Git branch bug

Git branch bug

[Guido Ostkamp]

Hello,

I would like to report the following bug:

$ mkdir test
$ cd test
$ git init
Initialized empty Git repository in .git/
$ git branch experimental
fatal: Not a valid object name: 'master'.

So, it seems it is not possible to create a named branch on an empty 
repository.

I'm not subscribed to the list, so please keep me directly on CC for any 
reply, thanks.

Regards

Guido

Re: Git branch bug

[Johannes Schindelin]

Hi,

On Sat, 5 May 2007, Guido Ostkamp wrote:

> $ mkdir test
> $ cd test
> $ git init
> Initialized empty Git repository in .git/
> $ git branch experimental
> fatal: Not a valid object name: 'master'.
> 
> So, it seems it is not possible to create a named branch on an empty
> repository.

Actually, it is not possible to branch from a non-existing branch. So, 
this is somewhat expected from my POV.

Ciao,
Dscho

Re: Git branch bug

[Daniel Barkalow]

On Sat, 5 May 2007, Johannes Schindelin wrote:

> Hi,
> 
> On Sat, 5 May 2007, Guido Ostkamp wrote:
> 
> > $ mkdir test
> > $ cd test
> > $ git init
> > Initialized empty Git repository in .git/
> > $ git branch experimental
> > fatal: Not a valid object name: 'master'.
> > 
> > So, it seems it is not possible to create a named branch on an empty
> > repository.
> 
> Actually, it is not possible to branch from a non-existing branch. So, 
> this is somewhat expected from my POV.

This leaves open the question of how you make your initial commit in a 
branch that isn't master. I think the answer should be:

 $ git checkout -b experimental
 warning: You appear to be on a branch yet to be born.
 warning: Forcing checkout of HEAD.
 fatal: just how do you expect me to merge 0 trees?

Which should probably be:

 $ git checkout -b experimental
 warning: You appear to be on a branch yet to be born.
 warning: Putting you on a new branch yet to be born.

And leaving .git/HEAD pointing to refs/heads/experimental instead of 
refs/heads/master, with refs/heads/ still empty.

 $ git branch experimental

Should probably give a more informative error message, but branches with 
no commits that aren't the current branch has no existance at all, so it 
can't work to create them without switching to them.

	-Daniel
*This .sig left intentionally blank*

Re: Git branch bug

[Junio C Hamano]

Daniel Barkalow <barkalow@iabervon.org> writes:

> This leaves open the question of how you make your initial commit in a 
> branch that isn't master. I think the answer should be:
>
>  $ git checkout -b experimental
>  warning: You appear to be on a branch yet to be born.
>  warning: Forcing checkout of HEAD.
>  fatal: just how do you expect me to merge 0 trees?
>
> Which should probably be:
>
>  $ git checkout -b experimental
>  warning: You appear to be on a branch yet to be born.
>  warning: Putting you on a new branch yet to be born.
>
> And leaving .git/HEAD pointing to refs/heads/experimental instead of 
> refs/heads/master, with refs/heads/ still empty.

While I agree that is probably correct, it would not be useful
in real-life that much.  When you do not even have 'master', I
do not think there is much point of being able to create two
useless, yet-to-be-born branches.

Also, once you make a commit on 'master', you would not be able
to create a new disconnected root with your suggestion alone.  I
am not convinced that being able to start a disconnected root is
useful to begin with, but it is a different matter.  We allow
fetching and merging from a different repository to cause
disconnected roots to exist in a repository anyway, so I do not
see any reason to disallow it either.

Assuming that the ability to switch to a new yet-to-be-born
branch is useful, I think the right thing to do is:

 - Regardless of your HEAD state, you may want to have a way to
   create a yet-to-be-born branch and switch to it.  Perhaps
   "git checkout -e new" make HEAD point at refs/heads/new
   without creating one (I picked -e randomly for "empty").

 - When you are on a yet-to-be-born branch, "git checkout -b
   new" and "git checkout -b new HEAD" makes HEAD point at
   refs/heads/new without creating one.  I happen to consider
   this a not so useful corner case, but that is to make things
   consistent.

 - When you are on a yet-to-be-born branch, "git branch new" and
   "git branch new HEAD" is a nonsense operation.  We should
   clearly state that it is nonsense (I am agreeing with the
   last point in your message).

Re: Git branch bug

[Daniel Barkalow]

On Sat, 5 May 2007, Junio C Hamano wrote:

> Daniel Barkalow <barkalow@iabervon.org> writes:
> 
> > This leaves open the question of how you make your initial commit in a 
> > branch that isn't master. I think the answer should be:
> >
> >  $ git checkout -b experimental
> >  warning: You appear to be on a branch yet to be born.
> >  warning: Forcing checkout of HEAD.
> >  fatal: just how do you expect me to merge 0 trees?
> >
> > Which should probably be:
> >
> >  $ git checkout -b experimental
> >  warning: You appear to be on a branch yet to be born.
> >  warning: Putting you on a new branch yet to be born.
> >
> > And leaving .git/HEAD pointing to refs/heads/experimental instead of 
> > refs/heads/master, with refs/heads/ still empty.
> 
> While I agree that is probably correct, it would not be useful
> in real-life that much.  When you do not even have 'master', I
> do not think there is much point of being able to create two
> useless, yet-to-be-born branches.

The point is that you may not want to have a branch named "master", but 
rather want your first branch to be named something else. Your project 
could be such that "master" would be a confusing term to use for a 
default branch. Currently, it's awkward to start a project that avoids 
that particular name, which is what it seemed to me that the original 
poster was trying to do.

I don't think it's useful to be able to switch to a new yet-to-be-born 
branch from any other state, but I do think it's useful to be able to 
choose what the first branch is.

	-Daniel
*This .sig left intentionally blank*

Re: Git branch bug

[Guido Ostkamp]

> On Sat, 5 May 2007, Guido Ostkamp wrote:
>
>> $ mkdir test
>> $ cd test
>> $ git init
>> Initialized empty Git repository in .git/
>> $ git branch experimental
>> fatal: Not a valid object name: 'master'.
>>
>> So, it seems it is not possible to create a named branch on an empty
>> repository.
>
> Actually, it is not possible to branch from a non-existing branch. So,
> this is somewhat expected from my POV.

>From a git beginners point of view, I would expect that the 'master' 
branch is automatically created when performing the 'git init'. Why is 
this not the case?

Regards

Guido

Re: Git branch bug

[Johannes Schindelin]

Hi,

On Sat, 5 May 2007, Guido Ostkamp wrote:

> > On Sat, 5 May 2007, Guido Ostkamp wrote:
> > 
> > > $ mkdir test
> > > $ cd test
> > > $ git init
> > > Initialized empty Git repository in .git/
> > > $ git branch experimental
> > > fatal: Not a valid object name: 'master'.
> > > 
> > > So, it seems it is not possible to create a named branch on an empty
> > > repository.
> > 
> > Actually, it is not possible to branch from a non-existing branch. So,
> > this is somewhat expected from my POV.
> 
> From a git beginners point of view, I would expect that the 'master' 
> branch is automatically created when performing the 'git init'. Why is 
> this not the case?

Well, one side of the branch _is_ created. But you have to provide initial 
content, to make it a proper branch, from which you can actually branch 
off.

Having said that, and having read Daniels mail, I agree that this 
behaviour should be changed, along with the behaviour of "git checkout -b 
<branch>" in a fresh repo.

It might be the illogical thing, to allow branching of a 
yet-to-be-initialised branch, but it does not hurt either.

Ciao,
Dscho