Description

benmss

opened on Aug 6, 2025

Issue body actions

Tags found in the packed-refs file that can be created by the Git command git pack-refs do not have to be UTF8 encoded. In these cases, GitPython fails to read these tags due to assuming they should be UTF8 compatible.

The source for this issue is found here:
https://github.com/gitpython-developers/GitPython/blob/main/git/refs/symbolic.py#L124

with open(cls._get_packed_refs_path(repo), "rt", encoding="UTF-8") as fp:

A working example (using a repository with a non-utf8 tag):

git clone https://github.com/ACRA/acra
cd acra

Create a Python script with the following content:

import git
repo = git.Repo(".")
print(repo.tags)

Execute the script:

python script.py

Result:
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc3 in position 6216: invalid continuation byte

Activity

1. Byron

   added

   acknowledged

   help wanted

   on Aug 6, 2025
2. Byron commented on Aug 6, 2025

   

   Byron

   on Aug 6, 2025

   Member

   More actions

   Thanks for reporting!

   GitPython absolutely has a problem with encodings, in that it should use bytes in most places where it now uses strings, and the same is true for filesystem paths.

   Reacted by Ben Selwyn-Smith
3. Byron

   added the

   Bug

   issue type on Jul 20, 2026
4. Byron

   added this to the v4.0 milestone on Jul 26, 2026
5. Byron commented on Jul 26, 2026

   

   Byron

   on Jul 26, 2026

   Member

   More actions

   When reviewing #2190 (comment) I realize that this can't be fixed properly without changing the types that are created by the ref traversal, after all, these can't have an encoding attached to them anymore.

   So no need to attempt this in V3.
6. Byron

   added a commit that references this issue on Oct 25, 2026

   fix(refs): preserve non-UTF-8 names at native API boundaries (#2064)

   70d2fd5
7. Byron

   closed this as completedin #2308on Oct 25, 2026

Issue actions