Drops IBS.SetIncarnation / IBS.GetIncarnation and the prevInc/PrevIncarnation
+ IncarnationPath versionRead/versionWritten plumbing in IBS.Selfdestruct,
IBS.CreateAccount, IBS.MakeWriteSet, IBS.ApplyVersionedWrites,
IBS.refreshVersionedAccount and IBS.HasStorage. Storage clearing on
SELFDESTRUCT + recreate now flows through an explicit
stateObject.recreatedFromDestructed flag → DeleteAccount, instead of the
implicit `original.Incarnation > account.Incarnation` predicate that
required the genesis Inc=1 fixup to fire. Continues #12440.

Inspired by zilkworm/zilk_core/core/state/intra_block_state.cpp.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…allel exec

In parallel mode, CreateAccount on an address that was self-destructed by
an earlier tx in the same block synthesises a non-nil "previous" stateObject
with selfdestructed=true (intra_block_state.go around the readAccount!=nil
branch). createObject then journals resetObjectChange, whose dirtied()
returns false — so the address is never added to journal.dirties.
SoftFinalise → MakeWriteSet then sees isDirty=false, skips
Writer.UpdateAccountData, and the apply-time writeset is missing
NoncePath/IncarnationPath for the resurrected empty account.

Symptom: EEST test_double_kill[fork_Frontier|fork_Homestead] failed with a
block-2 state-root mismatch under ERIGON_EXEC3_PARALLEL=true. Serial mode
passed because each tx runs in a fresh IBS against the post-apply DB, so
the SD'd account is gone and previous==nil → createObjectChange (which
dirties).

Fix: treat previous.selfdestructed as equivalent to previous==nil for the
journal-entry choice — emit createObjectChange so the address is dirtied.
On revert createObjectChange clears stateObjects[addr]; the next
getStateObject re-discovers the SD via the versionMap marker path, so the
revert-then-read invariant still holds.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…e only

4ab0771 unconditionally treated previous.selfdestructed as previous==nil
in createObject's journal-entry choice. That broke TestDeleteCreateRevert
in the serial accumulating-IBS path: a CALL→CreateAccount→REVERT inside the
same block reverts createObjectChange, which deletes stateObjects[addr]
entirely — losing the destructed previous stateObject and letting the next
read of the address resurrect it from on-disk state, mid-block.

The parallel path is unaffected: the SD signal lives in the versionMap, so
the next getStateObject re-discovers it via the marker placeholder.

Fix: only switch to createObjectChange when sdb.versionMap != nil. Serial
mode keeps resetObjectChange so revert restores ch.prev.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…getStateObject

After stateReader.ReadAccountData returns nil, getStateObject probes
versionedRead(AddressPath) to catch addresses CreateAccount'd by a prior
same-block tx. The follow-up versionedRead(SelfDestructPath) branch that
constructed a synthetic placeholder there was dead: versionedRead's own
SelfDestructPath short-circuit returns the zero value (false for bool)
when SD=true is present, so the destructed flag in that branch can only
ever be false. With destructed=false the branch built a placeholder
stateObject with all-false flags — never useful — and on err returned
(nil, err) with that misleading stateObject left in stateObjects.

The cross-tx SD-via-versionMap case is handled by the SelfDestructPath
probe further down (after refreshVersionedAccount), which uses the same
synthetic-placeholder pattern with the correct flags. Removing the
unreachable branch is a no-op behaviourally — 8 net LOC saved, the path
clarified.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The IncarnationPath enum value never carried information beyond the
account's nonce / SelfDestructPath / on-disk Incarnation (#12440): IBS
no longer tracks incarnation, the on-disk Incarnation field is read
from sd.mem at apply time when needed, and SelfDestructPath drives all
SD-and-recreate semantics in normalizeWriteSet / applyVersionedWrites /
calc_state.

What this drops:
- IncarnationPath enum value, String case, valuesEqual case.
- All emit sites: versionedWriteCollector.UpdateAccountData,
  LightCollector.UpdateAccountData, SetAccountBalanceOrDelete,
  versionedStateReader.applyVersionedUpdates.
- applyVersionedWrites' addrState.incarnation field and
  flushVersionedWritesToAccumulator's pendingAccount.incarnation.
- versionedRead's account-field-path lists (4 sites) shrink from 4 to 3.
- normalizeWriteSet's sdSet drop / filter switch / addrFields tracking /
  missing-field fill loop / empty-removal cleanup / resolveStorageWrites
  all shrink from 4 to 3 paths.
- calcState.calcAccountState.Incarnation field, the SD-zeroes-Incarnation
  line, and FlushToUpdates' "Deleted && Incarnation > 0" branch
  (unreachable from real writesets — ensureAccount never copied
  Incarnation from on-disk in the first place).
- The matching unit tests:
  TestApplyWrites_IncarnationPath (whole purpose was the dead branch),
  TestFlushToUpdates_DeletedWithIncarnation_EmitsZeroAccountUpdate
  (covered the same dead branch).
- Stale comments mentioning IncarnationPath = preInc emits in
  IBS.Selfdestruct and similar.

Net delta: 11 files, +100 / -298 (-198 LOC).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Companion to the earlier IBS/IncarnationPath removal — the per-account
incarnation counter is now gone from everywhere it used to participate
in storage-key composition.

DB migration
- New migration `drop_incarnation_from_storage` (db/migrations/) clears
  the five pre-E3 tables whose storage keys embedded an 8-byte
  incarnation between [addr] and [storage_key] (PlainState's dupsort
  storage rows; HashedStorageDeprecated; AccountChangeSetDeprecated;
  StorageChangeSetDeprecated; HashedAccountsDeprecated). None of these
  tables are read by E3 execution — kv.StorageDomain is the active store
  and is already incarnation-free.
- The constants are moved into ChaindataDeprecatedTables so the empty
  buckets are dropped on the next exclusive open; the live
  ChaindataTables list no longer recreates them.

Helper signatures
- dbutils.GenerateCompositeStorageKey(addr, seckey) loses the
  incarnation arg and now returns 64 bytes (addr+key).
- dbutils.ParseCompositeStorageKey returns (addr, key); ParseStoragePrefix
  is gone.
- common.StorageKey shrinks from 72 to 64 bytes (StorageKeyLen = 2*length.Hash).

Legacy MPT path (execution/commitment/trie)
- The 8-byte incarnation that used to sit between the account hash and
  the storage subtree in the compact-key form is dropped from
  NewProofRetainer's compactEncoded, from stream.go's
  AccountNode-with-Storage descent (twice), and from
  findSubTriesToLoad's nibble-path construction. Storage subtree is
  now rooted directly under the account hash.
- HashWithModifications callers pass storagePrefixLen=32 (was 40).
- Tests TestCreateLoadingPrefixes / TestHashWithModifications*
  updated for the shorter dbPrefixes / fixedbits / hook nibbles.

Caller cleanups
- t8ntool's CalculateStateRoot: removed the PlainState→HashedStorage
  hashing loop entirely (PlainState empty on E3, ComputeCommitment
  works off kv.CommitmentDomain).
- kvcache.AssertCheckValues: stubbed to a no-op (was comparing cache
  rows against kv.PlainState, which no longer exists in the schema).
- cmd/integration/commands/refetence_db.go and cmd/pics/state.go drop
  the deprecated tables from their whitelists.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Two CI failures introduced by 70c6991 (the storage-key incarnation
removal):

1. TestManagedTx (db/kv/mdbx/kv_abstract_test.go) was picking the first
   two entries of kv.ChaindataTables alphabetically and relying on them
   having matching DupSort flags. That invariant held when the list
   started with the deprecated AccountChangeSet / AccountHistory pair
   (both Flags=0) but broke once the deprecated tables moved to
   ChaindataDeprecatedTables — AccountHistory (Flags=0) and
   AccountHistoryKeys (Flags=DupSort) now lead the list, so the
   identical-operation-different-result asymmetry surfaced as a long
   chain of cursor-Equal failures. Switch the test to explicit
   TblAccountVals / TblStorageVals so it doesn't depend on the sort
   order.

2. TestEthCallToPrunedBlock (rpc/jsonrpc/eth_call_test.go) called
   PruneTableDupSort on kv.StorageChangeSetDeprecated, which the
   schema no longer materialises in ChaindataTables — cursor open
   panics with MDBX_BAD_DBI. Drop the prune of that now-empty table;
   the active prune on TblAccountVals exercises the same code path.

Also restore the DupSort flag entries for PlainState /
HashedStorageDeprecated / AccountChangeSetDeprecated /
StorageChangeSetDeprecated in ChaindataTablesCfg so the
drop_incarnation_from_storage migration can open the buckets with the
correct flags on a pre-E3 database before clearing them.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

E2AccountsHistory and E2StorageHistory are pre-E3 history-index tables
(roaring bitmaps mapping addr/storage-key → list of blocks that touched
it). The E3 execution path reads history from kv.AccountsDomain /
kv.StorageDomain inverted indices, not these — they were dead in
production and only kept alive by two debug callers.

Schema
- Moved E2AccountsHistory / E2StorageHistory from ChaindataTables to
  ChaindataDeprecatedTables so the empty buckets are dropped on the
  next exclusive open and the live schema no longer recreates them.

Migration
- The pre-existing drop_incarnation_from_storage migration is renamed
  to drop_legacy_e2_tables — the scope is now "clear dead pre-E3
  tables" rather than strictly incarnation-bearing ones — and the two
  history tables are added to its ClearTable loop. The migration is
  still idempotent (ClearTable is a no-op on already-empty tables).

Callers
- cmd/integration/commands/refetence_db.go: dropped the two entries
  from stateBuckets (the debug-compare whitelist) — TxLookup is all
  that remains there.
- cmd/rpctest/rpctest/account_range_verify.go: used kv.E2AccountsHistory
  purely as the bucket name for JSON-dumping eth_accountRange results
  into a fresh temp MDBX env for geth/erigon comparison — nothing to
  do with the actual E2 history index semantics. Switched the tool to
  a local "AccountDump" bucket via mdbx.WithTableCfg so the temp DB
  is fully decoupled from the chain schema.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…clear

Address Copilot review on #21280:

- drop_legacy_e2_tables migration was calling tx.ClearTable, which only
  empties the bucket (MDBX Drop with del=false) and leaves the bucket
  metadata in place. The doc comments on both the migration and the
  matching ChaindataDeprecatedTables / ChaindataTablesCfg entries said
  "drop" or "clears + drops" — overstated relative to the implementation.
  Switch to tx.DropTable: it requires IsDeprecated=true (which these
  tables now have), reclaims the bucket metadata, and is idempotent —
  dropEvenIfBucketIsNotDeprecated handles the NonExistingDBI case and
  returns nil if a re-open with mdbx.IsNotFound fails.
- eth_call_test.go's doPrune was leaking the time.NewTicker(20s) it
  created for the prune progress log; nothing called Stop() so the
  underlying ticker goroutine survived each test run. Add the missing
  defer.
- Tighten the comment in doPrune to say StorageChangeSetDeprecated is
  no longer "part of the active schema" rather than "materialised in
  the active schema".

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The AccountChange message (used by both the in-process Accumulator and
the gRPC StateChanges stream) carried a `uint64 incarnation` field that
existed solely to tag pre-E3 storage rows whose keys baked the
per-account incarnation counter. After the incarnation removal from
storage keys + commitment path (#21256), no live consumer needs it:

- the only production reader was txpool.processNewBlock, which used
  `change.Incarnation > 0` to skip senders that had just been
  self-destructed-and-recreated; with incarnation gone this filter
  was permanently false and the skip is dead;
- the kvcache write path baked an 8-byte incarnation slot between
  address and storage-location in its composite key but no reader
  ever looked at storage cache entries (only 20-byte account keys
  are served via `CacheView.Get`).

Proto / generated
- kv.proto: AccountChange.incarnation removed; field 2 / name
  "incarnation" marked reserved so the slot can't be silently
  reused by a later field with different semantics.
- kv.pb.go regenerated (only this file kept; the version-only churn
  in 18 other *.pb.go files was reverted to keep the diff scoped).

Native side
- notifications.AccountChange (Go) drops the Incarnation field.
- Accumulator.ChangeAccount / ChangeCode / ChangeStorage drop the
  `incarnation uint64` parameter; the "if newer incarnation, start
  a new change entry" branch was dead (incarnation always 0) and
  is removed.

Callers
- execution/state/rw_v3: serial-mode UpdateAccountData /
  UpdateAccountCode / WriteAccountStorage and the parallel-exec
  versionedio flush no longer pass incarnation.
- execution/stagedsync/stage_execute: unwind-time
  ChangeAccount/ChangeStorage drop incarnation; the now-unused
  `currentInc` local is removed.
- txnprovider/txpool/pool: the `change.Incarnation > 0` skip in the
  Action_UPSERT/UPSERT_CODE branch is removed — every changed
  sender now enters sendersWithChangedState.
- db/kv/kvcache: storage cache key shrinks from [addr(20)+inc(8)+
  loc(32)] = 60 bytes to [addr(20)+loc(32)] = 52 bytes; the
  encoding/binary import is no longer needed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Follow-up to the AccountChange.incarnation removal: across tests that
construct accounts.Account{...} purely to feed the AccountsDomain (or
the Accumulator) and don't exercise incarnation semantics, the
hard-coded Incarnation: 0/1/2/3 was just setup noise. Remove it where
removal leaves the test's intent unchanged.

Touched (incarnation removed):
  txnprovider/txpool/pool_test.go                              — 18
  execution/state/finalize_reader_blockcache_test.go           —  2
  execution/stagedsync/exec3_2cache_test.go                    —  1
  execution/stagedsync/exec3_finalize_test.go                  —  1
  execution/commitment/commitment_calculator_test.go           —  3
  execution/commitment/trie/retain_list_test.go                —  1
  db/state/aggregator_fuzz_test.go                             —  2
  db/state/domain_test.go                                      —  1
  db/state/execctx/domain_shared_test.go                       —  2
  db/state/squeeze_test.go                                     —  7
  db/state/squeeze_concurrent_rebuild_test.go                  —  3
  db/test/aggregator_ext_test.go                               —  6
  db/test/domain_shared_bench_test.go                          —  3
  db/test/domains_restart_test.go                              —  1
  db/integrity/commitment_state_verify_test.go                 —  1

Kept (intentionally exercises the field):
  execution/types/accounts/account_test.go                     — 9, SerializeV3 round-trip
  execution/commitment/trie/account_node_test.go               — 3, DeepEqual roundtrip
  execution/commitment/trie/delete_subrtee_test.go             — 1, DeepEqual roundtrip
  node/shards/state_cache_test.go                              — 1, directly tests acc.Incarnation
  db/kv/kvcache/cache_test.go                                  — 1, t.Skip()-ed test using Incarnation to vary encoded bytes

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Follow-up on the legacy MPT incarnation removal. The
structInfoReceiver.accountLeaf / accountLeafHash methods carried an
incarnation uint64 parameter that flowed into hb.acc.Incarnation, but
no production caller now produces a non-zero value:

- stream.go's StreamHash (the legacy MPT root path) used to copy
  a.Incarnation from accounts.Account into GenStructStepAccountData
  for every account leaf — drop that copy and the field itself.
- trie_from_witness.go already passed a literal 0 with an apologetic
  comment ("Stateless clients don't access the DB so we can just
  pass 0 here") — drop the local and the comment.
- HashBuilder.accountLeaf / accountLeafHash drop the parameter and
  the hb.acc.Incarnation = incarnation assignment. hb.acc.Incarnation
  stays at its zero value (no other writer), so account RLP encoded
  via accountLeafHashWithKey continues to omit the incarnation slot
  exactly as it does today for incarnation-0 accounts.

Net: four files, -21 / +12, no behaviour change on the live
(incarnation-free) state that the legacy MPT path consumes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The StateReader.ReadAccountIncarnation method was the last live reader
of pre-E3 incarnation semantics. After the rest of the incarnation
removal on this branch, every implementation either returned 0
unconditionally, decoded the account body and returned acc.Incarnation
(-1), or delegated to an inner reader that did one of those — no
caller in the tree consumes the result.

Removed:
- StateReader.ReadAccountIncarnation interface method (db/state).
- All implementations:
  - NoopReader, CachedReader, CachedReader3, ReaderV3, bufferedReader,
    HistoryReaderV3, Stateless, versionedStateReader (execution/state).
  - cachedHistoryReaderV3, simulationIntraBlockStateReader,
    RecordingState, witnessStateless (rpc/{rpchelper,jsonrpc}).
- Test stubs (parallel_fixes_test, versionedio_test, calc_state_test,
  exec3_finalize_test).

Removed alongside:
- TrieDbState.incarnationMap field and its four constructor inits
  (NewTrieDbState, Copy, WithNewBuffer, WithLastBuffer). The only
  writer was TrieStateWriter.DeleteAccount caching original.Incarnation
  for a later ReadAccountIncarnation hit; the only reader was
  TrieDbState.ReadAccountIncarnation itself. Both go in one step.
- TrieStateWriter.DeleteAccount's `if original.Incarnation > 0`
  branch (only used to populate incarnationMap).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

valuesEqual's AddressPath branch compared accounts.Account.Incarnation
across the read and write values for parallel-exec validation. With
the account-body Incarnation field now always 0 in production
(per the rest of #12440), the comparison is redundant. Drop it.

The Version{TxIndex, Incarnation} re-execution counter elsewhere in
this file is unrelated to account incarnation (it's the Block-STM
optimistic-concurrency counter) and is preserved.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Final step of the incarnation removal (#12440). The on-disk Incarnation
field is gone from the Account struct and its E3 SerialiseV3 layout.

Encoding changes
- Account struct loses Incarnation and PrevIncarnation.
- SerialiseV3 now writes a 3-section value
    [nonceLen|nonce][balLen|bal][codeLen|codeHash]
  instead of the 4-section
    [nonceLen|nonce][balLen|bal][codeLen|codeHash][incLen|inc]
  Saves one byte per account at minimum (and more for the rare rows
  whose Incarnation was non-zero before the cleanup).
- DeserialiseV3 is tolerant: reads 3 sections, silently skips any
  trailing bytes (so legacy 4-section rows in frozen AccountsDomain
  snapshot files keep decoding correctly). Added explicit bounds
  checking so truncated input now returns an error instead of
  panicking.
- DecodeForStorage's fieldset-bit-4 (Incarnation) handling becomes a
  no-op walk-past; the legacy E2 storage tables are dropped by the
  earlier drop_legacy_e2_tables migration anyway, but the safety net
  costs nothing.
- Helpers GetIncarnation / SetIncarnation / DecodeIncarnationFromStorage
  are removed; Equals / Copy / Reset drop the Incarnation handling.

Migration
- New drop_account_incarnation migration re-encodes every row in
  kv.TblAccountVals and kv.TblAccountHistoryVals via ETL: decode
  (tolerantly), re-encode in the 3-section format, ClearTable +
  bulk Load. Snapshot .kv segments are immutable and stay in the
  legacy format until rebuilt; the tolerant decoder covers them.

Production callers
- node/shards/trie_cache.go: AccountItem.HasPrefix no longer compares
  incarnations.
- execution/state/{rw_v3,triedb_state,intra_block_state,state_object,
  database}.go: dropped Incarnation arguments / map writes / trace
  fields; UpdateAccountCode and WriteAccountStorage are called with
  literal 0 for the now-unused incarnation parameter (signatures
  preserved — orthogonal cleanup).
- execution/commitment/trie/trie.go: AccountNode embeds Account so
  n.Incarnation is gone; the five call sites that plumbed the value
  through getNode / touchAll / deleteRecursive / notifyUnloadRecursive
  pass literal 0 (the parameter itself is dead but kept for now to
  keep the diff focused).
- execution/stagedsync/stage_execute.go + db/state/changeset: the
  "Inc:" field dropped from debug printfs.
- rpc/jsonrpc/otterscan_contract_creator.go: GetContractCreator
  simplified — used to compare acc.Incarnation against latest state
  to skip stale shards; now stops at the first non-empty value.

Tests
- execution/types/accounts/account_test.go rewritten — old file
  was almost entirely Incarnation roundtrip tests. New file covers
  SerialiseV3/DeserialiseV3 happy path + a legacy-format compat test
  with hand-crafted 4-section input to verify the tolerant decoder.
- Across the codebase: Incarnation: N test fixtures replaced with
  Nonce: N where the test relied on distinct encoded bytes;
  Incarnation == 0 assertions dropped.
- txnprovider/txpool/pool_test.go: TestValidateTxReturnsSenderInfoError
  + TestValidateAuthorityReturnsNonce previously fed []byte{0,0,0} as
  a "malformed" sender state — that's a valid 3-section empty account
  under the new format, so the test now uses []byte{0,1} (declares a
  1-byte balance, supplies none) to force a clean DeserialiseV3 error.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The file's exported types (ReconState, NewReconState, ReconnWork,
reconPair, ReconnLess) had zero callers in-tree and inside the file
they only reference each other — pure dead code from a pre-E3
recon path that was never decommissioned. -290 lines.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Two real conflicts, both fallout from prior moksha cleanups landing
on main via separate spinoff PRs while moksha's local versions of
those cleanups were still in flight:

1) execution/commitment/trie/gen_struct_step.go — main inlined
   GenStructStepEx into GenStructStep and dropped GenStructStepOld /
   the orphan HashCollector type aliases via #21284 (the spinoff
   originally cut from this branch). Moksha still had the
   pre-inlining layout. Took main's file as the base and re-applied
   moksha's "drop incarnation from structInfoReceiver" change:
   accountLeaf / accountLeafHash signatures lose `incarnation uint64`,
   GenStructStepAccountData loses its Incarnation field, the call
   sites in the inlined GenStructStep drop the argument.

2) execution/state/triedb_state.go — main dropped the dead
   TrieDbState.hashBuilder field via #21284. Moksha had concurrently
   dropped the dead incarnationMap field. Both removals are correct
   on the merged branch; resolution removes both struct fields and
   their four corresponding constructor inits (NewTrieDbState, Copy,
   WithNewBuffer, WithLastBuffer).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Two conflicts:

1) execution/state/versionmap.go — main split validateRead into a
   thin wrapper plus a new validateReadImpl(..., recursive bool) so
   recursive cross-validates can opt out of the source!=MapRead+nil-
   readVal tiebreaker (PR #21294). The AddressPath cross-check call
   conflicted with moksha's expanded comment (CreateContractPath +
   the deliberate EOA-creation-via-CALL+value gap, post-incarnation
   removal). Resolution: keep moksha's comment + the
   CreateContractPath signal, switch the cross-validate call to the
   new validateReadImpl(..., true) signature.

   Main also added IncarnationPath to the new
   "path != ... else cross-check" guard at line ~422; on moksha
   IncarnationPath does not exist, so the term is dropped.

2) execution/state/state_test.go — main extended SetCode with a
   tracing.CodeChangeReason argument and added a setIncarnation(1)
   call to the test scaffold. moksha already adopted the new SetCode
   signature from main; setIncarnation is gone with the rest of the
   incarnation removal. Resolution: take main's SetCode call, drop
   the setIncarnation line.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Five conflicts:

1) cmd/rpctest/rpctest/account_range_verify.go — main added defer
   Close() on the two MDBX handles. Trivial: kept both.

2) db/migrations/migrations.go — main landed #21280 (the legacy E2
   tables drop spinoff originally cut from this branch), which adds
   dropLegacyE2Tables. Moksha has both dropLegacyE2Tables and
   dropAccountIncarnation registered. Resolution: keep both in the
   ChainDB sequence.

3) db/kv/kvcache/cache.go — main removed AssertCheckValues entirely
   (the txpool stub call site went away with #21280's caller cleanup).
   Moksha had it as a no-op stub. Take main's deletion; the only
   reference left was a commented-out test line.

4) execution/stagedsync/exec3_finalize_test.go — PR #21211 deleted
   ~240 lines of test scenarios (TestFinalizeTx_SimpleTransfer,
   _London, _AllScenarios, coinbaseIsRecipientScenario,
   selfTransferScenario, hasCoinbaseDelta, adjustForTransferDelta)
   because the finalizeWithIBS / finalizeTx (delta-args) code paths
   they exercised were dead. Take main's deletion.

5) execution/state/intra_block_state.go — three blocks:

   a) Same-block-revival check in CreateAccount. Main (#21319)
      simplified it to `!account.Empty()` on the version-map-refreshed
      record. Moksha had a path-by-path LatestTxIndex scan
      (BalancePath/NoncePath/CodeHashPath > destructTxIndex). Take
      main's cleaner check — semantically equivalent on the same
      input (the refreshed `account` reflects exactly those higher-
      txIndex writes).

   b) prevInc / IncarnationPath bookkeeping for CreateAccount on
      main. Entirely incarnation-dependent; moksha has no
      IncarnationPath and no Account.Incarnation. Take HEAD's empty
      resolution.

   c) Synthetic versionRead(IncarnationPath, ...) in CreateAccount on
      main. Same reason. Take HEAD's empty resolution.

Plus two non-conflicting touch-ups dropping IncarnationPath references
that the merge silently brought in via main-side changes:

- execution/state/versionmap.go MVReadResultNone/MapRead recursive-
  cross-check whitelist dropped IncarnationPath from the path list.
- execution/state/versionedio.go SD short-circuit zero-value-fallback
  whitelist dropped IncarnationPath from the path list.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Re-align the accounts.Account struct literal whose Incarnation field was removed when merging origin/main into this incarnation-dropping branch.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…ionMap

createObject's two gates were anchored on `sdb.versionMap == nil`, which
also short-circuited the blockgen+versionMap path (chain generation runs
the versioned-IBS whenever EXEC3_PARALLEL is on). Result: tx0 SD then
tx1 value-transfer-revive (or revive-then-revert) left AA's stale
storage subtree in the trie root that blockgen wrote to header.Root,
and InsertChain's correctly-wiped root then mismatched, failing
TestDeleteRecreateAccount and TestDeleteCreateRevert under
ERIGON_EXEC3_PARALLEL=true.

Re-anchor both gates on versionMapMarker / hasWrite(SelfDestructPath)
instead: only synthetic markers (parallel-apply per-tx IBS) keep
createObjectChange / skip recreatedFromDestructed; real same-IBS prior
state (serial and blockgen+versionMap) restores via resetObjectChange
and sets the flag, so MakeWriteSet emits DeleteAccount and the storage
prefix wipe propagates into the header root.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

- trie: getNode no longer returns an always-zero incarnation; touchAll,
  deleteRecursive and notifyUnloadRecursive no longer take an unread
  incarnation parameter. Strip a stale incarnation comment on insert and
  on findSubTriesToLoad's storage-subtree arm.
- StateWriter: UpdateAccountCode and WriteAccountStorage no longer take
  an incarnation argument; updated every implementation and caller
  (NoopWriter, Stateless, TrieStateWriter, CachedWriter, Writer,
  versionedWriteCollector, LightCollector, diffTrackingWriter, StateDiff,
  RecordingState, witnessStateless and their callsites). TrieStateWriter
  no longer writes the now-unused storageIncarnation map field.
- composite_keys: drop the historical "incarnation between hashes"
  paragraph from GenerateCompositeStorageKey's docstring.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

…tion

Both maps on Buffer were write-only — created in initialise, deleted
from in DeleteAccount / CreateContract / merge, and copied into the
aggregate buffer — but never iterated or read for any computation.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

- stream.go: remove the historical "16 nibbles of incarnation between
  account hash and storage subtree" comment on AccountNode handling; the
  trie hasn't laid those nibbles down since the IBS-side counter was
  removed.
- structural_test.go: TestEmbeddedStorage built an `incarnation` byte
  slice and never used it (the locationKeys are pure addrHash + location
  hash). Drop the dead local and rewrite the docstring to match what the
  test actually does.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Rewrite or delete comments that still described the removed account-
incarnation counter — references to "next incarnations" to compute on
contract creation, "original.Incarnation > account.Incarnation" cleanup
predicates, "incarnation prefix nibbles" in the storage trie path,
"all 4 account fields (incl. incarnation)" emit lists, and "but maybe
not our Incarnation" in Otterscan's history scan. Also drop two dead
commented-out blocks: a `kv.IncarnationMap` dump in reset_state and a
WalkWrites assertion stub in state_cache_test.

Migration / deprecated-table docs and Block-STM tx-incarnation
(execution-retry counter — a different concept) are left untouched.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

db_walkthrough.MD documented the E2 layout (PlainState, HashedState,
HistoryOfAccounts, AccountChanges/StorageChanges, IntermediateTrieHashes
- now all *Deprecated in db/kv/tables.go and slated for deletion by
db/migrations/drop_legacy_e2_tables.go). E3's domain layout
(AccountsDomain, StorageDomain, CommitmentDomain backed by snapshots)
isn't covered anywhere in the doc, so it's actively misleading rather
than incomplete.

Removes the .MD, all 30 supporting genesis/block1/geth screenshots, and
the three remaining references (README, dupsort.md, two tables.go
comments).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

erigon.readthedocs.io is dead — the project slug exists on RTD but
/en/latest/ 404s; no build has been published in a long time. The seven
.rst sources have only seen mechanical import-path refactors since
Jan 2024; the substantive content has drifted (e.g. stagedsync.rst
walks through the long-gone SpawnHeaderDownloadStage shape, types.rst
claims Address still lives in common as [20]byte, snappy.rst is about
the pre-E3 block-body compression). Nobody is minding it.

Removes docs/readthedocs/, .readthedocs.yml, and the docs/readthedocs/build
entry from .gitignore.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

…otifications README

- guide.md: delete the "Attack - by delete account with huge state"
  subsection. It described the PlainState/HashedState/IntermediateTrieHash
  incarnation-tagging mitigation that no longer exists, and linked to
  execution/commitment/trie/trie_root.go and an accAddrHashWithInc symbol
  that have both been removed.
- notifications/README.md: strip the incarnation argument from the three
  Accumulator.Change* call signatures in the sender example, matching
  the current accumulator.go method set.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

# Conflicts:
#	execution/state/database.go
#	execution/state/rw_v3.go
#	execution/state/versionedio.go