fix(f02-mocks): match encoding format against f02's own vectors

[rvagg]

Also:

• one set of CBOR primitives
• addresses encode in both forms a contract can hold
• add tryStepWeightRecords over a shared encoder, and guard each field against the width f02 declares
• FVMRewardWire.t.sol mirrors actors/reward/tests/types_test.rs, asserting the bytes the mock received

[Copilot]

Pull request overview

This PR aligns the Solidity f02 (Reward actor) wire encoding/decoding with f02’s own DAG-CBOR serialization vectors, so the production encoder can be validated against canonical on-chain expectations instead of only matching the local mock.

Changes:

• Reworks FVMRewards to use shared CBOR primitives, shortest-form integer encoding, explicit width-guards, and adds tryStepWeightRecords.
• Updates the mock Reward actor and its tests to match the new parameter tuple shapes (single-field tuple wrappers, distribution encoding, cancel-pending null-id semantics).
• Adds FVMRewardWire.t.sol vector tests that assert the exact bytes dispatched to the mock, mirroring f02’s upstream vectors.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
src/lib/FVMRewards.sol	Replaces ad-hoc encoders with shared DAG-CBOR primitives, adds width guards and StepWeightRecords support.
test/mocks/FVMRewardActor.sol	Updates mock decoding/behavior to match updated wire format; adds param-recording and claim-return override hooks.
test/mocks/FVMRewardActor.t.sol	Refactors tests to use new library wrappers and new registration/distribution semantics.
test/mocks/FVMRewardWire.t.sol	New “golden vector” tests pinning exact bytes emitted by the production encoder.

Suppressed comments (2)

test/mocks/FVMRewardActor.sol:818

• _decodeAddress accepts any protocol-4 byte string of length 22 as an EVM address, but the comment specifies protocol-4 with the EAM namespace (0x0a). Without checking the namespace byte, non-EAM delegated addresses would be mis-decoded as valid EVM addresses.

        } else if (protocol == 4 && len == 22) {
            assembly ("memory-safe") {
                addr := shr(96, calldataload(add(pos, 2)))
            }
        }

test/mocks/FVMRewardActor.sol:787

• This doc comment still describes decoding an f410 address and/or CBOR null, but _decodeAddress no longer handles null and now supports both protocol-0 ID addresses and protocol-4 delegated addresses. Updating the comment avoids misleading future edits.

    /// @dev Decodes an f410 delegated address wrapped in a CBOR byte string (0x04, 0x0a, 20
    /// bytes) at absolute calldata position `pos`, or CBOR null (address(0)) for an IMPLICIT
    /// stream's absent writer. The address bytes are big-endian, so one `calldataload` shifted
    /// into place reads all 20 at once.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[rvagg]

@wjmelements anticipating one of your objections: I did replace your optimised, inlined cbor with a more compact form that could share across the implementations. forge snapshot suggests it's only a ~4k gas regression for most calls (EVM gas, but probably not a terrible proxy for scale in FEVM, and it's easily dwarfed by the call_actor either side of it anyway). IMO that's worth it for the clarity and maintainability, but it's up to you: you can unwind it if you want to - the tests in here assert against hex generated by f02 itself (actors/reward/tests/types_test.rs, and we'll mirror those in go-state-types too), so if they pass we agree with the chain.

[wjmelements]

nice

[wjmelements]

nice