Why is pushing to stash not allowed without an initial commit?

I recently ran `git stash push` in a newly initialized repository.

``` $ git init Initialized empty Git repository in <redacted>/.git/ $ touch a b $ git add a $ git stash push a You do not have the initial commit yet ```

On the other hand, with an initial commit:

```
$ git init
Initialized empty Git repository in <redacted>/.git/
$ touch a b
$ git add a
$ git commit -m a
[master (root-commit) 3bf80c1] a
 1 file changed, 0 insertions(+), 0 deletions(-)
 create mode 100644 a
$ git add b
$ git stash
Saved working directory and index state WIP on master: 3bf80c1 a
$ git log --oneline --graph --all
*   293076b (refs/stash) WIP on master: 3bf80c1 a
|\  
| * 64140e0 index on master: 3bf80c1 a
|/  
* 3bf80c1 (HEAD -> master) a
```

I see that the stash commit has two parents: the initial commit and the commit that actually holds the files that were stashed.

If git were to allow a stash entry with no initial commit, it would have to create a stash commit with only one parent (i.e. the commit labeled with 'index on <branch>'). I wonder if there is a reason this would be bad and therefore not allowed?

Thanks.

On Thu, Oct 17, 2024 at 08:51:29PM +0530, Abhijeet Sonar wrote:

Right, each stash entry has two parents:

  - The first parent is the state of HEAD in your repository when you
    created the stash entry,

  - The second parent is the state of the index at the same point

So there would be no reasonable value to substitute in for the first parent in the case where your repository does not yet have any commits.

As a concept, stashing without any commits in your repository doesn't seem fundamentally flawed, but it would make interpreting other stash entries substantially more awkward. The current model ensures that if you look at any stash entry S, then S^1 is the state of HEAD when S was created, and S^2 is likewise the state of the index.

That changes if you don't have a commit to store in S^1, since now the interpretation of S^1 depends on how many parents S has to begin with.

Thanks, Taylor

Taylor Blau <me@ttaylorr.com> writes:

I still wonder why Git has no concept of the ultimate null-commit from which all the Git world descends? Looks like it'd help to have one to handle such corner cases gracefully. Was it just an oversight of original design that isn't worth fixing already?