Cloning with remote callback results in HEAD pointing to user's `init.defaultBranch`

提供されている pygit2.clone_repository 呼び出しとその remote_cb コールバックから始め、コールバックありの場合となしの場合の動作を再現します。コールバックが実行される前に /tmp/pygit2/.git/HEAD を調べ、クローンされたリポジトリのデフォルトブランチと比較します。完了条件は、コールバック付きでクローンすると HEAD が init.defaultBranch ではなくクローンされたリポジトリのデフォルトブランチを指し、クローン動作がテストされているすべての箇所にカバレッジが追加されていることです。

Cloning with a remote callback results in HEAD pointing to the user's init.defaultBranch instead of the default branch of the cloned repository. See repro below.

Looks like an upstream bug to me, just confirming with you... thanks :)

I'll also note that HEAD already contains the wrong reference before the callback is executed. (Reproduce this e.g. by placing a breakpoint() in remote_cb below, and inspecting HEAD from the debugger prompt.) Still, the issue does not reproduce without the callback.

Repro

❯ docker run --rm -ti python bash
root@e747d803b7f3:/# pip install pygit2
Collecting pygit2
  Downloading pygit2-1.5.0-cp39-cp39-manylinux2014_x86_64.whl (3.1 MB)
     |████████████████████████████████| 3.1 MB 2.4 MB/s 
Collecting cffi>=1.4.0
  Downloading cffi-1.14.5-cp39-cp39-manylinux1_x86_64.whl (406 kB)
     |████████████████████████████████| 406 kB 15.7 MB/s 
Collecting cached-property
  Downloading cached_property-1.5.2-py2.py3-none-any.whl (7.6 kB)
Collecting pycparser
  Downloading pycparser-2.20-py2.py3-none-any.whl (112 kB)
     |████████████████████████████████| 112 kB 17.0 MB/s 
Installing collected packages: pycparser, cffi, cached-property, pygit2
Successfully installed cached-property-1.5.2 cffi-1.14.5 pycparser-2.20 pygit2-1.5.0
root@e747d803b7f3:/# cat >> ~/.gitconfig
[init]
    defaultBranch = teapot
root@e747d803b7f3:/# python
Python 3.9.4 (default, Apr 10 2021, 15:31:19) 
[GCC 8.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pygit2
>>> def remote_cb(repo, name, url):
...     n = name.decode()
...     repo.config[f"remote.{n}.mirror"] = True
...     return repo.remotes.create(name, url, "+refs/*:refs/*")
... 
>>> repo = pygit2.clone_repository("https://github.com/libgit2/pygit2", "/tmp/pygit2", remote=remote_cb)
>>> repo.head
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
_pygit2.GitError: reference 'refs/heads/teapot' not found
>>> print(open("/tmp/pygit2/.git/HEAD").read())
ref: refs/heads/teapot

Workaround

import contextlib
import pygit2

def _fix_repository_head(repository: pygit2.Repository) -> pygit2.Reference:
    """Work around a bug in libgit2 resulting in a bogus HEAD reference.

    Cloning with a remote callback results in HEAD pointing to the user's
    `init.defaultBranch` instead of the default branch of the cloned repository.
    """
    head = repository.references["HEAD"]

    with contextlib.suppress(KeyError):
        return head.resolve()

    for branch in ["main", "master"]:
        ref = f"refs/heads/{branch}"
        if head.target != ref and ref in repository.references:
            head.set_target(ref, message="repair broken HEAD after clone")

    return head.resolve()