[cloudsmith-iduffy]

Description

On Windows, the CLI reads config.ini and credentials.ini in the locale encoding, which is usually cp1252. click-configfile passes no encoding to configparser, and the code that changes the files (save_workspace after cloudsmith auth, and _set_api_key for login, logout and tokens ... --save-config) does the same. A UTF-8 file with Cyrillic text fails because cp1252 has no character for the bytes 0x81, 0x8D, 0x8F, 0x90 and 0x9D (for example, "с" is D1 81). So every command fails, not only auth and logout.

The CLI now picks the encoding for each file:

• If the bytes decode as UTF-8, the CLI uses UTF-8. A file that starts with a UTF-8 byte order mark keeps the mark.
• If not, the CLI uses the locale encoding, as it does today. A cp1252 file with José still loads, and a rewrite keeps the file byte for byte.
• A new file is UTF-8.

ConfigReader.read_config replaces the click-configfile version so that it can give the encoding to configparser for each file. The other logic is the same as the click-configfile code.

When the CLI cannot decode a file in either encoding, the error now gives the path. Sentry CLOUDSMITH-CLI-K is an example of this.

Before (Windows, UTF-8 config.ini with a Cyrillic comment):

> cloudsmith whoami
Error: UnicodeDecodeError: 'charmap' codec can't decode byte 0x81 in position 22: character maps to <undefined>

After: the command runs.

Before (a file that is not UTF-8, on macOS or Linux):

$ cloudsmith whoami --config-file config.ini
Error: UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 21: invalid start byte

After:

$ cloudsmith whoami --config-file config.ini
Error: Cannot read /home/me/config.ini as utf-8 text. Save the file as UTF-8.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Refactoring
• Other (please describe)

Additional Notes

• This is not a breaking change. A file that loads today still loads with the same values, except for one case. That case is a non-UTF-8 file whose bytes also happen to form valid UTF-8, such as cp1252 é. The exit code for a file that the CLI cannot decode stays 1. The error is now a clean message, not a raw UnicodeDecodeError, and the CLI still does not report it to Sentry.
• On Windows without UTF-8 mode, a UTF-16 file decodes as cp1252. It still fails with MissingSectionHeaderError, as it does today. That error also gives the path.
• The tests use cp1252 as the default encoding, in the same way as fix(no-ticket): read and write .npmrc as UTF-8 #555. They have the windows marker. This PR adds the marker to pyproject.toml with the same line as fix(no-ticket): print ASCII text in mcp configure, download and update #525 and fix(no-ticket): read and write .npmrc as UTF-8 #555.
• git merge-tree conflicts with other open PRs:
  • fix(no-ticket): keep credentials.ini sections and save missing API keys #533: cloudsmith_cli/cli/config.py, 2 hunks in _set_api_key. That PR adds lines next to the two open(...) lines. To resolve, keep its new lines and keep the encoding=_get_file_encoding(path) argument from this PR. There is also a conflict in CHANGELOG.md under ### Fixed.
  • fix(no-ticket): save the new API key to the user-level credentials file #541: CHANGELOG.md only. config.py merges cleanly. The save_workspace signature is the same as in fix(no-ticket): save the new API key to the user-level credentials file #541.
  • fix(no-ticket): save credentials.ini with mode 0600 #553: cloudsmith_cli/cli/config.py, 1 hunk where the chmod block meets the write open(...). Resolve it in the same way as fix(no-ticket): keep credentials.ini sections and save missing API keys #533. There is also a conflict in CHANGELOG.md.

🤖 Generated with Claude Code

[Copilot]

🟡 Changes recommended

Both encoding-test assertions reject native Windows line endings, causing all six parameterized cases to fail on the target platform.

1 open issue

• Normalize CRLF before comparing expected encoded bytes · New

What changed in this PR

This PR fixes Windows encoding failures when the CLI reads or updates config.ini and credentials.ini, while retaining support for locale-encoded files.

Changes:

• Detect each file’s encoding, preserving UTF-8 byte order marks and using UTF-8 for new files.
• Report the affected path when a configuration file cannot be decoded.
• Add encoding regression tests and register the Windows test marker.

File	Description
pyproject.toml	Registers the windows pytest marker.
cloudsmith_cli/​cli/​tests/​test_unexpected_errors.py	Checks actionable decoding errors without Sentry reporting.
cloudsmith_cli/​cli/​tests/​test_config_encoding.py	Tests UTF-8, BOM, and cp1252 read/rewrite behavior.
cloudsmith_cli/​cli/​config.py	Adds per-file encoding detection for reads and updates.
CHANGELOG.md	Documents encoding support and improved errors.

_{Review effort: Balanced}

[cloudsmith-iduffy]

@copilot resolve the merge conflicts in this pull request

[Copilot]

@copilot resolve the merge conflicts in this pull request

Resolved in 5709ff8 by merging master and retaining both sets of changelog entries. All 46 targeted tests and pre-commit checks passed; CodeQL found no alerts.