[PATCH] Define iszero as a function template for C++ [BZ #20715]

[Florian Weimer]

On 10/21/2016 06:12 PM, Joseph Myers wrote:
> On Fri, 21 Oct 2016, Florian Weimer wrote:
>
>> Tested on x86_64, ppc64, ppc64le, s390x, aarch64.
>>
>> I'm not sure if the new test is particularly meaningful, but at least it
>> catches syntax errors in <math.h>.
>
> This seems fine to me (with negative zero included in the testcase),

Oh.  How do I get a negative zero?  Would T{-0.0} be sufficient?  Or 
should I divide the negative denormal with the smallest absolute value 
by two?

> and
> in accordance with how C++ defines bindings for C99 macros such as
> infinite.  I presume you're deliberately leaving it to future libstdc++
> versions to handle the other is* macros from TS 18661-1 if they get into
> C2x and then into C++, as you haven't observed them causing problems for
> C++ code at present.

Right, I want to keep the difference between the C and C++ versions to a 
minimum, and only address problems we've seen so far.

Thanks,
Florian