Subject: Re: [boost] Official warnings policy?
From: Emil Dotchevski (emildotchevski_at_[hidden])
Date: 2009-11-12 12:52:30
On Thu, Nov 12, 2009 at 8:48 AM, Stewart, Robert <Robert.Stewart_at_[hidden]> wrote:
> Peter Dimov wrote:
>> Stewart, Robert wrote:
>> > The test tests that Variant produces a warning?
>> The test tests whether an int can be assigned to a variant, one
>> of whose types is a short. It should produce a warning; in this
>> context, this is a feature. User code that does the same should
>> also produce a warning. This is what warnings are for.
> Shouldn't the test confirm that an int can be assigned to a short in
> a variant and that the resulting short has the same value as when
> the same int is assigned to a short not in a variant? That test need
> not produce a warning because comparing the results from the two
> assignments proves or disproves the behavior of variant.
Do you agree that getting a warning in this use case is a good thing?
Assuming you do, don't you find value in a confirmation that the
warning is reported, as it should?
What's next, fixing compile errors in compile-fail tests? :)
Reverge Studios, Inc.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk