Re: [PATCH v2 0/3] Ensure unique worktree ids across repositories

From: shejialuo <shejialuo@gmail.com>
Date: Dec 2, 2024, 11:46 UTC
Message-ID: <Z02eKA8i51QyXgJ-@ArchLinux>
In-Reply-To: <xmqqed2qkfn0.fsf@gitster.g>

On Mon, Dec 02, 2024 at 11:00:51AM +0900, Junio C Hamano wrote:

Show 10 quoted lines

> Caleb White <cdwhite3@pm.me> writes:
> 
> > - Add the worktree id to `worktree list` output
> 
> I have always thought that we deliberately hid the "ID" from the end
> user's view because it is a mere implementation detail, and used the
> filesystem path of the worktree directory instead to identify each
> instance of the worktree in the end-user interaction.  It is unclear
> why this change is a good idea from that point of view.
>

I have a discussion with Caleb in the first version. Because appending a hash / random number will cause the worktree id has more digits, the user cannot easily use the following commands to make a ref in the main worktree point to linked worktree ref:

    ```sh
    git symbolic-ref refs/heads/foo \
        worktrees/<worktree id>/refs/worktree/foo
    ```

And I expressed my concern about above situation, the user types the extra hash to do above. So, Caleb decides to list the worktree id.

Actually, this usage should not be common. But when implementing the consistency check for files backend, you have told me that the above situation could happen. And you have said we _deliberately_ hide the "ID" from the end user's view.

But cross-ref operations between worktrees must explicitly specify the worktree id, so I am wondering whether we should allow the user do cross-ref operations in the first place:

  1. main worktree symref points to linked worktree ref.
  2. A linked worktree ref points to another linked worktree ref.

Thanks, Jialuo