Unicode file name do not checkout correctly on windows

[garyvdm]

Steps to reproduce:

dulwich clone https://github.com/garyvdm/git_unicode_files.git
dir git_unicode_files

expected: 1 file named À (which is u'\u00c0')
actual: the file is named À (which is u'\u00c3\u20ac')

the file name is what you get if you do u'\u00c0'.encode('utf8').decode('mbcs'). mbcs it the default filesystem charter encoding used on windows.

The git client handles this correctly. I'll take a look at their source code in the future to try figure out how they handle this.

[garyvdm]

This is what msysgit does: msysgit/git@19d1e75

[jelmer]

Has this been resolved?

[jelmer]

@garyvdm Has this been resolved?

[garyvdm]

No, not yet.

[guyskk]

I also encountered this problem.
see FriendCode/gittle#72

UnicodeDecodeError When filename is"article/python2编码问题.md" or has unicode char

dulwich/index.py(423) build_index_from_tree()
-> full_path = os.path.join(prefix, entry.path)
(Pdb) pp prefix
u'E:/work/py/kkblog/article_repo/\u54c8\u54c8\\guyskk\\webhooks_test'
(Pdb) pp entry.path
'article/python2\xe7\xbc\x96\xe7\xa0\x81\xe9\x97\xae\xe9\xa2\x98.md'
(Pdb) os.path.join(prefix, entry.path)
*** UnicodeDecodeError: 'ascii' codec can't decode byte 0xe7 in position 16: ordinal not in range(128)

My script:

# coding:utf-8

def pull_or_clone(dest, repo_url):

    from giturlparse import parse
    from gittle import Gittle
    import os
    p = parse(repo_url)
    user_repo_path = os.path.join(dest, p.owner, p.repo)
    if os.path.exists(user_repo_path):
        repo = Gittle(user_repo_path, origin_uri=repo_url)
        repo.pull()
    else:
        repo = Gittle.clone(repo_url, user_repo_path)

if __name__ == '__main__':
    dest = u"E:/work/py/kkblog/article_repo/哈哈"
    repo_url = u"https://github.com/guyskk/webhooks_test.git"
    pull_or_clone(dest, repo_url)

[jelmer]

It would be great if somebody could verify this still happens with Dulwich 0.20.3. The testsuite now passes on Windows, so if it still happens we can probably add a test & fix for it.