`no_std` regression in scorer

[devrandom]

https://gitlab.com/lightning-signer/validating-lightning-signer/-/merge_requests/210

error[E0599]: no method named `log10` found for type `f64` in the current scope
   --> /usr/local/cargo/registry/src/github.202132.xyz-1ecc6299db9ec823/lightning-0.0.105/src/routing/scoring.rs:744:26
    |
744 |         (-(success_probability.log10()) * liquidity_penalty_multiplier_msat as f64) as u64
    |                                ^^^^^ method not found in `f64`
For more information about this error, try `rustc --explain E0599`.

[TheBlueMatt]

Grrr, really annoying CI didn't somehow catch this. I'm in general a bit confused how cargo test --features no-std doesn't hit this? To address the immediate issue it seems we'll need to provide our own log method, given we can't rely on libm existing. Do you still have that log-approximation code handy @jkczyz ?

[jkczyz]

Yep, right here though not based on main: jkczyz@d2806cb

Will require updating the tests as well, unfortunately.

[jkczyz]

Here's the full branch as it uses a Probability enum which was ultimately removed.

https://github.com/jkczyz/rust-lightning/commits/2022-01-log-approximation

[devrandom]

I can't figure out why the crate compiles with --no-default-features --features no_std. f64::log10 is definitely in the std crate.

[jkczyz]

@TheBlueMatt Are you thinking we should just use log approximation unconditionally?

[TheBlueMatt]

I feel like that would be the simplest - if we're happy with the accuracy for no-std we should just use it for std, too, no?