Add 'remark' diagnostic type in clang

[Tobias Grosser]

On 02/28/2014 10:56 AM, Chandler Carruth wrote:
> On Fri, Feb 28, 2014 at 1:26 AM, Tobias Grosser <tobias at grosser.es> wrote:
>
>> I would have thought that the difference is that a remark indicates
>>> something that *isn't* a problem, and is just informational. That being
>>> the
>>> case, it doesn't make sense to me to upgrade remarks to warnings or to
>>> errors. It would seem bizarre to me if -Werror converts remarks to errors
>>> (and indeed, in this patch, it does not).
>>>
>>
>> Good point. I explicitly mentioned that '-Werror' does not trigger on
>> remarks in the commit message.
>>
>>
>>> I think we should also reject
>>
>>> -Werror=some-remark.
>>>
>>
>> Several people mentioned that they like this feature including Chris. This
>> either needs more discussion or just more experience when some
>> remarks have been added.
>
>
> I actually really agree with Richard here. If these were things that made
> sense to "escalate" they would be warnings. I think a good metric for
> whether something should be a warning or a remark is whether it really ever
> makes sense to break the compilation on it because it indicates a problem.

I was thinking the same. The only reason I did not want to set a strict 
border is that I see remarks more often coming from the backend, which 
means they may change between different compiler versions and 
optimization levels. This does not seem to make something that could 
reliably used as warning.

Tobias