Encode and estimate Rococo/Wococo/Kusama/Polkadot messages

[svyatonik]

closes #1319

The main thing here is EncodedOrDecodedCall, which has the same encoding as on-chain calls. But it has two variants - the call may be decoded (that's what we have been using previously). Or it may be encoded and only needs to be decoded when it needs to. In terms of relay subcommands, it means that if you're using pre-encoded calls in cli options and specify their dispatch weight, we'll never need to decode this call (this is the main problem from #1319 - we can't decode Rococo/Wococo/Kusama/Polkadot calls, because we have no access to the runtime from the relay).

Examples:

svyatonik@xwnotebook:~/dev/rococo-wococo.test$ ./bin/substrate-relay encode-message rococo-to-wococo call --sender 5CUXMS6Ja5E8HGa8qq2jumboFKjqpjA8mrdzyANhk7P2mELp --dispatch-weight 10000000000 raw 00010442
INFO bridge Created Message Payload: MessagePayload {
    spec_version: 9140,
    weight: 10000000000,
    origin: CallOrigin::SourceAccount(
        122a87a73f92704a4556131f8c2d14a32dc1d624fc523ba4e528bff3f9dfe55f (5CUXMS6J...),
    ),
    dispatch_fee_payment: DispatchFeePayment::AtSourceChain,
    call: 0x00010442,
}
INFO bridge Encoded Message Payload: 0xb423000000e40b540200000002122a87a73f92704a4556131f8c2d14a32dc1d624fc523ba4e528bff3f9dfe55f001000010442
0xb423000000e40b540200000002122a87a73f92704a4556131f8c2d14a32dc1d624fc523ba4e528bff3f9dfe55f001000010442
svyatonik@xwnotebook:~/dev/rococo-wococo.test$ ./bin/substrate-relay estimate-fee rococo-to-wococo call --sender 5CUXMS6Ja5E8HGa8qq2jumboFKjqpjA8mrdzyANhk7P2mELp --dispatch-weight 0 raw 00010442
INFO bridge Created Message Payload: MessagePayload {
    spec_version: 9140,
    weight: 0,
    origin: CallOrigin::SourceAccount(
        122a87a73f92704a4556131f8c2d14a32dc1d624fc523ba4e528bff3f9dfe55f (5CUXMS6J...),
    ),
    dispatch_fee_payment: DispatchFeePayment::AtSourceChain,
    call: 0x00010442,
}
INFO bridge Encoded Message Payload: 0xb4230000000000000000000002122a87a73f92704a4556131f8c2d14a32dc1d624fc523ba4e528bff3f9dfe55f001000010442
INFO bridge Fee: Balance(293479234556)
293479234556
svyatonik@xwnotebook:~/dev/rococo-wococo.test$ ./bin/substrate-relay send-message rococo-to-wococo --source-signer //Alice --dispatch-weight 100000000000 raw 00010442
INFO bridge Created Message Payload: MessagePayload {
    spec_version: 9140,
    weight: 100000000000,
    origin: CallOrigin::SourceAccount(
        d43593c715fdd31c61141abd04a99fd6822c8558854ccde39a5684e7a56da27d (5GrwvaEF...),
    ),
    dispatch_fee_payment: DispatchFeePayment::AtSourceChain,
    call: 0x00010442,
}
INFO bridge Encoded Message Payload: 0xb423000000e876481700000002d43593c715fdd31c61141abd04a99fd6822c8558854ccde39a5684e7a56da27d001000010442
INFO bridge Sending message to Wococo. Lane: [0, 0, 0, 0]. Size: 177. Dispatch weight: 100000000000. Fee: 1,173,439,257,663
INFO bridge The source account (d43593c715fdd31c61141abd04a99fd6822c8558854ccde39a5684e7a56da27d (5GrwvaEF...)) balance will be reduced by (at most) 1173439257663 (message fee) + 31605393717 (tx fee	) = 1205044651380 Rococo tokens
INFO bridge Signed Rococo Call: 0xc502bd028400d43593c715fdd31c61141abd04a99fd6822c8558854ccde39a5684e7a56da27d0180c4412df47caab679790c7992755661d891bc73ed75b632a46372ab23e2200fafec0946b26eaa091bbd04671261c06c15911e0f4e1a421d6943c66fc9c3ca850008002c0300000000b423000000e876481700000002d43593c715fdd31c61141abd04a99fd6822c8558854ccde39a5684e7a56da27d0010000104423f1c6e36110100000000000000000000

[svyatonik]

I'll merge this without approval for now - it needs to be tested on Rialto+Millau and we already have a bunch of unreviewed PRs. So maybe it'll be easier to review them in a polkadot-staging-update PR at once.

[acatangiu]

Do we forgo validation completely? Is this because we can't validate encoded call?

[svyatonik]

Yeah - this function is only called within runtime and we're using SignedExtensions just to get the encoded data (so we don't need any validation at all). More over, the default SignedExtension::validate actually does nothing: https://github.com/paritytech/substrate/blob/master/primitives/runtime/src/traits.rs#L862-L879

[acatangiu]

so this fn only returns EncodedOrDecodedCall::Encoded variant for Call::Raw, right?

[svyatonik]

That's correct. Why is it important for us - we can't maintain the whole copy of Polkadot/Kusama enum Call here, in substrate relay code. Instead, we allow to encode call externally and use this raw variant. We can't decode it here, because for most of calls, decode will fail (except for this remark). So yes - it only returns EncodedOrDecodedCall::Encoded for Call::Raw

[acatangiu]

I'm confused, for Millau the same trait fn encode_call() actually decodes the call when raw data is provided?

Why not use EncodedOrDecodedCall::Encoded(data.0.clone()) variant like in Kusama?

[svyatonik]

This can be changed. But:

1. imo earlier error = better error. Since Call of Rialto/Millau is the runtime call (not the mocked call as for the Polkadot/Kusama), decode only fail if the call is invalid. It is better to see decoding error here, than some strange error later (e.g. on the bridged chain during dispatch);
2. most of times (except when sending messages manually or using sending max-dispatch-weight messages) in Rialto<>Millau bridge, we'll be calling get_dispatch_info() on this call to get its dispatch weight. So it needs to be decoded anyways.

Feel free to insist, if you think it'll be better to change it to the Encoded here, though :)