Re: [BUG] Wrong worktree path when using multiple worktree
- From
Mike Rappazzo <rappazzo@gmail.com>
- Date
- Nov 4, 2015, 02:23 UTC
- Message-ID
- <CANoM8SXKmqq5qpv7vFobyAj9mL0wxFPqWeTt5oJT1EP0+=NqwA@mail.gmail.com>
- In-Reply-To
- <CANoM8SVOPpOvnykGxfw+dn1Twwe8S=WKfePogx+GRQeAjn6D7g@mail.gmail.com>
On Tue, Nov 3, 2015 at 5:27 PM, Mike Rappazzo <rappazzo@gmail.com> wrote:
Show 36 quoted lines
> On Tue, Nov 3, 2015 at 11:58 AM, Nicolas Morey-Chaisemartin
> <nicolas@morey-chaisemartin.com> wrote:
>> Hi,
>>
>> There seem to be an issue with the path computed for a worktree when multiple worktree were created (using git worktree)
>> In my Setup, I have 3 repos:
>> A/repo (main One)
>> A/repo-patches (worktree, using branch dev)
>> B/repo (worktree, using branch next)
>>
>> I'm working in A/repo-patches an run:
>> $ git checkout next
>> fatal: 'next' is already checked out at 'A/repo-patches'
>>
>> Which is partially true but not completely.
>> I looked a bit in the code and found that the issue comes from here (get_linked_worktree):
>> if (!strbuf_strip_suffix(&worktree_path, "/.git")) {
>> strbuf_reset(&worktree_path);
>> strbuf_addstr(&worktree_path, absolute_path("."));
>> strbuf_strip_suffix(&worktree_path, "/.");
>> }
>> Because it wrongfully assumes that I am in the linked worktree.
>> I checked in the .git/worktree files and couldn't see anything that actually points to where the repo are setup.
>>
>> Nicolas
>
> This is code that I worked on, but I am unable to reproduce it locally
> just yet. Before I dig too deep, could you report the contents of
> "A/repo/.git/worktrees/repo-patches/gitdir" (or similar)? There is
> another issue reported[1] where the contents of the gitdir for a
> linked worktree are overwritten in some cases. A fix for this is
> being worked on (see discussion).
>
> In the mean time, I will continue to try and reproduce.
>
> [1]: http://thread.gmane.org/gmane.comp.version-control.git/280307Followup: I was able to reproduce the error when I tried to simulate the aforementioned bug. Here is a test which I added to t2027 (not intended to publish):
+test_expect_success '"checkout" branch already checked out' ' + git worktree add -b linked_1_br linked_1 master && + git worktree add -b linked_2_br linked_2 master && + echo ".git" > .git/worktrees/linked_2/gitdir && + test_must_fail git -C linked_1 checkout linked_2_br +' +
Test run result: -------------------- Preparing linked_1 (identifier linked_1) HEAD is now at 2519212 init Preparing linked_2 (identifier linked_2) HEAD is now at 2519212 init fatal: 'linked_2_br' is already checked out at '/Users/mike/code/git-source/t/trash directory.t2027-worktree-list/linked_1'