Bump synedrion to 0.3.0 and use manul for protocol session loop

[ameba23]

Synedrion 0.3.0 and manul 0.2.0 have been released - this PR updates to using the new versions.

manul now gives as a function which implements the protocol execution loop. Which means we can move it out of entropy-protocol and use the manul-provided one. This is good as less code for us to maintain. However there are a couple of reasons why it is not a drop-in replacement:

• manul's par_run_session function takes a tokio::sync::mpsc::unbounded::Sender but we have a tokio::sync::broadcast::Sender
• Our ProtocolMessage needs to be replaced with manul::session::tokio::MessageIn and manul::session::tokio::MessageOut - but we would lose the extra bits protocol message gives us (eg: being able to pass the verifying key)

We get around this by using extra tasks/channels to pass messages into the manul execution loop. This feels a bit ugly but i don't think it has a big performance cost, so i am happy with it for now.

Other issues:

• I had issues with using the TestParams and switched to always using the production parameters in tests. This will increase our CI time - but i think not by much as many of our tests already used production parameters. If people think its important i can make a followup putting back the test params - i think we just to add elliptic-curve as a dependency to entropy-protocol.
• Manul now gives us a SessionReport with more details about protocol errors which are useful for slashing. But we don't yet do anything with it. I propose to do that in a follow up.

[ameba23]

@fjarri @dvdplm i am a bit stuck on how to do the serialization of the protocol messages.

manul::session::Message does not implement serde::Serialize or serde::Deserialize and has private fields preventing me from implementing them myself on a wrapper type.

I have tried my best to copy what you did with the serializer for postcard in the examples, with the serialization format we are using (bincode):

My code:

entropy-core/crates/protocol/src/lib.rs

Lines 127 to 167 in 1ed6627

	impl manul::session::SessionParameters for EntropySessionParameters {
	type Signer = execute_protocol::PairWrapper;
	type Verifier = PartyId;
	type Signature = sr25519::Signature;
	type Digest = Blake2s256;
	type WireFormat = BincodeWireFormat;
	}
	#[derive(Debug)]
	pub struct BincodeWireFormat;
	impl manul::session::WireFormat for BincodeWireFormat {
	fn serialize<T: Serialize>(value: T) -> Result<Box<[u8]>, manul::protocol::LocalError> {
	Ok(bincode::serialize(&value).unwrap().into())
	}
	type Deserializer<'de> = PersistentDeserializer<BincodeDeserializer<'de>>;
	fn deserializer(bytes: &[u8]) -> Self::Deserializer<'_> {
	PersistentDeserializer::new(BincodeDeserializer(bincode::de::Deserializer::from_slice(
	bytes,
	bincode::config::DefaultOptions::new(),
	)))
	}
	}
	/// A wrapper for a bincode deserializer.
	#[allow(missing_debug_implementations)]
	pub struct BincodeDeserializer<'de>(
	bincode::de::Deserializer<bincode::de::read::SliceReader<'de>, bincode::config::DefaultOptions>,
	);
	impl<'de> AsTransientDeserializer<'de> for BincodeDeserializer<'de> {
	type Error = bincode::Error;
	fn as_transient_deserializer<'a>(
	&'a mut self,
	) -> impl serde::Deserializer<'de, Error = Self::Error> {
	&mut self.0
	}
	}

Your code: https://github.com/entropyxyz/manul/blob/4f269080775aa6359d0d26d8b56940b4429882f7/manul/src/dev/wire_format.rs#L10-L38

But it seems to me like i don't have a example of how the serialization is used - in the examples messages don't actually go over the wire so they never need to be serialized / deserialized: https://github.com/entropyxyz/manul/blob/4f269080775aa6359d0d26d8b56940b4429882f7/manul/src/dev/run_sync.rs#L192-L221

[fjarri]

Should be fixed by entropyxyz/manul#98

[ameba23]

Thanks @fjarri that works - as in things now compile (tests not yet passing).

[ameba23]

It seems that for both the 'key init' and 'reshare' protocols, messages do get exchanged but then the session hangs before the round is finalized.

Eg: Here is the key init protocol running with two parties (the first stage of a DKG with 3 parties):

2025-04-28T10:58:06.830113Z DEBUG execute_dkg: entropy_protocol::execute_protocol: Executing DKG
2025-04-28T10:58:06.830334Z DEBUG execute_dkg: entropy_protocol::execute_protocol: Executing DKG
2025-04-28T10:58:06.830350Z DEBUG execute_dkg: entropy_protocol::execute_protocol: Executing DKG
2025-04-28T10:58:06.830631Z DEBUG execute_dkg: entropy_protocol::execute_protocol: Session starting
2025-04-28T10:58:06.830646Z DEBUG execute_dkg: manul::session::tokio: PartyId(AccountId32([208, 139, 181, 149, 207, 194, 199, 205, 52, 230, 2, 226, 50, 174, 189, 144, 20, 77, 65, 104, 216, 189, 155, 61, 133, 190, 124, 33, 30, 145, 43, 24])): *** starting round RoundId { round_nums: [1], is_echo: false } ***
2025-04-28T10:58:06.830733Z DEBUG execute_dkg: entropy_protocol::execute_protocol: Session starting
2025-04-28T10:58:06.830745Z DEBUG execute_dkg: manul::session::tokio: PartyId(AccountId32([76, 94, 254, 229, 66, 181, 20, 80, 20, 240, 21, 163, 237, 36, 116, 129, 134, 66, 26, 57, 48, 71, 122, 38, 221, 130, 178, 187, 170, 182, 73, 38])): *** starting round RoundId { round_nums: [1], is_echo: false } ***
2025-04-28T10:58:06.830919Z DEBUG manul::session::tokio: PartyId(AccountId32([208, 139, 181, 149, 207, 194, 199, 205, 52, 230, 2, 226, 50, 174, 189, 144, 20, 77, 65, 104, 216, 189, 155, 61, 133, 190, 124, 33, 30, 145, 43, 24])): Sending a message to PartyId(AccountId32([76, 94, 254, 229, 66, 181, 20, 80, 20, 240, 21, 163, 237, 36, 116, 129, 134, 66, 26, 57, 48, 71, 122, 38, 221, 130, 178, 187, 170, 182, 73, 38]))
2025-04-28T10:58:06.830945Z DEBUG manul::session::tokio: PartyId(AccountId32([76, 94, 254, 229, 66, 181, 20, 80, 20, 240, 21, 163, 237, 36, 116, 129, 134, 66, 26, 57, 48, 71, 122, 38, 221, 130, 178, 187, 170, 182, 73, 38])): Sending a message to PartyId(AccountId32([208, 139, 181, 149, 207, 194, 199, 205, 52, 230, 2, 226, 50, 174, 189, 144, 20, 77, 65, 104, 216, 189, 155, 61, 133, 190, 124, 33, 30, 145, 43, 24]))

At this point it hangs.

[ameba23]

entropy-protocol tests now pass - still need to update entropy-tss

[ameba23]

Current state - most tests passing - but the entropy-tss jumpstart and reshare tests fail.

[ameba23]

We get rid of this error type and just use ProtocolExecutionErr (below) - as things are simpler now.

[ameba23]

Thankfully we don't need this anymore as we can now use manul::dev::run_sync which gives us the same thing.