Tests hanging forever

[AndreaCrotti]

Thanks for this great utility, I'm trying to get it working now with a very simple example:

def test_connect_to_sftp(sftpserver):
    with sftpserver.serve_content({'a_dir': {'somefile.txt': "Some content"}}):
        trans, client = sftp_utils.connect_to_sftp(
            host=sftpserver.host,
            port=sftpserver.port,
            username='user',
            password='pw',
            dsp_name='name',
        )

Where the connect is very simple and just does this

def connect_to_sftp(host, port, username, password, dsp_name):
    logger.info('Connecting to %s: %s:%s', dsp_name, host, port)
    transport = Transport((host, port))
    transport.connect(username=username, password=password)
    logger.info('Connected to %s', dsp_name)

    return transport, SFTPClient.from_transport(transport)

It works in theory but the test runner never quits it just hangs there forever. Any idea why it's happening? Is there something I can do to kill the sftpserver thread? (I tried shutdown already but nothing..)

[mbyio]

It looks like you're using the paramiko client? If so, then this is actually a paramiko issue. It opens Python threads in the background to do the actual communication with the server. So you need to actually close the client or it might get stuck.

[ulope]

If this is indeed a paramiko issue I assume this issue can be closed?

Note to self: Maybe adding a hint in the readme wouldn't hurt though...

[alfredopalhares]

I am having ths problem, if its a paramiko issue, is there any links ? What is the status of this.

[christopherdoyle]

Also encountered this problem, I think sure this is the relevant paramiko issue: https://github.com/paramiko/paramiko/issues/520 (the issue is more widespread than the title). As above, explicitly closing the transport or client fixes it, e.g. from the orignal post adding trans.close() to the end of the test. I think the issue can be closed? I can't see any way of dealing with this in the fixture without monkey patching paramiko.

[matteosantama]

Similar problem here. Has anyone figured out how to gracefully kill the sftpserver at the end of a test?

[alfredopalhares]

Nope, still nothing on my end. Shame, because this library is not that useful withthis bug.

[fiendish]

Yeah, this happened to me too.

As above, explicitly closing the transport or client fixes it, e.g. from the orignal post adding trans.close() to the end of the test.

Not for me. :\

[fiendish]

@ulope running the server as a thread is probably a mistake. I was able to work around this issue by spawning the server in its own process to eliminate lock contention:

import time
from multiprocessing import Process, Queue
from pytest_sftpserver.sftp.server import SFTPServer


def run_server(q, contents):
    sftpserver = SFTPServer(content_object=contents)
    sftpserver.start()
    q.put((sftpserver.host, sftpserver.port))


def test_some_stuff():
    contents = {}  # my contents structure here

    queue = Queue()
    p = Process(target=run_server, args=(queue, contents), daemon=True)
    p.start()

    for _ in range(10):
        address = queue.get()
        if address:
            break
        time.sleep(0.5)

    assert address

    # <do my testing here>

[johannes-cogitaris]

To anyone coming here: This might be an alternative: https://github.com/oz123/pytest-localftpserver

edit: ftp only :-(

[tmieulet]

see : https://github.com/ulope/pytest-sftpserver/issues/30#issuecomment-1530896213 It worked for me:

@pytest.fixture
def sftp_fixture(sftpserver):
    # https://github.com/ulope/pytest-sftpserver/issues/30
    # Tests hanging forever
    sftpserver.daemon_threads = True
    sftpserver.block_on_close = False
    yield sftpserver