Comment 40 for bug 978896

I applied the patch to stable/essex and ran the unit tests without seeing any failures. That would lead me to believe there's something different in the upstream package/environment that's causing the issue. It's hard to debug that without more details.

Alternatively, the patch could simply be amended to wrap that del call inside an "if hasattr(request, "_cached_user")..." to allow it to fail more gracefully. That said, without knowing why that attribute is not present I can't definitely say that gracefully passing by that line won't cause other problems.

Perhaps Paul can comment on that.