Clear Open client state on close-response error

[henrik-nil-acc]

When receive() raised anything other than FileClosed, the
old close path propagated the exception with _connected still
True and the file_id still in session.open_table. A
subsequent close() reissued SMB CLOSE on a dead file_id, and
session.disconnect(close=True) walked the stale entry.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.78%. Comparing base (66ee483) to head (9fb8815).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #359      +/-   ##
==========================================
- Coverage   98.78%   98.78%   -0.01%     
==========================================
  Files          24       24              
  Lines        5289     5288       -1     
==========================================
- Hits         5225     5224       -1     
  Misses         64       64              

Flag	Coverage Δ
macOS	69.09% <52.94%> (+0.29%)	⬆️
py3.10	98.78% <100.00%> (-0.01%)	⬇️
py3.11	98.78% <100.00%> (-0.01%)	⬇️
py3.12	98.78% <100.00%> (-0.01%)	⬇️
py3.13	98.78% <100.00%> (-0.01%)	⬇️
py3.14	98.78% <100.00%> (-0.01%)	⬇️
ubuntu	96.70% <100.00%> (-0.01%)	⬇️
windows	98.71% <100.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jborean93]

Is there a real life scenario we can test with rather than doing it with the mocks? I definitely prefer proper integration tests over unit tests that are just testing the shape of the code.

[henrik-nil-acc]

Is there a real life scenario we can test with rather than doing it with the mocks? I definitely prefer proper integration tests over unit tests that are just testing the shape of the code.

Not very easy since it's a race condition.

This works against samba:
Point SMB_SERVER/SMB_SHARE/SMB_USER/SMB_PASSWORD at a writable share:

import gc, multiprocessing as mp, os, warnings

SERVER = os.environ["SMB_SERVER"]
SHARE = os.environ.get("SMB_SHARE", "share")
PORT = int(os.environ.get("SMB_PORT", "445"))
USER = os.environ.get("SMB_USER")
PASSWORD = os.environ.get("SMB_PASSWORD")
ROOT = rf"\\{SERVER}\{SHARE}\repro-359"
DIRS, FILES_PER_DIR, PROCESSES = 20, 50, 12


def _populate():
    from smbclient import makedirs, open_file, register_session
    from smbclient.shutil import rmtree
    register_session(SERVER, username=USER, password=PASSWORD, port=PORT)
    rmtree(ROOT, ignore_errors=True, port=PORT)
    makedirs(ROOT, exist_ok=True, port=PORT)
    for d in range(DIRS):
        sub = rf"{ROOT}\dir-{d:04d}"
        makedirs(sub, exist_ok=True, port=PORT)
        for f in range(FILES_PER_DIR):
            with open_file(rf"{sub}\file-{f:04d}.txt", mode="w", port=PORT) as fd:
                fd.write("x")


def _worker(queue):
    gc.set_threshold(1, 1, 1)
    count = 0

    def _show(message, category, *args, **kwargs):
        nonlocal count
        if category is ResourceWarning and "unclosed SMB handle" in str(message):
            count += 1

    warnings.simplefilter("always")
    warnings.showwarning = _show
    from smbclient import register_session
    from smbclient.shutil import rmtree
    register_session(SERVER, username=USER, password=PASSWORD, port=PORT)
    rmtree(ROOT, ignore_errors=True, port=PORT)
    gc.collect()
    queue.put(count)


def main():
    _populate()
    ctx = mp.get_context("spawn")
    queue = ctx.Queue()
    procs = [ctx.Process(target=_worker, args=(queue,)) for _ in range(PROCESSES)]
    for p in procs:
        p.start()
    total = sum(queue.get() for _ in procs)
    for p in procs:
        p.join()
    print(f"unclosed SMB handle warnings: {total}")


if __name__ == "__main__":
    main()

$ PYTHONWARNINGS=always python repro_359.py   # master
unclosed SMB handle warnings: 171
$ PYTHONWARNINGS=always python repro_359.py   # this PR
unclosed SMB handle warnings: 0

[jborean93]

Not very easy since it's a race condition.

That's fair enough thanks for clarifying.