Re: [PATCH 03/18] setup: stop using `the_repository` in `is_inside_git_dir()`
- From
Patrick Steinhardt <ps@pks.im>
- Date
- Apr 13, 2026, 05:47 UTC
- Message-ID
- <adyDiEW0-wRlgp6E@pks.im>
- In-Reply-To
- <CAOLa=ZS-bHiG0wyjwfzSbX4TgC_Jfpk2NXHAf=dUm6CvVxx80Q@mail.gmail.com>
On Thu, Apr 09, 2026 at 08:58:03AM -0400, Karthik Nayak wrote:
Show 10 quoted lines
> Patrick Steinhardt <ps@pks.im> writes: > > > Similar as with the preceding commit, `is_inside_git_dir()` determines > > whether the current working directory is located inside the gitdir of > > `the_repository`. Perform the same refactoring by dropping the caching > > mechanism and injecting the repository that shall be checked. > > > > The patch looks good. I do wonder if these functions need a new home, > 'setup.c' is fine, but it is also bloated.
I agree that they don't really make a ton of sense in "setup.c". In general, I would love to split up that file into two pieces:
- The pieces required to detect a repository and discover its exact
configuration. So scanning the current working directory and its
parent directories, paying attention to environment variables, and
detecting the repository format. - The pieces required to configure the repository based on the
findings.I would really like to completely detangle these two things from one another so that the detection logic will basically just give us a structure that contains all relevant information. And that info can then be passed into "repository.c" to create a new repository for us based on the info.
This would unlock a couple of benefits:
- Creating a repository will become "pure", as we only depend on
passed-in info and not on the environment anymore. - We can unify the creation of a repository into, instead of having
roughly the same logic in both "setup.c" and "repository.c". - We can remove some logic that essentially re-configures a repo
multiple times as we currently do in "setup.c", which should lead to
a much cleaner design.This is already thinking ahead quite a bit though. The next step would be to de-globalize some other functionality in "setup.c" first in the next patch series. But afterwards I'd indeed like to tackle this problem in the subsequent step.
Thanks!
Patrick