That jar fixed the re-init warnings for my environment.rb constants, the original NameError still persists.
I think the problem may be due to our code using custom load paths with non-namespaced constants.
For example, we have modules like Authentication that are in app/mixins/authentication.rb -- if I change the class name to Mixins::Authentication, GF seems to be able to load it. I realized this after I removed the observer class altogether -- there were more NameErrors, among them the one for the Authentication module. Subsequent NameErrors also popped up whenever I visited pages that reference models in subfolders (e.g. app/models/features/some_feature.rb) -- if I try to namespace these, I get a rails error saying the file didn't define the top-level constant rails expected...for example, if I change PointObserver to Observers::PointObserver, I get "Expected app/observers/point_observer.rb to define PointObserver".
This may not be the most scientific approach of isolating the issue but its really difficult to repro any other way (console & tests show no errors). Hopefully this will provide some direction for a resolution.
[Message sent by forum member 'mwadhera' (mwadhera)]
http://forums.java.net/jive/thread.jspa?messageID=315945