"Branch exists" error while trying to rename a non-existing branch to an existing one

"Branch exists" error while trying to rename a non-existing branch to an existing one

[Kaartic Sivaraam]

Hello all,

I recently got the following error message by change as a result of the
command,

    $ git branch -m no-branch master
    fatal: A branch named 'master' already exists.

Note: no-branch is an hypothetical branch that doesn't exist.

Shouldn't I get a 'no-branch' doesn't exist before that? Wouldn't this
behaviour make the users search for the non-existing 'no-branch' in
their repo?

I tried digging the implementation a little and what I could interpret
from it is,

    * only the validity of new branch name (master, in the above case)
    is checked
    * checking for existence of the branch being renamed(no-branch) is
    not done at all. It seems to be left to the lower level commands to
    identify.

I'm puzzled by seeing this. Why isn't there any check for the existence
of the branch being renamed and warning the user about that first?

-- 
Kaartic

Re: "Branch exists" error while trying to rename a non-existing branch to an existing one

[Junio C Hamano]

Kaartic Sivaraam <kaarticsivaraam91196@gmail.com> writes:

> I recently got the following error message by change as a result of the
> command,
>
>     $ git branch -m no-branch master
>     fatal: A branch named 'master' already exists.
>
> Note: no-branch is an hypothetical branch that doesn't exist.
>
> Shouldn't I get a 'no-branch' doesn't exist before that?

This is borderline "meh" at least to me.  An argument against a
hypothetical version of Git that "fixes" your issue would be that no
matter what the source of renaming is, as long as 'master' exists,
"branch -m" shouldn't overwrite it, and it is a good thing to remind
the user that 'master' exists and the user meant to rename it to
something else.

Of course you can make the error message three-way, but I do not
think it is worth it.

Re: "Branch exists" error while trying to rename a non-existing branch to an existing one

[Kaartic Sivaraam]

On Sun, 2017-07-09 at 11:57 -0700, Junio C Hamano wrote:
> This is borderline "meh" at least to me.  An argument against a
> hypothetical version of Git that "fixes" your issue would be that no
> matter what the source of renaming is, as long as 'master' exists,
> "branch -m" shouldn't overwrite it, and it is a good thing to remind
> the user that 'master' exists and the user meant to rename it to
> something else.
> 
I'm not against the fact of reminding the user about an existing
branch. I'm with the fact that, warn him when he really has to care
about a branch being overwritten i.e., when he tries to rename an
"existing" branch to one that refers to another existing branch.

I found this behaviour odd as I try to relate it with the 'mv' command.
It's behaviour is as follows,

    $ ls
    file  some_file
    $ mv nothing file
    mv: cannot stat 'nothing': No such file or directory

If I haven't missed anything the following patch seems to fix the
problem,

diff --git a/builtin/branch.c b/builtin/branch.c
index 48a513a84..2869aaca8 100644
--- a/builtin/branch.c
+++ b/builtin/branch.c
@@ -476,7 +476,10 @@ static void rename_branch(const char *oldname, const char *newname, int force)
         */
        clobber_head_ok = !strcmp(oldname, newname);
 
-       validate_new_branchname(newname, &newref, force, clobber_head_ok);
+       if(ref_exists(oldref.buf))
+               validate_new_branchname(newname, &newref, force, clobber_head_ok);
+       else
+               die(_("Branch '%s' does not exist."), oldname);
 
        reject_rebase_or_bisect_branch(oldref.buf);


-- 
Kaartic