Handle non- utf-8 characters

[adamtheturtle]

subprocess-tee errors when the subprocess prints a non- utf-8 character.
This is a difference between subprocess and subprocess-tee.

Reproduction

# my_script.sh
echo -e "\xC0\x80"

# reproducer.py
import subprocess
import subprocess_tee

print("Subprocess:")

subprocess.run(args=["bash", "my_script.sh"])

print("Subprocess tee:")

subprocess_tee.run(args=["bash", "my_script.sh"])

subprocess will replace invalid bytes with a placeholder character, while subprocess-tee errors with:

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc0 in position 0: invalid start byte

Fix

I suggest using errors="replace" on line.decode().

[ssbarnea]

Hmm... not sure but okey. I will accept a patch.

[adamtheturtle]

Thanks @ssbarnea .

The test might look something like:

def test_run_compat_not_utf8(tmp_path: Path) -> None:
    """Assure compatiblity with subprocess.run() when command output is not UTF-8."""
    script = tmp_path / "script.sh"
    script_content = 'echo -e "\xC0\x80"'
    script.write_text(data=script_content, encoding="latin1")
    cmd = ["bash", str(script)]
    ours = run(cmd)
    original = subprocess.run(
        cmd,
        stdout=subprocess.PIPE,
        stderr=subprocess.PIPE,
        check=False,
    )
    assert ours.returncode == original.returncode
    assert ours.stdout == original.stdout.decode("latin1")
    assert ours.stderr == original.stderr.decode("latin1")
    assert ours.args == original.args

Note that universal_newlines is not set in subprocess.run, and this contradicts the README:

Keep in mind that universal_newlines=True is implied as we expect text
processing, this being a divergence from the original subprocess.run.