lp.testing.publication.test_traverse doesn't set the current participation

[Michael Hudson-Doyle]

test_traverse's docstring claims that it "uses the currently logged in user" which is all well and good, but this means that it does not set up a new interaction, which in turn means that you will end up with a different request object being current as per get_current_browser_request than the one that's having .traverse() called on it. This can be very confusing.

I have a two line fix, but I'm not completely sure of it...

[Gary Poster (gary)]

From one perspective, a reasonable fix would be to do the following within test_traverse.

(before traversing)
from zope.security.management import endInteraction, newInteraction, restoreInteraction
endInteraction()
newInteraction(request)

(after traversing)
restoreInteraction()

We also could rip out the current participation (request) from the current interaction and replace it temporarily. This is a bit riskier; typically you want to make sure that the interaction (security policy instance) has the ability to clear any decision caches.

What did you have in mind?

[Michael Hudson-Doyle (mwhudson)]

On Fri, 30 Jul 2010 21:57:41 -0000, Gary Poster <email address hidden> wrote:
> >From one perspective, a reasonable fix would be to do the following
> within test_traverse.
>
> (before traversing)
> from zope.security.management import endInteraction, newInteraction, restoreInteraction
> endInteraction()
> newInteraction(request)
>
> (after traversing)
> restoreInteraction()
>
> We also could rip out the current participation (request) from the
> current interaction and replace it temporarily. This is a bit riskier;
> typically you want to make sure that the interaction (security policy
> instance) has the ability to clear any decision caches.
>
> What did you have in mind?

I had something like the first solution you proposed in mind, although I
hadn't heard of restoreInteraction().

I got bogged down in trying to write tests for test_traverse though --
slightly scary stuff! And then my lucid install ate itself :(

Cheers,
mwh

[Michael Hudson-Doyle (mwhudson)]

Gary, I think I have a fix in the attached branch. If you could have a look, that would be great. If you don't get to it, I'll add some more tests and comments and stuff tomorrow and propose it for review.

[Michael Hudson-Doyle (mwhudson)]

I found a few minutes to write some more tests, fix a few more bugs and propose the branch for merging.

[Ursula Junque (ursinha)]

Fixed in stable r11300 <http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/11300>.