GNU bug report logs - #74412
31.0.50; Treesit-outline conflict with outli in Emacs 31

Previous Next

Full log

Message #70 received at 74412 <at> debbugs.gnu.org (full text, mbox):

From: Yuan Fu <casouri <at> gmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 74412 <at> debbugs.gnu.org, liaotx2 <at> gmail.com, Daan Ro <daanturo <at> gmail.com>
Subject: Re: bug#74412: 31.0.50; Treesit-outline conflict with outli in Emacs
 31
Date: Sat, 23 Nov 2024 12:46:18 -0800

> On Nov 21, 2024, at 1:07 AM, Eli Zaretskii <eliz <at> gnu.org> wrote:
> 
>> Date: Thu, 21 Nov 2024 15:16:14 +0700
>> From: Daan Ro <daanturo <at> gmail.com>
>> Cc: "liaotx2 <at> gmail.com" <liaotx2 <at> gmail.com>, 
>> "74412 <at> debbugs.gnu.org" <74412 <at> debbugs.gnu.org>
>> 
>> Before treesit.el went to set those variables, there have been already a
>> bunch of packages based on outline-minor-mode for partitioning code into
>> sections using comments only by highlighting headers and providing
>> navigation, folding, etc. inspired by emacs-lisp-mode's outline-regexp
>> settings.
>> 
>> Such ones enable a feel of "literate" programming by bringing
>> outline-mode features to code section delimited by headings (special
>> comments like ";;;", ";; *") using just comments regardless of code
>> semantic.
>> 
>> The above is my use of outline-minor-mode, maybe other people have
>> different use cases. treesit.el makes the configuration harder by the
>> additional variable setting. I wish there's a hook like
>> treesit-after-major-mode-setup-hook to let people fine-tune or even undo
>> settings, but of course treesit-disable-features to prevent any
>> execution from the start is still better.
> 
> There's a mode hook run by every mode based on treesit.el, so if the
> settings by treesit.el are inappropriate, and the major mode itself
> somehow decided to use them anyway, and the user doesn't like that,
> the user should be able to countermand the treesit.el settings in that
> mode hook.
> 
> In addition, packages that use outline-minor-mode in ways that are
> incompatible with what treesit.el does should be able to reset these
> variables to nil or to some other function, as they need.
> 
> The purpose of treesit.el doing this is to make it easier for
> treesit-based modes to set up their support for outline-minor-mode, so
> that we won't need to perform the same settings in every major mode
> based on tree-sitter.  I don't see anything wrong here.  Modes and
> packages which for some reason don't like that should override these
> values, and that is easy enough to do without adding any user options.
> Let's keep in mind that every new user option makes Emacs more complex
> to use for users, which need to learn about the option and modify
> their init files.  Instead, modes and packages should DTRT by default
> without requiring users to get their (mode's) act together.


The issue is usually that tree-sitter automatically sets outline-search-function, and the package that uses outline-minor-mode only sets outline-regexp and doesn’t set outline-search-function to nil. Since outline prioritizes outline-search-function over outline-regex, it appears that tree-sitter’s outline setting overrides that of the package.

Most outline-related packages don’t set outline-search-function to nil since it’s usually nil by default. So tree-sitter’s behavior will surely break these packages. But since tree-sitter’s setup runs before minor modes, we can’t really tell if the user will later enable a minor mode that tree-sitter’s setup might break.

TRT here is probably for the packages to explicitly set outline-search-function to nil.

Yuan

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.