numtraits libm feature and num-bigint-dig directly depends on C library libm

[AlexanderSchuetz97]

Hello there,
you are including the numtraits crate with the libm feature enabled.
This causes linking with libm.

I am trying to use this crate in a scenario where linking to libm (or other C code) is not possible (Complex cross compilation).
I assumed since you mention that this is pure rust that it does not rely on C code.

If its possible could you perhaps evaluate removing this dependency on C code and if it seems feasible remove it?

If its not possible or would require too much effort then feel free to close this issue,
but perhaps a small clarification in the readme that this crate does
link libm would have saved me some time.

[tarcieri]

Can you please file an issue against crypto-primes? I didn't realize they made libm a hard dependency or I would've recommended against that:

https://github.com/entropyxyz/crypto-primes

Looks like it got added in this PR which could at least trivially be made feature-gated or implemented without libm:

entropyxyz/crypto-primes#81

[AlexanderSchuetz97]

Hello there,
I don't think this crates is the culprit.

The dependency tree is like this:
rsa->num-bigint-dig-->libm
aswell as
rsa->num-bigint-dig--> numtraits -->libm

I overlooked this num-bigint-dig direct dependency when creating the issue. I will adjust the title, Its not just numtraits.

The num-bigint-dig crate already has a issue open where its stated that the libm dependency is there to stay. (Please double check I am unsure If I am reading this correctly)
dignifiedquire/num-bigint#24

So I dont think this can be resolved without removing num-bigint-dig from this crate.

So I will not be opening a Issue on crypto-primes because I do not even see this crate in my dependency tree.
I am currently using rsa 0.9.10.
This may be different on the 0.10 release but since its a pre release I did not select it for usage.

Sincerely
Alexander Schütz

[tarcieri]

We can't really do anything about the v0.9 series, but if you don't want this to be an issue in v0.10 where we have migrated from num-bigint-dig to crypto-bigint, I would suggest opening an issue on crypto-primes in advance.

We will hopefully have rsa v0.10 stable out in a few months.

[AlexanderSchuetz97]

I am fine with this being only solved in the 0.10.X versions.

I will see if I can find the time to check if a 0.10.X pre release would work on my application in general and then hopefully verify the crypt-primes libm thing aswell as check if theres any other C code dependencies.