.
>        (or (not (eq (indent--funcall-widened indent-line-function) 'noindent))
> 
> 
> Feel free to change the commit if you prefer this way.
> 
> Then, the semantic of word-or-paren and word-or-paren-or-punct is 
> not correctly implemented : in the current state, if 
> tab-first-completion is set to word-or-paren, and if 
> tab-always-indent is set to complete, and we press tab in the 
> middle of a word, the word will get autocompleted despite the 
> docstring promising the contrary because :
> 
> The following form will correctly return nil :
> 
> (and (memq tab-first-completion
>            '(word word-or-paren word-or-paren-or-punct))
>      (not (memq 2 syn)))
> 
> But this one will return non-nil :
> 
> (and (memq tab-first-completion
>            '(word-or-paren word-or-paren-or-punct))
>      (not (or (eql 4 syn)
>               (eql 5 syn))))
> 
> Since syn is equal to (2) (we are within a word).
> 
> The constraints need to be cumulative, since they are evaluated 
> until one succeeds. So we simply cumulate them so that 
> word-or-paren cannot succeed where word could not, and 
> word-or-paren-or-punct cannot succeed when word-or-paren could 
> not.
> 
> This is my first contribution with email. I have tried to follow 
> the guidelines specified in CONTRIBUTE and Sending-Patches. Feel 
> free to change the commit message or ask me to do it. I have 
> already attributed the Copyright to the FSF because of a previous 
> contribution.

Thanks.

I added Stefan to this discussion.