While looking at a bug I found that expr-cumulative does not handle one
case of explicit calls to an operator:
(gdb) p operator==(s1,s2)
A syntax error in expression, near `operator==(s1,s2)'.
The bug is that the name_not_typename production does not handle
operator.
This patch fixes the problem and adds a couple new test cases -- one for
the bug and one for an explicit call to an operator as a member
function; I couldn't find one of these in the test suite.
Built and regtested on x86-64 (compile farm).
Ok?