Out-of-order stdout behavior from python buffering

test.py

import subprocess

print("=== Start of script")
subprocess.run(['ls', '-ld', '/'], check=True)
print("=== End of script")

# $ python3 test.py 
# === Start of script
# drwxr-xr-x 24 root root 4096 Jul  2  2025 /
# === End of script
#
#
# $ python3 test.py | cat
# drwxr-xr-x 24 root root 4096 Jul  2  2025 /
# === Start of script
# === End of script
#
#
# $ PYTHONUNBUFFERED=1 python3 test.py | cat
# === Start of script
# drwxr-xr-x 24 root root 4096 Jul  2  2025 /
# === End of script

Even happens with print(..., file=sys.stdout) and subprocess.run(..., stdout=sys.stdout).

It turns out that sys.stdout is a wrapper that includes a buffer under some circumstances, but subprocess ganks the file descriptor out of the wrapper and passes that directly to the child process. So subprocess is unbuffered, but print is buffered (sometimes).

More detail: Python buffering can be disabled, line-based, or full. If stdout is a TTY we get line-based, but otherwise it's full buffering.