Some std::f32 constants are actually f64s

[zhanrnl]

std::f32::MAX_VALUE, std::f32::MIN_VALUE, and std::f32::EPSILON are all actually f64s, not f32s as expected. FLT_MAX and FLT_MIN in C are all floats, not doubles, for example. If this is intentional, it's very confusing to me why any constants in the std::f32 module should be f64s, and probably should be explained in the documentation.

[adrientetar]

MAX_VALUE isn't defined in f32.rs. I don't know how it works but it must be kinda redirected to f64.

[zhanrnl]

Unless I'm completely misunderstanding something, I think it is. Interestingly, the source references issue #1433, which doesn't seem related.

[huonw]

@zhanrnl #1433 is related (to the constants themselves, but not to this issue), because the statics should be written in hex to guarantee that they have the exactly correct value.

[aturon]

Note: this issue and issue 13297 should be resolved together.