ops/demo.py: subprocess.run(text=True) without encoding fails on cp1252 Windows; real error swallowed into a misleading TypeError

[RRDANAI]

Summary

ops/demo.py --snapshots fails on a Windows box whose locale encoding is cp1252.
subprocess.run(..., text=True) is called without encoding, so the child's
UTF-8 output is decoded as cp1252. The resulting UnicodeDecodeError is raised
on subprocess's reader thread, where it cannot propagate — stdout comes
back as None, and the run dies much later with a TypeError that points at an
unrelated line.

Reproduction

py ops/demo.py --snapshots out

on Windows with locale.getpreferredencoding() == cp1252. Exit code 1.

Traceback

The real error, on the reader thread:

  File ".../Lib/subprocess.py", line 1615, in _readerthread
    buffer.append(fh.read())
  File ".../Lib/encodings/cp1252.py", line 23, in decode
    return codecs.charmap_decode(input,self.errors,decoding_table)[0]
UnicodeDecodeError: 'charmap' codec can't decode byte 0x90 in position 2672:
character maps to <undefined>

The error that actually stops the process:

  File ".../ops/demo.py", line 1619, in _render_isolated
    render_scenario(env, fixture, tmpdir, session_id, cfg, out_dir, theme=theme)
  File ".../ops/demo.py", line 1589, in render_scenario
    dest.write_text('\n\n'+out+'\n\n')
TypeError: can only concatenate str (not "NoneType") to str

Nothing in the second traceback points at an encoding problem, which is what
makes this expensive to diagnose.

Cause

ops/demo.py:743-748:

result = subprocess.run(
    [sys.executable, str(STATUSLINE_SCRIPT)],
    input=payload,
    text=True,
    env=env,
    capture_output=True,

text=True with no encoding uses the locale default. Byte 0x90 is an
undefined slot in cp1252, so the decode raises rather than producing mojibake.

ops/demo.py:1589 has the same latent defect on the write side:

dest.write_text('\n\n'+out+'\n\n')

Path.write_text with no encoding also uses the locale default, so it would
fail to encode the statusline's glyphs on cp1252 even once out is a valid
string.

Suggested fix

Pass encoding='utf-8' at both sites:

result = subprocess.run(..., text=True, encoding='utf-8', ...)
dest.write_text('\n\n'+out+'\n\n', encoding='utf-8')

The statusline already forces UTF-8 on its own stdout (app.py:76-77 calls
sys.stdout.reconfigure(encoding='utf-8')), so the producer side is
unambiguous — only the consumer is guessing.

Workaround

PYTHONUTF8=1 fixes it. With the flag set the same command exits 0 and writes
all 43 snapshot files.

Environment

Key	Value
OS	Windows 11 Home 10.0.26200
Python	3.13.14
Locale	cp1252
Terminal	WezTerm
Commit	74f2add