fwk upgrade to 1.34.0 to dev

[dhaval-supraoracles]

Below are the PR list that have been merged into this one.

• PR Attempt to address merge conflicts related to automation feature #291
• PR Issue/fix building issue on upgrade #315

[isaacdoidge]

We need to resolve https://hexens.io/research/aptos-hijack-bug before merging this PR. Investigation is ongoing.

https://github.com/Entropy-Foundation/supra-aptos-core-private/issues/9

[isaacdoidge]

Heads-up for this upgrade: the v1.34.0-era code is affected by the Hexens type-tag-cache partial-flush bug (https://hexens.io/research/aptos-hijack-bug — check_ready() flushing the struct-name-index map without the type-tag cache, enabling resource type confusion once the SNI exceeds max_struct_name_index_map_num_entries).

Good news: on #367's head (aeb003df) the fixed code is present — the over-size branch of code_cache_global_manager.rs::check_ready() calls the combined flush_struct_name_and_tag_caches(). Before this merges, please make sure the final merged tree keeps that combined flush (and that no path flushes struct_name_index_map alone). Verification checklist is tracked in Entropy-Foundation/smr-moonshot#3077.

[isaacdoidge]

Review of the v1.34.0 content-port. Since the raw 10.8k-file diff is mostly upstream-verbatim, this was reviewed as the Supra delta vs clean upstream (git diff aptos-node-v1.34.0-rc <branch>, ~1,400 files), plus the final merged tree at feat/merge_with_dev (aeb003df). Line references below are on aeb003df since that's the tree that will ship.

⚠️ Do not merge this PR without #367. This branch's aptos_features.rs has Supra feature flags 97–101 (SUPRA_BLS_KEYS, SUPRA_BCFT_CERTIFICATES, SUPRA_DKG, SUPRA_TRANSACTIONS_INCLUSION_PROOFS, SUPRA_AUTOMATION_V2_1) overwritten by upstream flags; only the #367 merge restores them. Recommend landing them together (or merging #367's branch into this one first).

Findings on the port itself

H1 — TransactionShufflerType variant 3 repurposed breaks decoding of persisted execution configs (types/src/on_chain_config/execution_config.rs:200–210). On dev, BCS index 3 is Fairness { u32 × 3 } (12 payload bytes) and OnChainExecutionConfig::default_for_genesis() uses it — every existing Supra chain has that shape persisted. On this branch index 3 is the unit variant DeprecatedFairness (upstream's deprecation), so decoding the persisted config leaves 12 trailing bytes → decode fails → fetch_config returns None → the block executor silently falls back to default_if_missing, i.e. nodes shuffle transactions with a different algorithm than the persisted config specifies. Mixed-version validator sets and historical replay diverge. Per the pre-production config policy a break may be acceptable, but it must be explicit: either migrate/rewrite the on-chain execution config as part of the upgrade rollout, or keep a decode-compatible Fairness variant that maps to the new behavior — and document the decision. Please also confirm how replay of pre-upgrade blocks treats this.

M1 — nonce_validation is never initialized at genesis (aptos-move/framework/supra-framework/sources/genesis.move). Upstream v1.34 calls nonce_validation::initialize(...) in genesis (upstream genesis.move:134); the Supra port imports the module (line 43) but never calls it — the unused import suggests the call was dropped accidentally. The prologue's check_for_replay_protection_orderless_txn asserts exists<NonceHistory> and aborts otherwise. Since ORDERLESS_TRANSACTIONS is enabled in default_features() on this branch, a fresh chain accepts orderless transactions at the flag level but aborts every one of them in the prologue until governance manually runs initialize_nonce_table. Fix: restore the genesis call (or explicitly disable ORDERLESS_TRANSACTIONS by default and document why). Note this interacts with the features.move numbering defect raised on #367.

M2 — coin.move retains pre-1.34 MigrationFlag-gated FA routing (can_receive_paired_fungible_asset ~line 1043, tail of maybe_convert_to_fungible_store ~745). Upstream v1.34 routes deposits to the paired FA store whenever it exists; the port still requires MigrationFlag (or the default-to-FA feature) and still writes MigrationFlag. For an account whose primary FA store was created by direct FA interaction (no flag), an incoming coin deposit lands in the legacy CoinStore here vs the FA store upstream — aggregate balances match but physical stores and events diverge from v1.34 behavior. If this is an intentional retention of Supra's migration semantics, please confirm and document; otherwise adopt the upstream form.

L1 — ValidatorTransaction::DKGResult arm returns Err(UNREACHABLE) that would fail block execution (aptos-move/aptos-vm/src/validator_txns/mod.rs:32–37). Unreachable in forward operation (Supra consensus only emits DKG), but if any historical block carried a DKGResult validator txn, replay of it now errors instead of executing. Please confirm no such txn exists in any chain history we intend to replay.

L2 — GenesisEvmContract.deploy_address changed Option<String> → String (types/src/on_chain_config/evm_genesis_config.rs:66): BCS layout change for a persisted on-chain config. Impact bounded to chains that persisted the old shape (EVM testnets, if any) — flagging for awareness given the EVM-activation timeline.

L3 — db.insert(...).unwrap() inside the eth_trie native (aptos-move/framework/src/natives/cryptography/eth_trie.rs:66): effectively infallible with MemoryDB, but natives should map to SafeNativeError, not unwrap. Related: the reachable-panic fix for this native (ce3cebc85a, PRs #5/#6) exists on dev but is missing from the merge head — raised on #367.

Verified clean (for the record)

• Automation processors + gas.rs + validator_txns/dkg.rs: faithful loader-V2 ports; gas metering matches the user-txn path; module-publish invalidation correctly delegated to the unified resolve_pending_code_publish_and_finish_user_session (no manual cache invalidation gaps found beyond the known Hexens item, whose fix is present on the merge head).
• Supra natives (rlp, eth_trie, vector_utils, bls12381_*, class_groups): all charge base + per-unit gas before work; no zero-charge path found.
• DKG/automation/on-chain-config types: no BCS layout changes vs dev besides the two flagged above; FeatureFlag renames keep explicit discriminants; OnChainConsensusConfig::V4 appended, not reordered.
• Supra framework Move port: permissioned_signer/account_abstraction/account byte-identical to upstream v1.34 after namespace normalization (fresh base, not a stale copy); reconfiguration_with_dkg preserves the features-before-stake ordering invariant and the SUPRA_DKG branch; governance hash-gated resolution intact; no spec files dropped; no stray aptos_framework:: references.

Separately, a pre-existing bug (also on dev, not introduced here) was found in stake.move::assert_consensus_pubkey_unique_in_validator_list: it compares raw serialized key blobs, which becomes vacuous after set_dkg_output_keys writes threshold-augmented keys back into consensus_pubkey. Filing as its own issue since it's not this PR's regression.