DetachedInstanceError on login

[Zitrax]

This might be due to my setup, but I am not yet sure exactly why I see this so I am filing an issue.

The confusing part is that I see this on one machine but not the other, and I have not yet pinpointed the difference. Could be some setup/environment that is different - or simply some timing issue.

Anyway this happens consistently when I try to login:

Traceback (most recent call last):
  File "..\lib\site-packages\waitress\channel.py", line 338, in service
    task.service()
  File "..\lib\site-packages\waitress\task.py", line 169, in service
    self.execute()
  File "..\lib\site-packages\waitress\task.py", line 399, in execute
    app_iter = self.channel.server.application(env, start_response)
  File "..\lib\site-packages\pyramid\router.py", line 223, in __call__
    response = self.invoke_subrequest(request, use_tweens=True)
  File "..\lib\site-packages\pyramid\router.py", line 198, in invoke_subrequest
    response = handle_request(request)
  File "..\lib\site-packages\pyramid_debugtoolbar\toolbar.py", line 266, in toolbar_tween
    toolbar.process_response(request, response)
  File "..\lib\site-packages\pyramid_debugtoolbar\toolbar.py", line 87, in process_response
    panel.process_response(response)
  File "..\lib\site-packages\pyramid_debugtoolbar\panels\request_vars.py", line 112, in process_response
    extracted_attributes = extract_request_attributes(self.request)
  File "..\lib\site-packages\pyramid_debugtoolbar\panels\request_vars.py", line 56, in extract_request_attributes
    if not hasattr(request, attr_):
  File "..\lib\site-packages\pyramid\security.py", line 344, in authenticated_userid
    return policy.authenticated_userid(self)
  File "..\lib\site-packages\pyramid\authentication.py", line 90, in authenticated_userid
    callback_ok = self.callback(userid, request)
  File "..\lib\site-packages\pyramid_fullauth\auth.py", line 28, in groupfinder
    if user and user.id == userid:
  File "..\lib\site-packages\sqlalchemy\orm\attributes.py", line 237, in __get__
    return self.impl.get(instance_state(instance), dict_)
  File "..\lib\site-packages\sqlalchemy\orm\attributes.py", line 578, in get
    value = state._load_expired(state, passive)
  File "..\lib\site-packages\sqlalchemy\orm\state.py", line 474, in _load_expired
    self.manager.deferred_scalar_loader(self, toload)
  File "..\lib\site-packages\sqlalchemy\orm\loading.py", line 610, in load_scalar_attributes
    (state_str(state)))
sqlalchemy.orm.exc.DetachedInstanceError: Instance <User at 0x497af10> is not bound to a Session; attribute refresh operation cannot proceed

so apparently the request.user instance is no longer bound to a session. Was it closed too early ?

Adding these lines to groupfinder() in auth.py:

    if user:
        from sqlalchemy.orm.session import object_session
        object_session(user).refresh(user)

right before the user variable is used reattaches it and all seem to work fine. Possibly there could also be a check to only use refresh if really needed.

However, I am not sure why I am even seeing this problem. Any idea ?

[Zitrax]

The workaround above still failed in some cases, where object_session returned None.

Another seemingly more reliable version of avoiding that exception is to change the implementation of user(request) in request.py to:

    if request.unauthenticated_userid:
        try:
            user = pyramid_basemodel.Session.query(User).filter(
                User.id == request.unauthenticated_userid
            ).options(joinedload('groups')).one()
            pyramid_basemodel.Session.expunge(user)
            return user
        except NoResultFound:  # pragma: no cover
            pass

The call to expunge to make sure the user object can be used even if detached, and the joinedload of groups since we access that too.

[fizyk]

Okay, it looks like it fails on the first request to load?
What's your session configuration? I have to admit It's the place I'd expect this to happen the least.

[Zitrax]

I am importing Session from pyramid_basemodel.

I noticed one machine was on pyramid 1.6.1 and the other 1.7. But updating that made no difference.

[Zitrax]

I did notice that the traceback contained pyramid_debugtoolbar and indeed turning that off removed the problem. Turns out with that enabled groupfinder() is called twice and fails with the exception the the second time.

[fizyk]

Okay then... this should not have happened... groupfinder doesn't have much in it to justify such behaviour. and neither does request.user.... You're using transactions package, right?