Re: Thoughts about the -m option of cherry-pick and revert

From: Junio C Hamano <gitster@pobox.com>
Date: Jun 21, 2024, 02:03 UTC
Message-ID: <xmqqa5jfoxvh.fsf@gitster.g>
In-Reply-To: <e60a8b1a-98c8-4ac7-b966-ff9635bb781d@haller-berlin.de>

Stefan Haller <lists@haller-berlin.de> writes:

Show 9 quoted lines

> I have two questions:
>
> - What are real-world scenarios where you would use a mainline number
>   other than 1? I could only come up with a single example myself, which
>   is that you have a topic branch, and right before merging it back to
>   main, you merge main into the topic branch; and then you merge it to
>   main with a fast-forward merge. If you then want to cherry-pick or
>   revert that topic, you'd have to use -m2 on that last merge from main.
>   Any other examples?

I do think your example is a real issue that is helped by using -m2; I do not think of any other cases offhand myself.

> - Wouldn't it make sense to default to -m1 when no -m option is given?
>   It seems that this would do the expected thing in the vast majority of
>   cases.

I do agree -m2 or higher would be rare when doing "git revert".

Given that the current behaviour was chosen to make sure that the user is aware that the commit being reverted/cherry-picked is a merge and has a chance to choose the right parent (as opposed to blindly picking the first parent that happened to be the right one by accident), I am not sure if it is prudent to change the behaviour.

If I were simplifying this, I would probably

 (1) disallow cherry-picking a merge (and suggest redoing the same
     merge, possibly after rebasing the copy of the merged history
     to an appropriate base as needed), and
 (2) allowing reverting a merge only wrt the first parent,

but that is a different story.