On Thu, Oct 17, 2019 at 12:28:25PM -0400, Peter Jones wrote:
> Currently if you do, for example:
>
> $ git worktree add path foo
>
> And "foo" has already been checked out at some other path, but the user
> has removed it without pruning, you'll get an error that the branch is
> already checked out. It isn't meaningfully checked out, the repo's
> data is just stale and no longer reflects reality.
>
> This makes it so that if nothing is present where a worktree is
> supposedly checked out, we ignore that the worktree exists, and let it
> get cleaned up the next time worktrees are pruned.
>
> (I would prune it instead, but prune isn't available from libgit
> currently.)
>
> Signed-off-by: Peter Jones <[email protected]>
> ---
> branch.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/branch.c b/branch.c
> index 579494738a7..60322ded953 100644
> --- a/branch.c
> +++ b/branch.c
> @@ -360,6 +360,9 @@ void die_if_checked_out(const char *branch, int
> ignore_current_worktree)
> wt = find_shared_symref("HEAD", branch);
> if (!wt || (ignore_current_worktree && wt->is_current))
> return;
> + if (access(wt->path, F_OK) < 0 &&
> + (errno == ENOENT || errno == ENOTDIR))
> + return;
I think this check is insuffient: even if the directory of the working
tree is not present, the working tree might still exist, and should
not be ignored (or deleted/pruned in the second patch).
See the description of 'git worktree lock' for details.
> skip_prefix(branch, "refs/heads/", &branch);
> die(_("'%s' is already checked out at '%s'"),
> branch, wt->path);
> --
> 2.23.0
>
