I just tripped across this (prematurely closed) bug:
http://dev.rubyonrails.org/ticket/7460
In a nutshell, the wrong exception is being rescued, LoadError instead
of MissingSourceFile, and thus if a LoadError is thrown (for instance
if the helper file doesn't define the expected helper module) it gets
masked by a NoMethodError due to the non-existent #is_missing? getting
called on the LoadError.
The patch is a one-liner to change the rescue clause - do I hear some +1's?
