Re: [PATCH v6 03/11] replay: stop using `the_repository`
- From
Patrick Steinhardt <ps@pks.im>
- Date
- Dec 2, 2025, 18:47 UTC
- Message-ID
- <aS8zvg1WquqJiC4H@pks.im>
- In-Reply-To
- <CABPp-BH7E1Bh2g0vR3T4NEsv34DvFQPzMuJSsqtOAaWY-fFCxg@mail.gmail.com>
On Wed, Nov 19, 2025 at 11:01:29PM -0800, Elijah Newren wrote:
Show 28 quoted lines
> On Mon, Oct 27, 2025 at 4:34 AM Patrick Steinhardt <ps@pks.im> wrote: > > > > In `create_commit()` we're using `the_repository` even though we already > > have a repository passed to use as an argument. Fix this. > > I feel like I've fixed this multiple times, but it keeps coming back. > In fact, I have this same fix locally in my replay-edit work. Thanks > for fixing it. > > > > > Note that we still cannot get rid of `USE_THE_REPOSITORY_VARIABLE`. This > > is because we use `DEFAULT_ABBREV and `get_commit_output_encoding()`, > > both of which are stored as global variables that can be modified via > > the Git configuration. > > Indeed. > > Going on a tangent for a second...I feel like I've had to remove > "the_repository" from builtin/replay.c multiple times. In my local > replay-edit work, I actually added a "#define the_repository > DO_NOT_USE_THE_REPOSITORY" in builtin/replay.c, after all the header > includes, because the_repository isn't what builtin/replay.c is using, > it's these other two things that are also only included if > USE_THE_REPOSITORY_VARIABALE is defined. That obviously doesn't need > to be part of your series, but what would you think if I were to > submit that? Is it too ugly/weird of a way to avoid the_repository > being added back to builtin/replay.c so we can stop having to remove > it again?
It does feel somewhat ugly, and the better solution would of course be to refactor both `DEFAULT_ABBREV` and `get_commit_output_encoding()` to accept a repository as input. But if I remember correctly that was a nontrivial endeavour, so your proposed hack might be the next-best solution.
Patrick