test_console_encoding.py

"""The CLI must survive a legacy-codepage console (issue #140).

Windows consoles default to cp1252, which cannot encode the arrows and
middots in our output. Before the fix, `waku dashboard` crashed with
UnicodeEncodeError while printing the startup banner — before the server
even started. The entrypoint now reconfigures stdout/stderr with
errors="replace", so unencodable characters degrade to "?" instead of
killing the process. This test forces the failing encoding on any OS.
"""

import os
import subprocess
import sys


def test_usage_text_survives_a_cp1252_console():
    """`waku <unknown>` prints the usage text, which contains a real arrow.
    Under a cp1252 stdout that print must degrade, not raise."""
    env = {**os.environ, "PYTHONIOENCODING": "cp1252"}
    # Compare bytes: the child's output IS cp1252, and decoding it with the
    # parent's own stdio encoding would make this test depend on the host.
    result = subprocess.run(
        [sys.executable, "-m", "prometheus", "definitely-not-a-command"],
        capture_output=True,
        env=env,
        timeout=60,
        check=False,
    )
    assert b"UnicodeEncodeError" not in result.stderr
    assert result.returncode == 1  # the usage path's normal exit
    assert b"prometheus dashboard" in result.stdout  # usage text actually printed