EDIT: this is no longer the case, the API break is moved to #507

Since the change to Result<(), io::Error> is an API break, I'm willing to make a variant of this PR that just removes the debug_assert in PublicKey::write_into first together with all the other (not-breaking) changes and then PR the API change separately.

@stevenroose Yeah, I'd prefer that. This way we can get a "fix" in 0.26 even before we properly fix it.

K, I changed this PR to no longer change the API. Then I'll make another one for that.

LGTM
FWIW most of these asserts (the size_of + bytes) will be optimized out in release builds anyway

Oh is that like a cargo optimization? I didn't know about that, but well this approach more explicitly expresses that intent, IMO. And perhaps other Rust compilers aren't that smart 😅

Can you add a commit which bumps the minor crate version and adds a changelog entry? I think this is the only thing we're targeting for 0.25.2.

Done.

LGTM
FWIW most of these asserts (the size_of + bytes) will be optimized out in release builds anyway

Oh is that like a cargo optimization? I didn't know about that, but well this approach more explicitly expresses that intent, IMO. And perhaps other Rust compilers aren't that smart sweat_smile

No, it's called "constant propogation", when the compiler see things that it can evaluate in compile time it will evaluate and change the branches accordingly,
i.e.

if 5< 3 {
    println!("One");
} else {
    println!("Two");
}

Will const-propogate to:

println!("Two");

so also:
assert_eq!(::std::mem::size_of::<$type>(), $byte_len); will macro expand into i.e.
assert_eq!(std::mem::size_of::<u64>(), 8); -> assert_eq!(8, 8); -> {if 8 != 8 {panic!("assertion failed....")}} -> {} :)