Findings from CI in Multiprecision on MSVC 14_0

[ckormanyos]

In Multiprecision Issues like this we find stuff maybe relevant for Math.

• MSVC 14.0 does not comprehend /std:c++17 and the CI run(s) should use [14, latest] in the standard matrix.
• MSVC 14.0 seems to not comprehend the preprocessor jargon for testing if it has the header <stdfloat>. We need to do some preprocessor exclusions around here. This one is probably running wild in the field so maybe we get to it for 1.84.

On the second topic, something like below?

#if !(defined(_MSC_VER) && (_MSC_VER <= 1900))
#if __has_include(<stdfloat>)
#  include <stdfloat>
#endif

But I don't know 1f 14.1 comprehends __has_include so we need to check that too.

[ckormanyos]

I can take a crack at these items.

Cc: @jzmaddock and @mborland

[jzmaddock]

You should use #ifdef __has_Include to verify it's existence.

[ckormanyos]

use #ifdef __has_Include to verify it's existence

#if (defined(__has_include) && __has_include(<stdfloat>))
#  include <stdfloat>
#endif

Like that. Thx John.
I will try. But it was a problem with lexical parsing so I'm not sure it works. I will report after trying it out...

[ckormanyos]

See also #1017

[mborland]

Since MSVC 14.0 is not really C++14 compliant this is another vote from me to drop support for it entirely. I don't think its worth the effort to add workarounds for something that we're going to continue to break as the code is modernized/features added.

[ckormanyos]

Since MSVC 14.0 is not really C++14 compliant this is another vote from me to drop support for it entirely. I don't think its worth the effort to add workarounds for something that we're going to continue to break as the code is modernized/features added.

Yea. I'd be ready to drop it any time. It's going to be a nuisance with constexpr anyway, ...

Yet, this workaround at the moment is just about finished. I'd just go ahead and finish it. Then we could continue the drop vc140 discussion at another place/time?