Exception if GIT_COMMITTER_NAME contains UTF-8 encoded Umlaut

[guettli]

If I run the python script using gitpython with this env var:

export GIT_COMMITTER_NAME="Thomas Müller"

I get this exception:

Error
Traceback (most recent call last):
  File "/home/foo_eins_d/src/scmtools/scm/tests/test_git.py", line 25, in test_git_status
    repo.index.commit('initial commit')
  File "/home/foo_eins_d/local/lib/python2.7/site-packages/git/index/base.py", line 900, in commit
    head, author=author, committer=committer)
  File "/home/foo_eins_d/local/lib/python2.7/site-packages/git/objects/commit.py", line 349, in create_from_tree
    new_commit._serialize(stream)
  File "/home/foo_eins_d/local/lib/python2.7/site-packages/git/objects/commit.py", line 392, in _serialize
    altz_to_utctz_str(self.committer_tz_offset))).encode(self.encoding))
UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 18: ordinal not in range(128)

Version: GitPython==0.3.5

[Byron]

This is an interesting one ! git-python does handle non-ascii encodings, and defaults to utf-8.
In your case, there either is a non-obvious implicit conversion, or self.encoding ended up being 'ascii'.

I will review how encodings are handled to make sure something like this can't easily happen.

In the meanwhile, can you post the test in test_git.py here to make sure I can reproduce the issue ?

[Byron]

You can watch the archived development stream on youtube

[guettli]

thank you very much for the fast fix.

[saheel1115]

@Byron A similar issue occurs in GitPython-1.0.1 (tested with Python 2.7 and 3.4) when I try to do Repo.git.blame() and the blame output contains some funny characters. Is this issue fixed somewhere upstream, so that I use a later version?

[Byron]

@saheel1115 Repo.git.blame() executes the git command and attempts to interpret the returned bytes as UTF-8 string. If that is not expected to work, you can keep the bytes unchanged like so. Repo.git.blame(stdout_as_string=False).

[saheel1115]

@Byron Cool, sounds perfect. Thanks!

[jrydberg]

The stdout_as_string=False workaround for blame isn't working anymore since that is already passed to self.git.blame. defencis utf-8 but the funny character in the committer name gets the parser off track.

[Byron]

@jrydberg It could be that whatever it is trying to parse is not actually encoded in UTF-8. The current implementation doesn't handle encodings properly at all, unfortunately, and I am afraid there is no fix for that unless one changes all str usages into bytes.