Bug 23845 – std.math.traits.copysign returns the wrong value

As written, the function simply returns its first parameter. Should likely return cast(R) *pto, though my pointer syntax is rusty—surely doing all that work with pto should have *some* point (no pun intended). I’d honestly bump this up to critical, but if no one’s noticed this yet, maybe it shouldn’t be—except copysign() can be called from cbrt(), at the very least. It keeps cbrt() from being inferred pure on the MS runtime, I think; if I’m wrong, that’s a separate bug.

Did this slip through because unit tests are run at compile time? Because the unit tests sure as hell should’ve caught this. I’d also recommend keeping template parameter order consistent between the versions of copysign(); it’s not a bug, but it could damned easily cause one if someone’s not paying attention.

My pointer syntax is much rustier than I’d realized, so…ignore this. cbrt() should still be pure, though; a cube root function *not* being pure is…surprising, at the least.