Re: [PATCH] Add option -b/--branch to clone for select a new HEAD

["Björn Steinbrink" <B.Steinbrink@gmx.de>]

On 2009.08.26 15:53:58 +0400, Kirill A. Korinskiy wrote:
> At Wed, 26 Aug 2009 00:36:37 +0200,
> Björn Steinbrink <B.Steinbrink@gmx.de> wrote:
> 
> 
> > > @@ -518,7 +521,21 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
> > >  
> > >  		mapped_refs = write_remote_refs(refs, refspec, reflog_msg.buf);
> > >  
> > > -		remote_head = find_ref_by_name(refs, "HEAD");
> > > +		if (option_branch) {
> > > +			strbuf_addf(&branch_head, "%s%s", src_ref_prefix, option_branch);
> > > +
> > > +			remote_head = find_ref_by_name(refs, branch_head.buf);
> > > +		}
> > > +
> > > +		if (!remote_head) {
> > > +			if (option_branch)
> > > +				warning("Remote branch %s not found in upstream %s"
> > > +					", using HEAD instead",
> > > +					option_branch, option_origin);
> > > +
> > > +			remote_head = find_ref_by_name(refs, "HEAD");
> > > +		}
> > > +
> > >  		head_points_at = guess_remote_head(remote_head, mapped_refs, 0);
> > 
> > This would still pick refs/heads/master if refs/heads/master and
> > refs/heads/<branch> reference the same commit. That's due to the check
> > in guess_remote_head() which prefers refs/heads/master over all other
> > refs. While this is acceptable for the HEAD lookup, I'd treat that as a
> > bug for this new option.
> > 
> 
> My english is not a good and I don't understand it, sorry.

guess_remote_head() compares the object ids from remote_head and all of
the remote's refs to guess which is the right one.

Let's say that the repo has:

refs/heads/master: object1
refs/heads/foo: object2
refs/heads/bar: object1

If you do "git clone -b foo ...", then remote_head->old_sha1 will be
"object2". guess_remote_head() compares that to all the remote heads. In
this case, it will find refs/heads/foo (as expected).

But when you do "git clone -b bar", then remote_head->old_sha1 will be
"object1". And guess_remote_head() will then take refs/heads/master,
as it prefers that one.

doener@atjola:h $ mkdir a; cd a; git init
Initialized empty Git repository in /home/doener/h/a/.git/
doener@atjola:a (master) $ git commit --allow-empty -m init
[master (root-commit) a7a0b54] init
doener@atjola:a (master) $ git branch bar
doener@atjola:a (master) $ git checkout -b foo
Switched to a new branch 'foo'
doener@atjola:a (foo) $ git commit --allow-empty -m on_foo
[foo 375047e] on_foo
doener@atjola:a (foo) $ cd ..
doener@atjola:h $ (git clone -b foo a foo; cd foo; git branch)
Initialized empty Git repository in /home/doener/h/foo/.git/
* foo
doener@atjola:h $ (git clone -b bar a bar; cd bar; git branch)
Initialized empty Git repository in /home/doener/h/bar/.git/
* master


That said, I actually wonder why you don't simple set HEAD in the
original repo so that you get whichever branch you want by default
anyway.

Björn