Nathanael Nerode wrote:
I actually thought that there should be two keywords for this:
missing-diagnostic (There should be a warning, but there isn't.)
misleading-diagnostic (There's a warning which shouldn't be there, or
should be replaced by a different warning.)
The latter (misleading-diagnostic) can apply also to error messages. There
are situations where error messages are totally wrong (for instance, they
show unreadable names because they can't pretty-print correctly them, or
they are totally unrelated to the real error), or could be improved (for
instance, a simply "parse error", or "expected primary-expression" instead
of something more detailed which would help the user understanding the
problem).
But in the first place, why do you think that two keywords could be useful?
I think the biggest difference is between diagnostic bugs which are request
for enhancements and not strictly necessary, and diagnostic bugs which are
really bugs (or at least, can be considered as such). And this difference is
already taken into account by the Severity of the bug. In other words I
suggest that your missing-diagnostic be simply a diagnostic keyword with
Severity = enhancement, and you misleading-diagnostic be a diagnostic
keyword with Severity != enhancement.