pull: Pulling an imported file in the same repo produce a "neither origin_head nor branch are provided" error

[courentin]

Bug Report

Description

When pulling a dvc file that is dvc imported from the same git repo (using . for the url) in detached HEAD (eg. in CI), the cloning step raise the following error:

2023-03-28 14:09:35,165 ERROR: failed to pull data from the cloud - SCM error: Failed to clone repo '../' to '/tmp/tmpl_adca1idvc-clone': neither origin_head nor branch are provided
Traceback (most recent call last):
  File "/__w/speech-lab/speech-lab/modelling/.venv/lib/python3.9/site-packages/scmrepo/git/backend/dulwich/__init__.py", line 219, in clone
    repo = clone_from()
  File "/__w/speech-lab/speech-lab/modelling/.venv/lib/python3.9/site-packages/dulwich/porcelain.py", line 508, in clone
    return client.clone(
  File "/__w/speech-lab/speech-lab/modelling/.venv/lib/python3.9/site-packages/dulwich/client.py", line 716, in clone
    head_ref = _set_default_branch(
  File "/__w/speech-lab/speech-lab/modelling/.venv/lib/python3.9/site-packages/dulwich/refs.py", line 1221, in _set_default_branch
    raise ValueError('neither origin_head nor branch are provided')
ValueError: neither origin_head nor branch are provided

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/__w/speech-lab/speech-lab/modelling/.venv/lib/python3.9/site-packages/dvc/scm.py", line 160, in clone
    git = Git.clone(url, to_path, progress=pbar.update_git, **kwargs)
  File "/__w/speech-lab/speech-lab/modelling/.venv/lib/python3.9/site-packages/scmrepo/git/__init__.py", line 142, in clone
    backend.clone(url, to_path, **kwargs)
  File "/__w/speech-lab/speech-lab/modelling/.venv/lib/python3.9/site-packages/scmrepo/git/backend/dulwich/__init__.py", line 224, in clone
    raise CloneError(url, to_path) from exc
scmrepo.exceptions.CloneError: Failed to clone repo '../' to '/tmp/tmpl_adca1idvc-clone'

The issue appears even if I use cml ci --fetch-depth=0 or:

uses: actions/checkout@v3
with:
  fetch-depth: 0

Reproduce

dvc import a file from the same repo and dvc pull this file in datached mode.
Also, We have two dvc instances in the same git repo in two subdirectories.

Expected

I expect no exception or a clearer one.

Environment information

The problem only appear in a CI environment where the repo is detached from HEAD.

[UPDATE]: I was able to reproduce it locally by being in detached mode.

Output of dvc doctor:

DVC version: 2.51.0 (pip)
-------------------------
Platform: Python 3.9.5 on Linux-4.14.304-226.531.amzn2.x86_64-x86_64-with-glibc2.31
Subprojects:
        dvc_data = 0.44.1
        dvc_objects = 0.21.1
        dvc_render = 0.3.1
        dvc_task = 0.2.0
        scmrepo = 0.1.17
Supports:
        http (aiohttp = 3.8.1, aiohttp-retry = 2.8.3),
        https (aiohttp = 3.8.1, aiohttp-retry = 2.8.3),
        s3 (s3fs = 2023.1.0, boto3 = 1.24.59)
Cache types: symlink
Cache directory: lustre on 172.31.75.85@tcp:/o4e3rbev
Caches: local
Remotes: s3, s3
Workspace directory: xfs on /dev/nvme0n1p1
Repo: dvc (subdir), git
Repo.site_cache_dir: /var/tmp/dvc/repo/639387cceca8b895ded2b9292d87a036

[courentin]

Alternatively, we tried to specify the git url instead of the local url when importing the file but we ran into another issue where the Clone step during the pull is stuck and the memory keeps increasing until the systems runs out of memory.

I was not able to reproduce it locally (even in detached mode) though

[shcheklein]

@pmrowla hey, I see that it was fixed in Dulwich, are we waiting for the new release?

[pmrowla]

It's fixed in dulwich but has not been released yet. Anyone who needs the fix can install from dulwich master via pip in the meantime

pip install git+https://github.com/jelmer/dulwich.git

[pmrowla]

This is released in dulwich==0.21.5, for binary packages it will be fixed in the next DVC release