You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Traceback (most recent call last):
File "gitl.py", line 13, in <module>
main()
File "gitl.py", line 9, in main
"/home/vlad/AltSTU-Submodules/Диплом/Microcontroller-Unit-Testing/scripts/")
File "/home/vlad/.local/lib/python3.5/site-packages/git/repo/base.py", line 982, in clone_from
return cls._clone(git, url, to_path, GitCmdObjectDB, progress, **kwargs)
File "/home/vlad/.local/lib/python3.5/site-packages/git/repo/base.py", line 933, in _clone
finalize_process(proc, stderr=stderr)
File "/home/vlad/.local/lib/python3.5/site-packages/git/util.py", line 333, in finalize_process
proc.wait(**kwargs)
File "/home/vlad/.local/lib/python3.5/site-packages/git/cmd.py", line 402, in wait
stderr = force_bytes(stderr)
File "/home/vlad/.local/lib/python3.5/site-packages/gitdb/utils/encoding.py", line 16, in force_bytes
return data.encode(encoding)
UnicodeEncodeError: 'ascii' codec can't encode characters in position 54-59: ordinal not in range(128)
Back in the days when this library was originally conceived, it was very easy to get text encodings wrong, unfortunately.
Given the current state of affairs, is there any good way to deal with this in what I would call 'modern' python?
After all, the actual OS encoding can only be guessed if it is not provided from the outside, and today one might argue that utf-8 is a better default than ascii.
Well, as far as i know, python 3 uses UTF-8 by default, so adding
# -*- coding: utf-8 -*-
to my code didn't worked out. I don't know where to look for this bug: in system locale configuration or in library. Can you confirm that, this actually a bug in library?
It is a problem with GitPython as it tries to decode something that is not ascii encoded as ascii.
The coding declaration you provided tells python how to decode source code files while making sense of them, and thus not affect the runtime.
Activity
Byron commentedon Jun 5, 2018
Back in the days when this library was originally conceived, it was very easy to get text encodings wrong, unfortunately.
Given the current state of affairs, is there any good way to deal with this in what I would call 'modern' python?
After all, the actual OS encoding can only be guessed if it is not provided from the outside, and today one might argue that
utf-8is a better default thanascii.vtsisyk commentedon Jun 7, 2018
Well, as far as i know, python 3 uses UTF-8 by default, so adding
to my code didn't worked out. I don't know where to look for this bug: in system locale configuration or in library. Can you confirm that, this actually a bug in library?
Byron commentedon Jun 10, 2018
It is a problem with GitPython as it tries to decode something that is not
asciiencoded asascii.The
codingdeclaration you provided tellspythonhow to decode source code files while making sense of them, and thus not affect the runtime.brizjin commentedon Aug 13, 2018
I have the same error in GitPython 2.1.11. Is there any solutions?
Byron commentedon Oct 14, 2018
@brizjin I don't think there is unless someone contributes a patch.
Harmon758 commentedon Feb 18, 2020
This should be fixed with ebf4656 (#955) in v3.0.5. See #920.