run_local not working on windows

My system is Python 3.7.0 running Windows 10 and I receive an error when I try to run anything with run_local on windows (which affects all backends). The stack trace looks like this:

venv\lib\site-packages\testinfra\host.py:71: in run
    return self.backend.run(command, *args, **kwargs)
venv\lib\site-packages\testinfra\backend\docker.py:35: in run
    "docker exec %s /bin/sh -c %s", self.name, cmd)
venv\lib\site-packages\testinfra\backend\base.py:203: in run_local
    stderr=subprocess.PIPE,
..\..\..\appdata\local\programs\python\python37-32\Lib\subprocess.py:756: in __init__
    restore_signals, start_new_session)
..\..\..\appdata\local\programs\python\python37-32\Lib\subprocess.py:1100: in _execute_child
    args = list2cmdline(args)
..\..\..\appdata\local\programs\python\python37-32\Lib\subprocess.py:511: TypeError
    TypeError: argument of type 'int' is not iterable

Looks like the fact that the command is being encoded before running is to blame. Popen in Windows doesn’t handle byte arrays the same way it does on Linux. Was able to fix by modifying code to this:

def run_local(self, command, *args):
    command = self.quote(command, *args)
    if os.name != 'nt':
        command = self.encode(command)
    p = subprocess.Popen(
        command, shell=True,
        stdin=subprocess.PIPE,
        stdout=subprocess.PIPE,
        stderr=subprocess.PIPE,
    )
    stdout, stderr = p.communicate()
    result = self.result(p.returncode, command, stdout, stderr)
    return result

Basically only encode when not on windows

[viniciusartur]

How did you patch the run_local(self, command, *args) method and how can I do this with the approach above? For me it looks like as if my patched run_local method is not used.

In my conftest.py:

def patch_testinfra():
    def run_local_new(self, command, *args):
        command = self.quote(command, *args)
        p = subprocess.Popen(
            command, shell=True,
            stdin=subprocess.PIPE,
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
        )
        stdout, stderr = p.communicate()
        result = self.result(p.returncode, command, stdout, stderr)
        return result

    from testinfra.backend.base import BaseBackend
    BaseBackend.run_local = run_local_new

def pytest_generate_tests(metafunc):
    patch_testinfra()

[thomasleveil]

@viniciusartur solution wasn’t suffisent in my case (I’m using testinfra with docker backend), but combining its solution with https://github.com/pytest-dev/pytest-testinfra/issues/375#issue-360991022 made it work :

# conftest.py
import os
import re
import subprocess
import testinfra

def patch_testinfra():
    if os.name == 'nt':
        """
        Making testinfa work from Windows with Docker backend.
        Inspired from 
        - https://github.com/pytest-dev/pytest-testinfra/issues/411#issuecomment-782729362
        - https://github.com/pytest-dev/pytest-testinfra/issues/375#issue-360991022 
        """
        def quote(command, *args):
            def anon_1(arg):
                if re.match(r'/("|\s|\')', arg) != None:
                    return arg

                arg = re.sub('"', '\"', arg)
                return '"%s"' % (arg)

            if args:
                return command % tuple(anon_1(a) for a in args)
                # return command % tuple(pipes.quote(a) for a in args)
            return command

        def run_local_new(self, command, *args):
            command = quote(command, *args)
            p = subprocess.Popen(
                command,
                shell=True,
                stdin=subprocess.PIPE,
                stdout=subprocess.PIPE,
                stderr=subprocess.PIPE,
            )
            stdout, stderr = p.communicate()
            result = self.result(p.returncode, command, stdout, stderr)
            return result

        from testinfra.backend.base import BaseBackend

        BaseBackend.run_local = run_local_new


def pytest_generate_tests(metafunc):
    patch_testinfra()