Better constraint that Output is ProjectivePoint

[daxpedda]

I encountered quiet the issue:

fn test<C: GroupDigest>() {
    let var: ProjectivePoint<C> =
        C::hash_from_bytes::<ExpandMsgXmd<sha2::Sha256>>(&[b"test"], b"test");
}

Gives the following error

error[E0308]: mismatched types
  --> src/group/elliptic_curve.rs:34:9
   |
33 |     let var: ProjectivePoint<C> =
   |              ------------------ expected due to this
34 |         C::hash_from_bytes::<ExpandMsgXmd<sha2::Sha256>>(&[b"test"], b"test");
   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected associated type, found enum `core::result::Result`
   |
   = note: expected associated type `<C as GroupDigest>::Output`
                         found enum `core::result::Result<<C as GroupDigest>::Output, elliptic_curve::Error>`
help: consider constraining the associated type `<C as GroupDigest>::Output` to `core::result::Result<<C as GroupDigest>::Output, elliptic_curve::Error>`
   |
32 | fn test<C: GroupDigest<Output = core::result::Result<<C as GroupDigest>::Output, elliptic_curve::Error>>>() {
   |                       ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

For more information about this error, try `rustc --explain E0308`.

Doing this:

fn test<C>()
where
    C: GroupDigest<Output = ProjectivePoint<C>>,
{
    let var: ProjectivePoint<C> =
        C::hash_from_bytes::<ExpandMsgXmd<sha2::Sha256>>(&[b"test"], b"test");
}

Doesn't help either

error[E0275]: overflow evaluating the requirement `<C as ProjectiveArithmetic>::ProjectivePoint == _`
  --> src/group/elliptic_curve.rs:32:1
   |
32 | / fn test<C>()
33 | | where
34 | |     C: GroupDigest<Output = ProjectivePoint<C>>,
   | |________________________________________________^

For more information about this error, try `rustc --explain E0275`.

I couldn't find any other solution to this then removing the associated type Output and adding a where-clause to constraint ProjectivePoint to CofactorGroup.

Alternatively I would suggest adding clear_cofactor to some other trait required by ProjectivePoint or require CofactorGroup for ProjectivePoint.

CC @tarcieri @mikelodder7

[tarcieri]

Seems fine to me. @mikelodder7 any thoughts?

[mikelodder7]

Seems fine to me. This whole thing is complex unfortunately.