[Alpenj]

Description

On Windows, the locale encoding is a legacy code page (for example cp1252, or cp949 on Korean Windows). Two code paths used it for text that is often outside that code page:

• cz commit --edit wrote the message to a temporary file with the locale encoding and read it back the same way. A message with characters outside the code page failed before the editor opened:
  UnicodeEncodeError: 'cp949' codec can't encode character '\u2728' in position 25: illegal multibyte sequence
  It now uses the new commit_encoding setting. When it is not set, it uses encoding ("utf-8" by default). A file that the editor saves as UTF-8 is also read back correctly.
• cz init read pyproject.toml with the locale encoding to choose the default version provider. TOML files are UTF-8, so a pyproject.toml with non-ASCII text (for example an author name) failed:
  UnicodeDecodeError: 'cp949' codec can't decode byte 0xe2 in position 50: illegal multibyte sequence
  It now reads the file as UTF-8.

commit_encoding also applies to the other files that hold a commit message: the --write-message-to-file file, the --retry backup file and the cz check --commit-msg-file file. They already used encoding, so they do not change by default.

Checklist

• I have read the contributing guidelines

Was generative AI tooling used to co-author this PR?

• Yes (please specify the tool below)

Generated-by: Claude Code (claude-opus-5-5) following the guidelines

Assisted By: AI

Code Changes

• Add test cases to all the changes you introduce
• Run uv run poe all locally to ensure this change passes linter check and tests
• Manually test the changes:
  • Verify the feature/bug fix works as expected in real-world scenarios
  • Test edge cases and error conditions
  • Ensure backward compatibility is maintained
  • Document any manual testing steps performed
• Update the documentation for the changes

Expected Behavior

cz commit --edit and cz init work with non-ASCII text whatever the locale encoding is.

Steps to Test This Pull Request

1. On Windows with a non-UTF-8 code page, run pytest tests/commands/test_commit_command.py tests/commands/test_check_command.py tests/test_project_info.py.
2. Without this change, test_manual_edit_keeps_non_ascii_message and the new test_get_default_version_provider case fail with the errors above. With this change, they pass.

Additional Context

What I ran on Windows 11 (cp949), Python 3.12, instead of uv run poe all:

• With the source files from master, the new tests fail. With this PR, tests/commands/test_commit_command.py, tests/commands/test_check_command.py, tests/test_conf.py, tests/test_project_info.py and tests/commands/test_init_command.py pass (204 passed).
• cz check --rev-range origin/master.. and ruff check pass. mypy reports no errors in the changed files; it reports the same 17 errors on master with the dependency versions installed here.

I did not run the full test suite with coverage. On Linux and macOS the locale encoding is usually UTF-8, so the new tests pass there with or without the fix; the Windows CI job runs them with the runner's cp1252 code page.

This PR was prepared and opened by Claude Code, an AI agent, on behalf of the account owner.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.29%. Comparing base (f690817) to head (c00f5dc).
⚠️ Report is 3 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2119   +/-   ##
=======================================
  Coverage   98.28%   98.29%           
=======================================
  Files          61       61           
  Lines        2864     2868    +4     
=======================================
+ Hits         2815     2819    +4     
  Misses         49       49           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

[woile]

I'm open to the change, a few details

[woile]

This should be commit_encoding and it should be initialized in

commitizen/commitizen/defaults.py

Line 32 in 33c248e

class Settings(TypedDict, total=False):

It should also have a default in
https://github.com/commitizen-tools/commitizen/blob/master/commitizen/defaults.py#L82-L88

Should the commit_encoding default be utf-8 or should it be None?

[Alpenj]

Done in c00f5dc. The default is None, which means "use encoding". So existing configs keep the same behavior, and the --edit file is still UTF-8 by default.

I also applied commit_encoding to the other files that hold a commit message: --write-message-to-file, the --retry backup and cz check --commit-msg-file. They already used encoding, so nothing changes by default, and one setting now controls all commit message files.

I documented it in docs/config/commit.md.