On Thu, Nov 12, 2009 at 11:51 AM, John Maddock <john@johnmaddock.co.uk> wrote:
Indeed, Variant shouldn't suppress the warning, but wasn't the original issue because a test was triggering it? IOW, it is the test code that should work around or suppress the warning.
This is what the test tests. If it "works around the warning" it is no longer a test. You may as well remove it, which will also silence the warning.
The test tests that Variant produces a warning?
No the test verifies that:
variant = type-convertible-to-actual-type-in-variant;
works as expected.
The warning is a byproduct of the fact that the type stored in the variant is a short, and the type convertible to it is an int.
This may seem too obvious and I'm probably not knowledgeable enough about the test, but does a widening conversion accomplish the same thing, only without the warning? e.g. variant stores int, try to assign short --Michael Fawcett