Re: [PATCH v7 1/1] pull: add ff-only option

From: Raymond E. Pasco <ray@ameretat.dev>
Date: Nov 24, 2020, 00:32 UTC
Message-ID: <C7B2K1BLPDIO.1D6XVJIWIG3UZ@ziyou.local>
In-Reply-To: <CAMP44s0r1vf7x-wHP-b6KQRU7mcPSZzcMRK2VqqxU_SJRDs+kg@mail.gmail.com>

On Mon Nov 23, 2020 at 7:08 PM EST, Felipe Contreras wrote:

Show 8 quoted lines

> I'd have no objection in refactoring this option so it's used for
> both, but this breaks existing behavior. A user that has
> "pull.ff=only", and does a "git pull --rebase" might expect it to
> work.
>
> My main interest is what happens when the user has not configured
> anything, which I defined with a "push.mode=default", any option (e.g.
> --rebase/--merge) would override that default.

I feel like the parsimonious change to make is simply defaulting the existing "pull.ff" to "only". I think someone who has set "pull.rebase" expects pull --rebase behavior just as much as someone who passes --rebase on the command line. The issue in question is what someone who has not made any changes to the settings expects to happen with a plain "git pull", and I certainly agree that people who are not power users expect a fast-forward (I try not to force my opinions on workflow or style when onboarding people to Git, but I do always recommend "pull.ff=only" because I know this is a perennial pitfall).

The problem is that, as it stands now, this would just leave the user with a cryptic error message ("Not possible to fast-forward, aborting") when they wanted to see remote changes. I think this might warrant an even more expanded message than the short one in this patch, but I'm not sure exactly what it should say - there are a few things the user might expect, but an error message isn't the best place for a crash course.