Re: [GSOC PATCH v2 2/2] attr: use `repo_settings_get_attributesfile_path()` and update callers

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 10, 2025, 23:08 UTC
Message-ID: <xmqqy0xc4h7b.fsf@gitster.g>
In-Reply-To: <CAOLa=ZT=zGTF2DLEy9VjXhcUN3wEi7_R=8O6nV-TtBXKT=ENXg@mail.gmail.com>

Karthik Nayak <karthik.188@gmail.com> writes:

Show 11 quoted lines

> Ayush Chandekar <ayu.chandekar@gmail.com> writes:
>
>> Update attribute-related functions to retrieve the "core.attributesfile"
>> configuration via the new repository-scoped accessor
>> `repo_settings_get_attributesfile_path()`. This improves behaviour in
>> multi-repository contexts and aligns with the goal of minimizing
>> reliance on global state.
>>
>
> We should also talk about the modifications made to pass around the
> repository struct.

Yes. We first should justify if it makes sense to cram attribute set into the repository object and pass it around in the first place. Many index-state related functions do not pass repository around because they work on index-state, so index-state is passed around instead. Perhaps the attribute subsystem should be the same way, in that their globals should belong to its own abstraction that is smaller scale than a repository object (it is permissible to have such an attribute-set object know about which repository instance it is related to, though).