Warn of obscure floating-point literals

[tamaroning]

What it does

This lint warns of misleading floating-point literals.
In the example, a has type i32 because floating-point literals must start with a decimal literal.
Instead of suffixes, we should use type ascriptions.

Lint Name

obscure_float_literal

Category

style

Advantage

• Avoid ambiguity

Drawbacks

none

Example

let a = 0x1_f32; // == 0x1F32

Could be written as:

let a = 0x1 as f32;

[xFrednet]

This can be a lint. However, it should be in the restriction or some other allow-by-default group, as this is very debatable. We have some lints like default_numeric_fallback that suggest using the type suffixes. I also prefer having the type as a suffix 🙃

I also prefer having the type as a suffix

The issue is that the _f32 here is not a type suffix.

It's literally just 0x1f32 (Which, in decimal, is 7986) which is a perfectly valid hex literal for an i32. You can't denote a hex literal to be a float.

So this is incredibly misleading code, and should probably be in suspicious, if not correctness (if the user writes _f32 or _f64)

[xFrednet]

Ahh yes, I misunderstood the issue. Thank you very much for pointing that out! Then it should be in suspicious or correctness as you suggested. 🙃

[hellow554]

@rustbot claim