statement-distribution: Add tests for incoming acknowledgements

[mrcnski]

In #2329 we noticed that some async backing functions were missing tests. This adds a few tests for receiving acknowledgements. More tests will come in a future PR.

I also started doing the refactor described in #2382, plus some more refactor ideas I had. It would have made this PR way too big to finish this refactor, so I only did it for the new tests. If it looks good, I'll raise another PR refactoring all the tests. Personally I think it makes the tests much easier to read, write, and reason about. For me it's important because I read the tests to find out how async backing is expected to behave in certain situations.

Related

See #2329
See #2382

[mrcnski]

@AndreiEres brought up a good point about these asserts:

assert_peer_reported(&mut overseer, peer_c, BENEFIT_VALID_STATEMENT).await;
assert_peer_reported(&mut overseer, peer_c, BENEFIT_VALID_STATEMENT).await;
assert_peer_reported(&mut overseer, peer_c, BENEFIT_VALID_STATEMENT).await;
assert_peer_reported(&mut overseer, peer_c, BENEFIT_VALID_RESPONSE).await;

If one of the asserts here fails, the error will point to inside the assert function and not where the assert is called in code (what we care about). Actually this is an existing issue with these tests and it's annoyed me before.

I believe we could fix it by making these asserts macros instead of functions. Indeed, I just tested it and it works.

tests/grid.rs:

assert_peer_reported!(&mut overseer, peer_c, BENEFIT_VALID_STATEMENT);
assert_peer_reported(&mut overseer, peer_c, BENEFIT_VALID_STATEMENT).await;
assert_peer_reported(&mut overseer, peer_c, BENEFIT_VALID_STATEMENT).await;
assert_peer_reported(&mut overseer, peer_c, BENEFIT_VALID_RESPONSE).await;

tests/mod.rs:

#[macro_export]
macro_rules! assert_peer_reported {
    ($virtual_overseer:expr, $peer_id:expr, $rep_change:expr) => {
        assert!(false);
        assert_matches!(
            $virtual_overseer.recv().await,
            AllMessages::NetworkBridgeTx(NetworkBridgeTxMessage::ReportPeer(ReportPeerMessage::Single(p, r)))
                if p == $peer_id && r == $rep_change.into()
        );
    }
}

Result:

thread 'v2::tests::grid::received_advertisement_after_backing_leads_to_acknowledgement' panicked at polkadot/node/network/statement-distribution/src/v2/tests/grid.rs:528:13:

tl;dr: TIL that custom asserts should always be macros.

[Sophia-Gold]

This naming was already confusing to me and this change makes it a little more so. Can we take this opportunity to rename it everywhere here? Maybe just use local_group and other_group (or our_group and outside_group like you have below) and have a comment that says we're targeting the other_group because this is testing grid mode.

[mrcnski]

Will do, thanks for calling it out. Not sure about our and outside - for groups it kind of makes sense, but IMO outside_para would be a head-scratcher. To me, local and other make sense: local_validator, local_para, other_para, etc. I'll standardize on those.