need explicit function call for c_long_double

[WalterBright]

Supplemental needed for dlang/druntime#958

Yes, this makes it a regression, but an unavoidable one.

[9rnsr]

Auto-merge toggled on

[dnadlinger]

So now we go out and tell everybody to change all their C API? Somehow, just making real 64 bit on Win64 (and maybe providing a special __real80 type or whatever) just seems a lot more attractive.

[WalterBright]

So now we go out and tell everybody to change all their C API?

This has always been true for Win64. It's why we have c_long_double.

Somehow, just making real 64 bit on Win64 (and maybe providing a special __real80 type or whatever) just seems a lot more attractive.

But what about the poor saps who need higher precision and have the hardware for it? If others are happy with 64 bit precision, they can use double and be done with it.

[d-random-contributor]

They would
import core.hw : float80;