threads / bug / 35616

[BUG] rebase --onto doesn't abort properly

Subject: [BUG] rebase --onto doesn't abort properly

## tl;dr

3 messages between Jan 6, 2014 and Jan 6, 2014.

replies: 2people: 2as markdown or json

Ramkumar Ramachandra· Jan 6, 2014, 17:42 UTC · lore

Hi,

On the latest git, I noticed that a rebase --onto doesn't abort properly. Steps to reproduce:

  # on some topic branch
  $ git rebase --onto master @~10
  ^C # quickly!
  $ git rebase --abort
  # HEAD is still detached

I tried going back a few revisions, and the bug seems to be very old; I'm surprised I didn't notice it earlier.

Are others able to reproduce this?

Thanks.

Ram

Junio C Hamano· Jan 6, 2014, 18:44 UTC · re: Ramkumar Ramachandra · lore

Re: [BUG] rebase --onto doesn't abort properly

Ramkumar Ramachandra <artagnon@gmail.com> writes:

Show 10 quoted lines

> Hi,
>
> On the latest git, I noticed that a rebase --onto doesn't abort
> properly. Steps to reproduce:
>
>   # on some topic branch
>   $ git rebase --onto master @~10
>   ^C # quickly!
>   $ git rebase --abort
>   # HEAD is still detached

I do not think --abort was designed to abort an uncontrolled stop like ^C in the first place. To allow that kind of "recovery", you need to teach "rebase" to first record the state you would want to go back to before doing anything, but then what happens if the ^C comes when you are still in the middle of doing so?

Ramkumar Ramachandra· Jan 6, 2014, 18:55 UTC · re: Junio C Hamano · lore

Re: [BUG] rebase --onto doesn't abort properly

Junio C Hamano wrote:

> I do not think --abort was designed to abort an uncontrolled stop
> like ^C in the first place.

Why not? All it requires is a reset --hard to .git/rebase-apply/head-name, as usual, no?

> To allow that kind of "recovery", you
> need to teach "rebase" to first record the state you would want to
> go back to before doing anything, but then what happens if the ^C
> comes when you are still in the middle of doing so?

I'm a bit puzzled: doesn't rebase write_basic_state() as soon as it starts? It's true that we can't recover from a ^C before that, but I would expect to be able to recover after a ^C somewhere in the middle.