Re: [PATCH] [GSOC] cherry-pick: fix bug when used with GIT_CHERRY_PICK_HELP

From: ZheNing Hu <adlternative@gmail.com>
Date: Jul 23, 2021, 09:37 UTC
Message-ID: <CAOLTT8Ty47fyY7T3d68CYPKh9k+HAHsnCLJ=F0KaLm+0gp3+EQ@mail.gmail.com>
In-Reply-To: <xmqqo8au82i0.fsf@gitster.g>

Junio C Hamano <gitster@pobox.com> 于2021年7月23日周五 上午5:25写道：

Show 14 quoted lines

>
> "ZheNing Hu via GitGitGadget" <gitgitgadget@gmail.com> writes:
>
> > From: ZheNing Hu <adlternative@gmail.com>
> >
> > If we set the value of the environment variable GIT_CHERRY_PICK_HELP
> > when using `git cherry-pick`, CHERRY_PICK_HEAD will be deleted, then
> > we will get an error when we try to use `git cherry-pick --continue`
> > or other cherr-pick command.
>
> I think that the GIT_CHERRY_PICK_HELP is an implemention detail for
> various forms of rebase to use cherry-pick as a backend and not for
> use by end users.
>

But someone complain to me that the cherry-pick advice is not good enough. Think about a git newbie is cherry-picking a patch series containing several commits, E.g.

git cherry-pick dev~3..dev

And then he (she) will see these advice info: hint: after resolving the conflicts, mark the corrected paths hint: with 'git add <paths>' or 'git rm <paths>' hint: and commit the result with 'git commit'

After he resolving git conflict, execute 'git commit' according to the prompt, A terrible thing happened: CHERRY_PICK_HEAD is deleted by git and no errors are output. But in fact .git/sequencer still exists, Wait until he uses the cherry-pick command next time, the error appears:

error: cherry-pick is already in progress hint: try "git cherry-pick (--continue | --abort | --quit)" fatal: cherry-pick failed

So we should not encourage users to use git commit when git cherry-pick. It would be great if it could provide advice similar to rebase, like this:

Once you are satisfied with your changes, run

  git cherry-pick --continue

Show 9 quoted lines

> I strongly suspect that the right solution for the breakage is to
> unsetenv(GIT_CHERRY_PICK_HELP) in cmd_cherry_pick() without touching
> sequencer.c at all.
>
> It _is_ ugly that a helper that is responsible for emitting an
> advise message also makes a decision whether the pseudo-ref gets
> deleted or not, but a fix to that problem should be done byy making
> the logic for the decision less complex, not more.
>

May be you are right :)

> So, I am not enthused to see this change.

Thanks. -- ZheNing Hu