Sync deposit ZKHALO2VERIFYWITHVK fixtures for chainId public input (#280)

* Sync deposit ZKHALO2VERIFYWITHVK fixtures for chainId public input

The deposit circuit gained a chainId public input at index 4 (bridge
branch pruvendo/pause-chainid-pi-multil2, commit c0da8a3), taking the
PI count from 11 to 12. Because the ZKHALO2VERIFYWITHVK opcode is
schema-agnostic (decode_instances_strict consumes any N × 32 B LE Fr
payload, VK carries num_instance_columns), no executor code changes
are needed — only test fixtures and PI-count metadata.

- Resync tvm_vm/halo2_test_data/deposit_10proofs/ from
  deposit-prover/fixtures/deposit_10proofs (Hermez SRS, k=18, Sepolia
  chain-id 11155111; VkBlob sha256
  006cca5ddd457065bef8469fc10dd231ad8fd13c895877c9deff5c51191dec05).
- Update PI-layout comments, log strings, and README to reflect the
  12-PI layout with chainId at index 4.
- Add a length-assertion regression guard in
  test_zkhalo2_with_vk_deposit_10_real_proofs to catch future fixture
  desyncs (expected 384 B per public_inputs.bin).
- Remove orphaned deposit_rlc_{vk_blob,proof,public_inputs,bundle}.bin
  singletons — the round_trip_deposit_rlc_* tests already load from
  deposit_10proofs/proof_00/*, so nothing referenced them.
- Add a design-doc footnote (§4) recording that this 11 → 12 PI
  transition was absorbed with zero opcode changes, empirically
  validating the Q-WIRE-3 / Q-WIRE-4 schema-agnostic decisions.

Verified: cargo test -r --lib -p tvm_vm -- halo2 → 41 passed, 0 failed
(3 pre-existing #[ignore]s on fallback Circuit 1B unrelated to this
change).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* tvm_vm: refresh deposit_10proofs fixtures against current bridge circuit_v2

Regenerated VkBlob + 10 proof triples against
bridge/deposit-prover/src/circuit_v2.rs, keyed on Hermez SRS
(s_g2 = 928fafb3…) at k=18. Sync driven by
scripts/sync_deposit_opcode_fixtures_to_tvm_sdk.sh.

Verified cryptographically: keygen-only regen from the current
circuit yielded a byte-identical VkBlob to the checked-in fixture
(sha256 9dacd998af5fd03af8097cb80a571df098c925bba235af61d920cc808360fae3),
proving the VkBlob is in sync with circuit_v2 and Hermez-keyed.

All 41 relevant halo2 tests pass, including
test_zkhalo2_with_vk_deposit_10_real_proofs.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Author: alinaT95

docs/zkhalo2verifywithvk_design.md

@@ -89,6 +89,18 @@ All length prefixes are `u32` LE. Producer-side cap: 16 MiB
 (`zk_halo2_with_vk_bundle::MAX_BUNDLE_BYTES`); the consumer fails
 structurally above that cap as a DoS guard.
 
+> **Schema-agnosticism, validated empirically (2026-07-29).** The deposit
+> circuit added a `chainId` public input (11 → 12 PIs, bridge commit
+> `c0da8a3`). Adjusting to it required **zero** changes to
+> `zk_halo2_with_vk.rs` / `zk_halo2_with_vk_bundle.rs`; only the on-disk
+> fixtures under `tvm_vm/halo2_test_data/deposit_10proofs/` had to be
+> resynced. `decode_instances_strict` chunks `instances_bytes` into
+> 32-byte LE Fr elements without any hardcoded count, and the VK itself
+> carries `num_instance_columns` / per-column lengths — mismatches
+> surface as cryptographic `false` (see
+> `instance_count_mismatch_rejected_as_false`), not `FatalError`. This
+> is exactly the Q-WIRE-3 / Q-WIRE-4 outcome we designed for.
+
 ### Safety of VK deserialisation
 
 The consumer deserialises the VK using

tvm_vm/doc/HERMEZ_KZG_MIGRATION_FOR_SERGEY.md

@@ -84,11 +84,15 @@ and need to be rebuilt against `kzg_bn254_20.srs` (Hermez):
 
 **Deposit circuit fixtures (your circuit, your fixtures):**
 * `deposit_10proofs/*` — the 10-proof deposit-event fixture pack
-  (`EthCircuitImpl<Fr, DepositEventCircuitV2>`, 11 public inputs).
-* `deposit_rlc_*.bin` — any `deposit_rlc_proof.bin`,
-  `deposit_rlc_instances.bin`, `deposit_rlc_vk.bin`, and the packed
-  `deposit_vk_blob.bin` (the `VkBlob`-format wrapper carrying the VK
-  plus the RLC `EthCircuitParams` JSON).
+  (`EthCircuitImpl<Fr, DepositEventCircuitV2>`, 12 public inputs since
+  2026-07-22 bridge commit `c0da8a3` — `chainId` inserted at index 4).
+* ~~`deposit_rlc_*.bin`~~ — the singleton `deposit_rlc_proof.bin`,
+  `deposit_rlc_public_inputs.bin`, `deposit_rlc_vk_blob.bin`,
+  `deposit_rlc_bundle.bin` files were **removed on 2026-07-29**
+  (branch `feature/deposit_circuit_chain_id_pi`). They were unreferenced
+  by any test — the `round_trip_deposit_rlc_*` tests in
+  `test_halo2_with_vk.rs` actually load from `deposit_10proofs/proof_00/*`,
+  which supplies the same VK+proof+PI triple in the fresh 12-PI layout.
 
 **Fallback (Circuit 1B) fixtures (my circuit, but *your* tvm-sdk
 fixtures + tests):**

tvm_vm/halo2_test_data/deposit_10proofs/README.md

@@ -4,18 +4,38 @@ These files exercise `ZKHALO2VERIFYWITHVK` in `tvm_vm` tests. They are the
 **three-operand ABI** only:
 
 ```text
-vk_cell              ← deposit_vk_blob.bin (VkBlob v2 RLC, 11 public inputs)
-public_inputs_cell   ← proof_NN/public_inputs.bin (11 × 32 B LE Fr)
+vk_cell              ← deposit_vk_blob.bin (VkBlob v2 RLC, 12 public inputs)
+public_inputs_cell   ← proof_NN/public_inputs.bin (12 × 32 B LE Fr = 384 B)
 proof_cell           ← proof_NN/proof.bin (Blake2b SHPLONK)
 ```
 
 ## KZG ceremony
 
 Keyed on the **Hermez Perpetual Powers of Tau** SRS (`data/kzg_params_18.srs`,
 `s_g2 = 928fafb3…`), matching `KZG_S_G2_BYTES` on
-`feature/hermez-kzg-resurrection`. Regenerated 2026-07-13 via
-`deposit-prover` `export_deposit_proof_set` (VkBlob sha256
-`304c1c4ed1e4cf09a00fb1d83a0ae2ba42db2afead035ae089a4faa85346251a`).
+`feature/hermez-kzg-resurrection`. Regenerated 2026-08-07 via
+`deposit-prover` `export_deposit_proof_set` after the circuit gained a
+`chainId` public input at index 4 (bridge branch
+`pruvendo/pause-chainid-pi-multil2`, commit `c0da8a3`). VkBlob sha256
+`9dacd998af5fd03af8097cb80a571df098c925bba235af61d920cc808360fae3` —
+update this line whenever the fixtures are resynced.
+
+Public-input layout (12 × 32 B LE Fr, in order):
+
+```
+0  depositId
+1  sender
+2  amount
+3  contractAddress
+4  chainId              ← added 2026-07-22 (bound via EIP-1559 tx MPT proof)
+5  dappIdHigh
+6  dappIdLow
+7  anAccountHigh
+8  anAccountLow
+9  blockHashHigh
+10 blockHashLow
+11 promiseCommit        (auto-appended by axiom-eth EthCircuitImpl)
+```
 
 ## What does **not** belong here
 
@@ -36,7 +56,8 @@ From `acki-nacki-bridge` (sibling repo):
 cd deposit-prover
 cargo run --release --example export_deposit_proof_set -- \
   --set-dir fixtures/deposit_10proofs --count 10 \
-  --degree 18 --max-data-byte-len 256 --max-log-num 20
+  --degree 18 --max-data-byte-len 256 --max-log-num 20 \
+  --chain-id 11155111    # Sepolia; select source chain for the fetch layer
 ../scripts/sync_deposit_opcode_fixtures_to_tvm_sdk.sh
 ```