feat!(seismic-tx): add auth_list to TxSeismic + bump deps

[samlaf]

Update seismic dependencies (alloy, revm, evm) and fix anything that needs updates in reth:

• Add authorization_list: Vec<SignedAuthorization> (EIP-7702) to TxSeismic, including Compact codec encoding/decoding and tests
• Add CodeOverrideNotPermitted variant to EthApiError to handle the new StateOverrideError variant from seismic-evm (Seismic disallows code overrides for privacy)
• Remove RngMode from all call sites — seismic-revm no longer exposes it on SeismicTransaction (rng mode is now determined internally)
• Delegate SeismicTypedTransaction::Seismic to TxEnv::from_recovered_tx now that seismic-evm provides the impl
• Collapse two raw-tx tests into one self-contained roundtrip test (hash was invalidated by the authorization_list field change)

Related PRs:

• feat(TxSeismic): add optional 7702 authorization_list seismic-alloy#82
• feat: Disallow code overrides seismic-evm#45
• Fix: veridise 801/825 refactor rng precompile seismic-revm#189

Note that this is a breaking change, since the tx's rlp encoding is changed, which breaks all of: wire format, db format (whats stored in tx trie), and signatures (signing over different rlp encoding).

[github-actions]

PR Review Summary

Reviewed commit: 7c4163f86a29a694bb35abafd077d5ad73d09417

Changes

Adds an authorization_list: Vec<SignedAuthorization> field to TxSeismic (EIP-7702 extension) by bumping the seismic-revm, seismic-alloy, and seismic-evm dependency revisions. The new field is wired through the compact codec, the FromRecoveredTx → TxEnv mapping, and all construction sites (tests, utils, fuzz harness). The PR acknowledges this is a breaking change to wire format, db format, and signatures.

---

Blocking Issues

1. alloy_compat.rs silently drops authorization_list from RPC-deserialized transactions
File: crates/seismic/primitives/src/alloy_compat.rs:96

When building a TxSeismic from an AnyRpcTransaction (e.g., from eth_getTransactionByHash), the authorization_list is unconditionally hardcoded to vec![]:

authorization_list: vec![],

If the Seismic node serializes a non-empty auth list in its RPC response (which the upstream seismic-alloy PR #82 presumably adds), this conversion will silently drop it. The downstream effect is a different sighash, causing incorrect sender recovery and breaking any code path that re-signs or re-verifies the deserialized transaction.

The fields map is already in scope for all other Seismic-specific fields. The fix is to parse authorization_list from the RPC response fields (or fall back to vec![]] if the field is absent/null), consistent with how other fields are extracted. At minimum, a TODO comment should note this is intentionally incomplete and enumerate the expected breakage.

---

2. Missing compact codec test for non-empty authorization_list
File: crates/storage/codecs/src/alloy/transaction/seismic.rs:434

The only compact roundtrip test constructs a TxSeismic with authorization_list: vec![]. The new field is placed in the middle of the compact struct (between seismic_elements and input), which is the most sensitive position for a field insertion because any decoding regression would silently corrupt input for old-format records. A test encoding and decoding a TxSeismic with at least one real SignedAuthorization is needed to verify:

• the SignedAuthorization Compact impl round-trips correctly in this context
• the input field following it is correctly positioned after decoding

---

Suggestions

1. TODO comments in Cargo.toml reference unmerged upstream PRs
The three new TODO(samlaf) comments note that the pinned revisions point to branch tips on seismic-revm PR #209, seismic-alloy PR #82, and seismic-evm PR #40. This means the code pulled in by this PR has not yet been reviewed/merged through the standard upstream gate. Consider:

• describing the surface-level changes in those PRs in the PR description (the diff does not show what changed at those revisions)
• tracking the cleanup as a follow-up issue rather than an in-file TODO, which is easy to forget

2. authorization_list field placement in compact codec struct
The field is inserted between seismic_elements and input rather than at the end of the struct. EIP-7702 convention (and all other reth codec structs for type-4 txs) places authorization_list last. The current placement is a valid choice for a new schema, but deviates from the established pattern and makes schema audits harder. Since this is already a breaking change to the db format, it may be worth moving the field to the end of the struct for consistency before the format is locked.

---

Positive Notes

• The FromRecoveredTx fix at crates/seismic/primitives/src/transaction/signed.rs:311 correctly maps each SignedAuthorization via Either::Left rather than defaulting, so the EVM TxEnv will actually carry the auth list through to execution.
• All affected construction sites (test utils, fuzz harness, RPC test helpers) are consistently updated with authorization_list: vec![].
• The compact codec struct properly delegates Compact derivation (#[derive(Compact)] on TxSeismic) and the Compact for AlloyTxSeismic manual impl is correctly updated on both the encode and decode sides.
• Dependency revisions are pinned by exact commit SHA, so the build is reproducible even while the upstream PRs are pending merge.

To request a re-review after pushing fixes, comment @claude on this PR.

[samlaf]

@claude fixed your 2 blocking issues:

1. merged refactor(seismic-primitive): delete unused alloy-compat code/feature #333 into seismic and then rebased this PR on top of that. That code was deadcode anyways so deleted it.
2. augmented the test to have a non-empty authorization-list in 35d0064

[github-actions]

Re-Review Summary

Reviewed commit: 61a01761a0a5b97cf94fdb84fafd36a1c0a46519
Previous review commit: 7c4163f86a29a694bb35abafd077d5ad73d09417

Previously Flagged — Now Resolved

1. alloy_compat.rs silently drops authorization_list — Resolved by merging PR refactor(seismic-primitive): delete unused alloy-compat code/feature #333, which deleted the entire alloy_compat dead-code module. The conversion path that was losing the auth list no longer exists.

2. Missing compact codec test for non-empty authorization_list — Resolved in commit 35d0064. The roundtrip test in crates/storage/codecs/src/alloy/transaction/seismic.rs now constructs a real SignedAuthorization (with a chain ID, address, nonce, and explicit r/s/v signature) and asserts it survives encode→decode intact. Field-ordering safety for input is also now exercised.

Previously Flagged — Still Unresolved

1. authorization_list field placement in compact struct (Suggestion) — The field remains between seismic_elements and input rather than at the end of the struct. This was a non-blocking suggestion; no action required before merge.

2. TODO comments referencing unmerged upstream PRs (Suggestion) — Still present in Cargo.toml. No action required before merge.

New Issues in Fix Commits

None found.

New Suggestions

1. test_recover_raw_tx → test_generate_and_recover_raw_tx refactor is sound but loses a cross-format sentinel. The old hardcoded raw-bytes test in crates/seismic/rpc/src/eth/transaction.rs caught accidental format regressions by verifying a known-good byte string from outside the current codebase. The replacement test only asserts self-consistency (generate then recover). Consider keeping one hardcoded expected hash (analogous to the one preserved in utils.rs) so a future accidental wire-format change will be caught at test time rather than silently passing.

[samlaf]

rebased on top of #336

[samlaf]

rebased on top of #340

[github-actions]

Updates seismic-revm, seismic-evm, and seismic-alloy dependencies and removes RngMode from transaction abstraction while adding EIP-7702 authorization list support.

Phase 1

• crates/storage/codecs/src/alloy/transaction/seismic.rs:64 — Adding authorization_list field to TxSeismic struct changes the compact encoding format. Since there's no backward compatibility handling visible for the old format without this field, existing MDBX data may fail to decode or deserialize with incorrect field mappings.

Phase 2

• crates/seismic/rpc/src/eth/utils.rs:702 — Test expectation changed from a9c1c87a4fa27002f9487ade27b5eb77ab3c82b284bc384609572f1eb8e171dc to 0xe82f9ce621da07a8ae10d330383275402ff9430dbe4b8b10cd0038ede3ef3718 due to authorization_list field addition. This suggests the transaction hash calculation has changed, which could affect transaction replay or verification of historical transactions.
• crates/seismic/rpc/src/eth/transaction.rs:662 — Test was marked #[ignore] and renamed from test_generate_raw_tx to test_generate_and_recover_raw_tx, removing the hardcoded transaction bytes check. This removes validation against a known good transaction encoding.

LGTM - The dependency updates clean up the RngMode abstraction properly across the codebase. The new code override rejection provides good security for Seismic's privacy model with comprehensive test coverage. The authorization_list addition follows EIP-7702 standards and includes proper roundtrip testing.

[HenryMBaldwin]

This overlaps with #367

I think this is fine, as the way we merged seismic-evm sort of makes this necessary. We can close that pr, but we need to just make sure we get everything from it. I think the only thing we're missing is the e2e tests here

If you add those and we can merge this pr and close 367.

[HenryMBaldwin]

I went ahead and added the tests in 5dd8b19. Assuming they pass in ci I think we're gtg.

[samlaf]

CI is stuck. closing and reopening PR.