LO calc formulas

[Stephan Bergmann]

On 14/12/2020 08:46, Rene Engelhard wrote:
> Am 14.12.20 um 07:57 schrieb Mike Kaganski:
>> On 11.12.2020 23:41, Dante Doménech wrote:
>>> We have our own isnan, isinf, isfinite, ... and some of them with
>>> quite strange code.
>>> Why is the default c++ standard library isn't good enough?
>>
>> See also https://bugs.documentfoundation.org/show_bug.cgi?id=130975
>> and https://bugs.documentfoundation.org/show_bug.cgi?id=130976.
>>
> I wonder why these commits didn't change sal to just call the library
> function?
> 
> Stephan only said that the function must not go because it's part of the
> interface, but that doesn't say one can't just make it a function
> calling the library function?
> 
> That should be ABI compatible, shouldn't it?

Taking <https://bugs.documentfoundation.org/show_bug.cgi?id=130975> 
"Drop rtl::math::isNan in favor of std::isnan" as an example: 
LO-internal code can and should call std::isnan directly.  But 
include/rtl/math.hxx must continue to provide an rtl::math::isNan that 
works with pre-C99, pre-C++11 compilers, so that inline function cannot 
call std::isnan internally.  (And as no internal code should call it, 
why bother with adding, say, a LIBO_INTERNAL_ONLY branch to its 
implementation that would call std::isnan.  What we could do though is 
mark rtl::math::isNan as deprecated, at least for the LIBO_INTERNAL_ONLY 
case.)