approx_constant "Known problems" section is incorrect

[kraai]

The documentation for approx_constant says in the "Known problems" section:

If you happen to have a value that is within 1/8192 of a known constant, but is not and should not be the same, this lint will report your value anyway. We have not yet noticed any false positives in code we tested clippy with (this includes servo), but YMMV.

The function that compares the values has the following header comment:

/// Returns `false` if the number of significant figures in `value` are                                                                               
/// less than `min_digits`; otherwise, returns true if `value` is equal                                                                               
/// to `constant`, rounded to the number of digits present in `value`.                                                                                

There's nothing in that comment, or that I can see in the code, referring to 1/8192, so I think the documentation is incorrect.

[flip1995]

I don't really know where the number 1/8192(=2^-13) is coming from. I assume it comes from the precision of float numbers expressed in decimal digits.

cc @llogiq wrote the lint and documentation ~4 years ago. Maybe he can shed some light onto this magic number.

[flip1995]

In the initial implementation there was a magic EPSILON_DIVISOR constant:

rust-clippy/src/approx_const.rs

Line 23 in a927bc0

const EPSILON_DIVISOR : f64 = 8192f64; //TODO: test to find a good value

(old like from 09/2015)

This constant was removed in #408 and I think this Known Problems is not an issue anymore.