Only suggest trait method completions when all (method) bounds are satisfied.

[Philipp-M]

Consider the following:

struct Color;
trait HasProperty<P> {}
trait Style {
    fn color(self)
    where
        Self: Sized + HasProperty<Color>,
    {
    }
}

impl<T> Style for T {}
struct WithoutColor;

I'd expect that WithoutColor.| doesn't show any color() as it doesn't fullfill the method trait bounds, but it does, and after completion cargo check fails (i.e. shows an error).

I haven't seen these kind of issues with super-trait-bounds on traits, something like Style: HasProperty<Color> (i.e. works as expected).

(This is a trimmed down example of an actual case in Xilem, in case you're interested, you can try the examples, there's a Style trait that does something similar).

[ChayimFriedman2]

Completion is already too slow due to a lot of trait solving. Evaluating the function's bounds will be even slower. I'm not saying we shouldn't do this, but we definitely should very carefully weigh the benefits against the costs.