fix machine type equality

[boggle]

float, int, uint should be equal to their machine type equivalents. This is currently not the case and originally was the reason for the introduction of trivial casts. Now that these are gone, this needs fixing, i.e.

let f: float = fXY::sin(1.0)

should compile.

[brson]

If we do this then the line let i: int = 0i32 would compile on 32-bit platforms but not on 64-bit platforms. This strikes me as undesirable.

[boggle]

Yes, i can see this issue. On the other hand it stops me from exporting core::f64|f32 math as core::float depending on architecture.

I'm not vested in any particular solution to this as long as there is one.

[graydon]

brson: that risk is already present with declarations like let i : int = 0x1000_0000_0000_0000;, it'll compile on x64 but not x86. This is the inherent meaning of using int or float: "please make the following code at least slightly platform-specific"

[boggle]

We agreed to go for this now, unless strong objections are raised.