rebase: do not try to fixup a skipped commit

[dscho]

When running an interactive rebase, it is possible that the target commit of a fixup! commit could not be applied due to merge conflicts (or because it has been upstreamed). In this instance, going forward with a fixup will amend the wrong commit.

Let's prevent that.

The specifics how to prevent that are not set in stone. The easiest solution would probably be to refuse to even start with the fixup command, to complain loudly, and to re-issue the command for the user to edit.

The tricky part will be to detect that situation because the target commit might have failed to apply cleanly, requiring manual fixing up followed by git add -u && git rebase --continue, but those manual fixups might have resulted in no more changes to be committed, either.

[dscho]

See git-for-windows#3981 (comment) for a real-world scenario where this bug caused real problems.

[derrickstolee]

By "refusing to start the fixup!" do you instead mean to keep the commit as a pick and not reorder it?

[dscho]

No, at the time we could reorder it, we won't have the information whether its target pick caused problems and was skipped by the user.

[samkergachu]

Hey dscho. Is anyone working on this, I would like to take it up as my outreachy microproject

[dscho]

@samkergachu I am unaware of anybody else working on this, so go right ahead!

[vinayakdsci]

Hey @dscho! Could you help me take up this issue as my GSoC microproject?
I would be very grateful.

[dscho]

Hey @dscho! Could you help me take up this issue as my GSoC microproject? I would be very grateful.

Unfortunately, I am pressed for time. You could reach out to the Git mailing list for guidance, though. You can maximize your chances by being specific about your needs.

[vinayakdsci]

Absolutely, I can understand. I will ask on the mailing list. Thanks!

[vinayakdsci]

Hey Cheskaqiqi! Yes, I am on it.

…

On Wed, Feb 8, 2023, 12:22 Cheskaqiqi ***@***.***> wrote: Hey dscho. Is still anyone working on this . I would like to have a try. — Reply to this email directly, view it on GitHub <#1321 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AY4VBINV5TCBO7KUD4UFMJLWWM7BVANCNFSM56K5A3PA> . You are receiving this because you commented.Message ID: ***@***.***>