Re: [PATCH] environment: move access to "core.attributesfile" into repo settings

From: Ayush Chandekar <ayu.chandekar@gmail.com>
Date: Mar 10, 2025, 17:21 UTC
Message-ID: <CAE7as+aSRuo9sFxSX8M66HB3EOH+_OwugAnAJfN800_6GiDqBQ@mail.gmail.com>
In-Reply-To: <xmqqwmcw97z2.fsf@gitster.g>

Show 13 quoted lines

> Another thing that felt awkward about this change is actually
> larger.  The current attributes globals are built around the notion
> that the functions involved work on a single set of attributes at a
> time.  Even in a single repository, when you are checking new contents
> into the object database and when you are checking objects out of
> the object database, you'd need to switch the direction manually,
> which means you always have two sets of attributes active that you
> can switch between (one is from the working tree and the other one
> is from the index, if I am recalling correctly).
>
> But step back and think.  What does it mean to make them belong to a
> repository instance?  Whose index and working tree does the attribute
> set that belongs to a repository that is not the_repository come from?

I'm trying my best to wrap my head around this. I definitely don't fully understand your review yet since I'm still quite new to the codebase. I get what you mean when you said which repo does it belong to. But in the long term, isn’t our goal to get rid of the_repository anyway? So at some point, wouldn't we need to either attach attributes to a repository or have the attribute set know about its repository?

Thanks, Ayush:)