Search code examples
pythonpython-3.xsubprocessvalueerror

Python write to Linux device using subprocess gives ValueError : embedded null byte

On a Debian Linux system I have a Python script that prepares a string to be written out to a USB CDC device at /dev/ttyACM0. This write occurs at the end of the script as follows...

USBpipe = open("/dev/ttyACM0", 'w')
shellCmd = subprocess.Popen(["echo", USBpacket], stdout = USBpipe)
USBpipe.close()

...where USBpacket is a string. If I make this a pure string such as using USBpacket = "test" the code executes correctly and I've verified the data appears on the USB device. However, during normal execution USBpacket gets bytes appended to it via the chr() function, and some of these may be zero. When this happens I get this error running the script:

Traceback (most recent call last):
  File "/root/gpstoPIC.py", line 247, in <module>
    shellCmd = subprocess.Popen(["echo", USBpacket], stdout = USBpipe)
  File "/usr/lib/python3.9/subprocess.py", line 951, in __init__
    self._execute_child(args, executable, preexec_fn, close_fds,
  File "/usr/lib/python3.9/subprocess.py", line 1756, in _execute_child
    self.pid = _posixsubprocess.fork_exec(
ValueError: embedded null byte

I've tried a couple of solutions that haven't worked, such as doing wb instead of w on the open of /dev/ttyACM0, and using bytes() to convert from string to binary. What is the correct way to write out this data?

Solution

  • The shell in general, and echo in particular, can't cope with arbitrary binary input. Lucky thing then that Python can, and so you don't need a subprocess here at all.

    with open("/dev/ttyACM0", "wb") as usb_pipe:
    usb_pipe.write(USBpacket)

    If you really wanted to use a subprocess, try cat (but then that's really a useless use of cat). Notice also how the open uses "wb" to open the filehandle in binary mode (plain "w" applies an encoding which again is not tolerant of or compatible with arbitrary binary data). And yes, if the data you want to send is really binary, USBpacket should indeed be a bytes object.

    Finally, perhaps notice how I preferred snake_case over CamelCase for local scalar variable names; perhaps USBpacket should be renamed to adhere to PEP-8 recommendations, too.