feat: miner aggregation support #131
Conversation
Fixes aggregation to reject proofs that share a block hash but disagree on block number, preventing mixed-block batches from being accepted. Adds a structured parser for layer-1 aggregated public inputs, documents the layout with version constants, and exposes the new parsing path through the verifier. Improves test coverage around malformed inputs and block-number mismatches to make the aggregation boundary safer to evolve.
The medium risk finding isn't valid since the verifier is not yet connected to the chain for layer-1. |
*Helps produce reproducible test fixtures from existing proof inputs and generated binaries, so integration tests can rely on consistent aggregate outputs. *Accepts command-line inputs, validates the expected proof count, verifies the generated aggregate, and exports both proof artifacts and a JSON bundle of expected public data for downstream test use.
*Writes fixture outputs only after aggregation succeeds and adds a binary aggregate artifact, which avoids leaving partial output behind and makes the proof easier to consume in downstream tooling. *Adds regression and validation coverage for aggregated public inputs so ABI layout drift and malformed payload shapes fail fast with clearer guarantees.
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 20317a9. Configure here.
|
|
||
| let err = Layer1AggregatedPublicCircuitInputs::try_from_u64_slice(&pis).unwrap_err(); | ||
| assert!(err.to_string().contains("inconsistent shape")); | ||
| } |
There was a problem hiding this comment.
Duplicate test masks unreachable cursor-mismatch check
Low Severity
layer1_public_inputs_reject_cursor_shape_mismatch is identical to layer1_public_inputs_reject_too_long_vector — same input construction, same assertion. The name suggests it was meant to test the "cursor mismatch" check at line 624, but that check is actually unreachable dead code: after the remaining % 4 == 0 and nullifier_count * 2 == total_exit_slots validations pass, cursor is guaranteed to equal pis.len(). The test instead hits the earlier "inconsistent shape" error, duplicating the other test.
Additional Locations (2)
Reviewed by Cursor Bugbot for commit 20317a9. Configure here.
| pub account_data: Vec<PublicInputsByAccount>, | ||
| /// Forwarded nullifiers from all child layer-0 proofs. | ||
| pub nullifiers: Vec<BytesDigest>, | ||
| /// Reserved for the future constrained bundle-root public input. |
There was a problem hiding this comment.
Do we need these "reserved" fields?
illuzen
left a comment
illuzen
left a comment
There was a problem hiding this comment.
Generally fine, looks like some unused fields in the struct, bugbot found a duplicate test
2 participants
Note
Medium Risk
Modifies layer-1 aggregation circuit constraints and public-input parsing, which can invalidate existing proofs/fixtures if the layout or invariants are wrong. Changes are well-scoped with added negative/ABI regression tests, but affect proof compatibility.
Overview
Adds first-class support for layer-1 aggregated proof public inputs: introduces
Layer1AggregatedPublicCircuitInputswith an explicit header/slot/nullifier layout, robust shape validation, and comprehensive parser regression/negative tests.Tightens the layer-1 aggregation circuit by enforcing block number consistency (in addition to block hash) across all child layer-0 proofs, plus a new negative test for mismatched block numbers.
Adds a
generate_l1_fixtureexample that takes a full batch of L0 proof hex files, produces an L1 aggregate proof (.hex+.bin), and writes anexpected_bundle.jsonderived from parsed L1 public inputs; verifier crate now re-exports layout versions and providesparse_layer1_aggregated_public_inputs.Reviewed by Cursor Bugbot for commit 22c6fd0. Bugbot is set up for automated code reviews on this repo. Configure here.