[m-dev.] question about warnings for should-be-singleton-but-aren't variables

[Zoltan Somogyi]

The attached diff describes a potential problem with
how we generate warnings for variables whose names start
with an underscore, but which occur more than once.
The new test case in the diff exhibits the issue, and asks
the question. The language manual does not explicitly say
what the answer should be either. What do you guys think?

Zoltan.