[nk-batman21]

On Windows with Python 3.11 to 3.14 (without UTF-8 mode), isort --check-only lets an unsorted file pass if it contains a character the ANSI code page cannot encode, such as an emoji, CJK text or ✓. For a file unsorted.py containing

import sys
import os

print('✓ done')

running python -m isort --check-only unsorted.py on a cp1252 machine prints UserWarning: Unable to parse file unsorted.py due to 'charmap' codec can't encode character '\u2713' in position 32: character maps to <undefined> and exits 0. isort.check_file() raises UnicodeEncodeError on the same file. 8.0.1 reports the file as incorrectly sorted and exits 1. The same thing happens in this repo: on Windows, isort --profile hug --check isort/ tests/ skips seven of the test files with this warning.

This is a regression from 9.0.0: since fed6c0c (Remove _EmptyIO, #2601), check_stream() writes the sorted output to open(os.devnull, "w"), which uses the locale encoding. That output is thrown away, so this change opens it as UTF-8.

The new test in tests/unit/test_api.py fails without the change on Windows with a cp1252 locale (checked on 3.11 and 3.14). On Linux and macOS it passes either way because the locale encoding is already UTF-8.

Ran: pytest tests/unit/test_api.py (18 passed) and coverage run -m pytest tests/unit/ -s (713 passed, 5 skipped, coverage 99%) on Windows with Python 3.14.
Checked: mypy, flake8, ruff check, ruff format --check, bandit -r isort/ and isort --profile hug --check on the changed files.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.38%. Comparing base (a514c71) to head (2e14bed).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2740   +/-   ##
=======================================
  Coverage   99.38%   99.38%           
=======================================
  Files          40       40           
  Lines        3274     3274           
  Branches      706      706           
=======================================
  Hits         3254     3254           
  Misses         12       12           
  Partials        8        8           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

🟢 Approved

The narrow encoding fix addresses the reported failure without changing sorting behavior, and the regression test covers both check outcomes.

No open issues 🎉

What changed in this PR

Fixes a Windows check-mode regression by explicitly using UTF-8 for discarded sorting output, preventing non-ASCII files from being skipped because of locale encoding errors.

Changes:

• Opens the null output stream with UTF-8 encoding.
• Adds regression coverage for sorted and unsorted files containing a non-ASCII character.

File	Description
tests/​unit/​test_api.py	Tests both check results with non-ASCII content.
isort/​api.py	Uses UTF-8 for discarded check-mode output.

_{Review effort: Balanced}