Cache code and invalid jump destination tables (fixes #2268) (#2404)

It is common for many accounts to share the same code - at the database
level, code is stored by hash meaning only one copy exists per unique
program but when loaded in memory, a copy is made for each account.

Further, every time we execute the code, it must be scanned for invalid
jump destinations which slows down EVM exeuction.

Finally, the extcodesize call causes code to be loaded even if only the
size is needed.

This PR improves on all these points by introducing a shared
CodeBytesRef type whose code section is immutable and that can be shared
between accounts. Further, a dedicated `len` API call is added so that
the EXTCODESIZE opcode can operate without polluting the GC and code
cache, for cases where only the size is requested - rocksdb will in this
case cache the code itself in the row cache meaning that lookup of the
code itself remains fast when length is asked for first.

With 16k code entries, there's a 90% hit rate which goes up to 99%
during the 2.3M attack - the cache significantly lowers memory
consumption and execution time not only during this event but across the
board.

examples/decompile_smart_contract.nim

@@ -1,5 +1,5 @@
 # Nimbus
-# Copyright (c) 2018-2023 Status Research & Development GmbH
+# Copyright (c) 2018-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or
 #    http://www.apache.org/licenses/LICENSE-2.0)
@@ -8,9 +8,10 @@
 # at your option. This file may not be copied, modified, or distributed except
 # according to those terms.
 
-import ../nimbus/vm/code_stream, strformat
+import ../nimbus/evm/code_stream, strformat
 
-var c = newCodeStreamFromUnescaped("0x6003600202600055")
+var c =
+  CodeStream.init(CodeBytesRef.fromHex("0x6003600202600055").expect("valid code"))
 
 let opcodes = c.decompile()
 for op in opcodes:

nimbus/core/eip4844.nim

@@ -18,8 +18,6 @@ import
   ../constants,
   ../common/common
 
-from std/sequtils import mapIt
-
 {.push raises: [].}
 
 type

nimbus/db/core_db/backend/aristo_db/handlers_kvt.nim

@@ -107,6 +107,19 @@ proc kvtMethods(cKvt: KvtCoreDbKvtRef): CoreDbKvtFns =
     else:
       rc.toRc(cKvt.base, info)
 
+  proc kvtLen(
+      cKvt: KvtCoreDbKvtRef;
+      k: openArray[byte];
+      info: static[string];
+        ): CoreDbRc[int] =
+    let rc = cKvt.base.api.len(cKvt.kvt, k)
+    if rc.isOk:
+      ok(rc.value)
+    elif rc.error == GetNotFound:
+      err(rc.error.toError(cKvt.base, info, KvtNotFound))
+    else:
+      rc.toRc(cKvt.base, info)
+
   proc kvtPut(
       cKvt: KvtCoreDbKvtRef;
       k: openArray[byte];
@@ -148,6 +161,9 @@ proc kvtMethods(cKvt: KvtCoreDbKvtRef): CoreDbKvtFns =
     getFn: proc(k: openArray[byte]): CoreDbRc[Blob] =
       cKvt.kvtGet(k, "getFn()"),
 
+    lenFn: proc(k: openArray[byte]): CoreDbRc[int] =
+      cKvt.kvtLen(k, "lenFn()"),
+
     delFn: proc(k: openArray[byte]): CoreDbRc[void] =
       cKvt.kvtDel(k, "delFn()"),
 

nimbus/db/core_db/base.nim

@@ -255,6 +255,12 @@ proc get*(kvt: CoreDbKvtRef; key: openArray[byte]): CoreDbRc[Blob] =
   result = kvt.methods.getFn key
   kvt.ifTrackNewApi: debug newApiTxt, api, elapsed, key=key.toStr, result
 
+proc len*(kvt: CoreDbKvtRef; key: openArray[byte]): CoreDbRc[int] =
+  ## This function always returns a non-empty `Blob` or an error code.
+  kvt.setTrackNewApi KvtLenFn
+  result = kvt.methods.lenFn key
+  kvt.ifTrackNewApi: debug newApiTxt, api, elapsed, key=key.toStr, result
+
 proc getOrEmpty*(kvt: CoreDbKvtRef; key: openArray[byte]): CoreDbRc[Blob] =
   ## This function sort of mimics the behaviour of the legacy database
   ## returning an empty `Blob` if the argument `key` is not found on the

nimbus/db/core_db/base/api_tracking.nim

@@ -70,6 +70,7 @@ type
     KvtDelFn            = "kvt/del"
     KvtForgetFn         = "kvt/forget"
     KvtGetFn            = "kvt/get"
+    KvtLenFn            = "kvt/len"
     KvtGetOrEmptyFn     = "kvt/getOrEmpty"
     KvtHasKeyFn         = "kvt/hasKey"
     KvtPairsIt          = "kvt/pairs"

nimbus/db/core_db/base/base_desc.nim

@@ -140,6 +140,7 @@ type
   # --------------------------------------------------
   CoreDbKvtBackendFn* = proc(): CoreDbKvtBackendRef {.noRaise.}
   CoreDbKvtGetFn* = proc(k: openArray[byte]): CoreDbRc[Blob] {.noRaise.}
+  CoreDbKvtLenFn* = proc(k: openArray[byte]): CoreDbRc[int] {.noRaise.}
   CoreDbKvtDelFn* = proc(k: openArray[byte]): CoreDbRc[void] {.noRaise.}
   CoreDbKvtPutFn* =
     proc(k: openArray[byte]; v: openArray[byte]): CoreDbRc[void] {.noRaise.}
@@ -150,6 +151,7 @@ type
     ## Methods for key-value table
     backendFn*:     CoreDbKvtBackendFn
     getFn*:         CoreDbKvtGetFn
+    lenFn*:         CoreDbKvtLenFn
     delFn*:         CoreDbKvtDelFn
     putFn*:         CoreDbKvtPutFn
     hasKeyFn*:      CoreDbKvtHasKeyFn

nimbus/db/core_db/base/validate.nim

@@ -44,6 +44,7 @@ proc validateMethodsDesc(base: CoreDbBaseFns) =
 proc validateMethodsDesc(kvt: CoreDbKvtFns) =
   doAssert not kvt.backendFn.isNil
   doAssert not kvt.getFn.isNil
+  doAssert not kvt.lenFn.isNil
   doAssert not kvt.delFn.isNil
   doAssert not kvt.putFn.isNil
   doAssert not kvt.hasKeyFn.isNil

nimbus/db/kvt/kvt_api.nim

@@ -50,6 +50,8 @@ type
     backLevel: int): Result[KvtDbRef,KvtError] {.noRaise.}
   KvtApiGetFn* = proc(db: KvtDbRef,
     key: openArray[byte]): Result[Blob,KvtError] {.noRaise.}
+  KvtApiLenFn* = proc(db: KvtDbRef,
+    key: openArray[byte]): Result[int,KvtError] {.noRaise.}
   KvtApiHasKeyFn* = proc(db: KvtDbRef,
     key: openArray[byte]): Result[bool,KvtError] {.noRaise.}
   KvtApiIsCentreFn* = proc(db: KvtDbRef): bool {.noRaise.}
@@ -76,6 +78,7 @@ type
     forget*: KvtApiForgetFn
     forkTx*: KvtApiForkTxFn
     get*: KvtApiGetFn
+    len*: KvtApiLenFn
     hasKey*: KvtApiHasKeyFn
     isCentre*: KvtApiIsCentreFn
     isTop*: KvtApiIsTopFn
@@ -100,6 +103,7 @@ type
     KvtApiProfForgetFn       = "forget"
     KvtApiProfForkTxFn       = "forkTx"
     KvtApiProfGetFn          = "get"
+    KvtApiProfLenFn          = "len"
     KvtApiProfHasKeyFn       = "hasKey"
     KvtApiProfIsCentreFn     = "isCentre"
     KvtApiProfIsTopFn        = "isTop"
@@ -114,6 +118,7 @@ type
     KvtApiProfTxTopFn        = "txTop"
 
     KvtApiProfBeGetKvpFn     = "be/getKvp"
+    KvtApiProfBeLenKvpFn     = "be/lenKvp"
     KvtApiProfBePutKvpFn     = "be/putKvp"
     KvtApiProfBePutEndFn     = "be/putEnd"
 
@@ -176,6 +181,7 @@ func init*(api: var KvtApiObj) =
   api.forget = forget
   api.forkTx = forkTx
   api.get = get
+  api.len = len
   api.hasKey = hasKey
   api.isCentre = isCentre
   api.isTop = isTop
@@ -203,6 +209,7 @@ func dup*(api: KvtApiRef): KvtApiRef =
     forget:     api.forget,
     forkTx:     api.forkTx,
     get:        api.get,
+    len:        api.len,
     hasKey:     api.hasKey,
     isCentre:   api.isCentre,
     isTop:      api.isTop,
@@ -275,6 +282,11 @@ func init*(
       KvtApiProfGetFn.profileRunner:
         result = api.get(a, b)
 
+  profApi.len =
+    proc(a: KvtDbRef, b: openArray[byte]): auto =
+      KvtApiProfLenFn.profileRunner:
+        result = api.len(a, b)
+
   profApi.hasKey =
     proc(a: KvtDbRef, b: openArray[byte]): auto =
       KvtApiProfHasKeyFn.profileRunner:
@@ -346,6 +358,12 @@ func init*(
           result = be.getKvpFn(a)
     data.list[KvtApiProfBeGetKvpFn.ord].masked = true
 
+    beDup.lenKvpFn =
+      proc(a: openArray[byte]): auto =
+        KvtApiProfBeLenKvpFn.profileRunner:
+          result = be.lenKvpFn(a)
+    data.list[KvtApiProfBeLenKvpFn.ord].masked = true
+
     beDup.putKvpFn =
       proc(a: PutHdlRef; b: openArray[(Blob,Blob)]) =
         be.putKvpFn(a,b)

nimbus/db/kvt/kvt_desc/desc_backend.nim

@@ -23,6 +23,9 @@ type
   GetKvpFn* =
     proc(key: openArray[byte]): Result[Blob,KvtError] {.gcsafe, raises: [].}
       ## Generic backend database retrieval function
+  LenKvpFn* =
+    proc(key: openArray[byte]): Result[int,KvtError] {.gcsafe, raises: [].}
+      ## Generic backend database retrieval function
 
   # -------------
 
@@ -76,6 +79,7 @@ type
     ## Backend interface.
 
     getKvpFn*: GetKvpFn              ## Read key-value pair
+    lenKvpFn*: LenKvpFn              ## Read key-value pair length
 
     putBegFn*: PutBegFn              ## Start bulk store session
     putKvpFn*: PutKvpFn              ## Bulk store key-value pairs
@@ -88,6 +92,7 @@ type
 
 proc init*(trg: var BackendObj; src: BackendObj) =
   trg.getKvpFn = src.getKvpFn
+  trg.lenKvpFn = src.lenKvpFn
   trg.putBegFn = src.putBegFn
   trg.putKvpFn = src.putKvpFn
   trg.putEndFn = src.putEndFn

nimbus/db/kvt/kvt_init/memory_db.nim

@@ -82,6 +82,16 @@ proc getKvpFn(db: MemBackendRef): GetKvpFn =
         return ok(move(data))
       err(GetNotFound)
 
+proc lenKvpFn(db: MemBackendRef): LenKvpFn =
+  result =
+    proc(key: openArray[byte]): Result[int,KvtError] =
+      if key.len == 0:
+        return err(KeyInvalid)
+      var data = db.mdb.tab.getOrVoid @key
+      if data.isValid:
+        return ok(data.len)
+      err(GetNotFound)
+
 # -------------
 
 proc putBegFn(db: MemBackendRef): PutBegFn =
@@ -140,6 +150,7 @@ proc memoryBackend*: BackendRef =
     mdb:    MemDbRef())
 
   db.getKvpFn = getKvpFn db
+  db.lenKvpFn = lenKvpFn db
 
   db.putBegFn = putBegFn db
   db.putKvpFn = putKvpFn db

nimbus/db/kvt/kvt_init/rocks_db.nim

@@ -92,6 +92,22 @@ proc getKvpFn(db: RdbBackendRef): GetKvpFn =
 
       err(GetNotFound)
 
+proc lenKvpFn(db: RdbBackendRef): LenKvpFn =
+  result =
+    proc(key: openArray[byte]): Result[int,KvtError] =
+
+      # Get data record
+      var len = db.rdb.len(key).valueOr:
+        when extraTraceMessages:
+          debug logTxt "lenKvpFn() failed", key, error=error[0], info=error[1]
+        return err(error[0])
+
+      # Return if non-empty
+      if 0 < len:
+        return ok(len)
+
+      err(GetNotFound)
+
 # -------------
 
 proc putBegFn(db: RdbBackendRef): PutBegFn =
@@ -268,6 +284,7 @@ proc rocksDbKvtBackend*(
     return err(error)
 
   db.getKvpFn = getKvpFn db
+  db.lenKvpFn = lenKvpFn db
 
   db.putBegFn = putBegFn db
   db.putKvpFn = putKvpFn db
@@ -297,6 +314,7 @@ proc rocksDbKvtTriggeredBackend*(
     return err((RdbBeHostError,$error))
 
   db.getKvpFn = getKvpFn db
+  db.lenKvpFn = lenKvpFn db
 
   db.putBegFn = putBegTriggeredFn db
   db.putKvpFn = putKvpFn db

nimbus/db/kvt/kvt_init/rocks_db/rdb_get.nim

@@ -53,6 +53,24 @@ proc get*(
     res = EmptyBlob
   ok move(res)
 
+proc len*(
+    rdb: RdbInst;
+    key: openArray[byte],
+      ): Result[int,(KvtError,string)] =
+  var res: int
+  let onData: DataProc = proc(data: openArray[byte]) =
+    res = data.len
+
+  let gotData = rdb.store[KvtGeneric].get(key, onData).valueOr:
+    const errSym = RdbBeDriverGetError
+    when extraTraceMessages:
+      trace logTxt "len", error=errSym, info=error
+    return err((errSym,error))
+
+  if not gotData:
+    res = 0
+  ok res
+
 # ------------------------------------------------------------------------------
 # End
 # ------------------------------------------------------------------------------

nimbus/db/kvt/kvt_layers.nim

@@ -31,18 +31,25 @@ func nLayersKeys*(db: KvtDbRef): int =
 # Public functions: get function
 # ------------------------------------------------------------------------------
 
-func layersHasKey*(db: KvtDbRef; key: openArray[byte]|seq[byte]): bool =
+func layersLen*(db: KvtDbRef; key: openArray[byte]|seq[byte]): Opt[int] =
   ## Return `true` id the argument key is cached.
   ##
   when key isnot seq[byte]:
     let key = @key
-  if db.top.delta.sTab.hasKey key:
+
-    return true
+  db.top.delta.sTab.withValue(key, item):
+    return Opt.some(item[].len())
 
   for w in db.rstack:
-    if w.delta.sTab.hasKey key:
+    w.delta.sTab.withValue(key, item):
-      return true
+      return Opt.some(item[].len())
 
+  Opt.none(int)
+
+func layersHasKey*(db: KvtDbRef; key: openArray[byte]|seq[byte]): bool =
+  ## Return `true` id the argument key is cached.
+  ##
+  db.layersLen(key).isSome()
 
 func layersGet*(db: KvtDbRef; key: openArray[byte]|seq[byte]): Opt[Blob] =
   ## Find an item on the cache layers. An `ok()` result might contain an

nimbus/db/kvt/kvt_utils.nim

@@ -36,6 +36,18 @@ proc getUbe*(
     return be.getKvpFn key
   err(GetNotFound)
 
+proc getUbeLen*(
+    db: KvtDbRef;                     # Database
+    key: openArray[byte];             # Key of database record
+      ): Result[int,KvtError] =
+  ## For the argument `key` return the associated value from the backend
+  ## database if available.
+  ##
+  let be = db.backend
+  if not be.isNil:
+    return be.lenKvpFn key
+  err(GetNotFound)
+
 proc getBe*(
     db: KvtDbRef;                     # Database
     key: openArray[byte];             # Key of database record
@@ -48,6 +60,18 @@ proc getBe*(
       return ok(w[])
   db.getUbe key
 
+proc getBeLen*(
+    db: KvtDbRef;                     # Database
+    key: openArray[byte];             # Key of database record
+      ): Result[int,KvtError] =
+  ## Get the vertex from the (filtered) backened if available.
+  if not db.balancer.isNil:
+    db.balancer.sTab.withValue(@key, w):
+      if w[].len == 0:
+        return err(GetNotFound)
+      return ok(w[].len)
+  db.getUbeLen key
+
 # ------------
 
 proc put*(
@@ -95,6 +119,19 @@ proc get*(
 
   return ok(move(data))
 
+proc len*(
+    db: KvtDbRef;                     # Database
+    key: openArray[byte];             # Key of database record
+      ): Result[int,KvtError] =
+  ## For the argument `key` return the associated value preferably from the
+  ## top layer, or the database otherwise.
+  ##
+  if key.len == 0:
+    return err(KeyInvalid)
+
+  let len = db.layersLen(key).valueOr:
+    return db.getBeLen key
+  ok(len)
 
 proc hasKey*(
     db: KvtDbRef;                     # Database

nimbus/db/ledger/accounts_ledger.nim

@@ -25,6 +25,7 @@
 ##
 
 import
+  stew/keyed_queue,
   std/[tables, hashes, sets],
   chronicles,
   eth/[common, rlp],
@@ -33,10 +34,20 @@ import
   "../.."/[constants, utils/utils],
   ../access_list as ac_access_list,
   ".."/[core_db, storage_types, transient_storage],
+  ../../evm/code_bytes,
   ./distinct_ledgers
 
+export code_bytes
+
 const
   debugAccountsLedgerRef = false
+  codeLruSize = 16*1024
+    # An LRU cache of 16K items gives roughly 90% hit rate anecdotally on a
+    # small range of test blocks - this number could be studied in more detail
+    # Per EIP-170, a the code of a contract can be up to `MAX_CODE_SIZE` = 24kb,
+    # which would cause a worst case of 386MB memory usage though in reality
+    # code sizes are much smaller - it would make sense to study these numbers
+    # in greater detail.
 
 type
   AccountFlag = enum
@@ -44,7 +55,6 @@ type
     IsNew
     Dirty
     Touched
-    CodeLoaded
     CodeChanged
     StorageChanged
     NewlyCreated # EIP-6780: self destruct only in same transaction
@@ -54,7 +64,7 @@ type
   AccountRef = ref object
     statement: CoreDbAccount
     flags: AccountFlags
-    code: seq[byte]
+    code: CodeBytesRef
     originalStorage: TableRef[UInt256, UInt256]
     overlayStorage: Table[UInt256, UInt256]
 
@@ -69,6 +79,16 @@ type
     witnessCache: Table[EthAddress, WitnessData]
     isDirty: bool
     ripemdSpecial: bool
+    code: KeyedQueue[Hash256, CodeBytesRef]
+      ## The code cache provides two main benefits:
+      ##
+      ## * duplicate code is shared in memory beween accounts
+      ## * the jump destination table does not have to be recomputed for every
+      ##   execution, for commonly called called contracts
+      ##
+      ## The former feature is specially important in the 2.3-2.7M block range
+      ## when underpriced code opcodes are being run en masse - both advantages
+      ## help performance broadly as well.
 
   ReadOnlyStateDB* = distinct AccountsLedgerRef
 
@@ -330,7 +350,7 @@ proc persistMode(acc: AccountRef): PersistMode =
 proc persistCode(acc: AccountRef, ac: AccountsLedgerRef) =
   if acc.code.len != 0:
     let rc = ac.kvt.put(
-      contractHashKey(acc.statement.codeHash).toOpenArray, acc.code)
+      contractHashKey(acc.statement.codeHash).toOpenArray, acc.code.bytes())
     if rc.isErr:
       warn logTxt "persistCode()",
        codeHash=acc.statement.codeHash, error=($$rc.error)
@@ -412,33 +432,49 @@ proc getNonce*(ac: AccountsLedgerRef, address: EthAddress): AccountNonce =
   if acc.isNil: emptyEthAccount.nonce
   else: acc.statement.nonce
 
-proc getCode(acc: AccountRef, kvt: CoreDbKvtRef): lent seq[byte] =
+proc getCode*(ac: AccountsLedgerRef, address: EthAddress): CodeBytesRef =
-  if CodeLoaded notin acc.flags and CodeChanged notin acc.flags:
+  # Always returns non-nil!
-    if acc.statement.codeHash != EMPTY_CODE_HASH:
-      var rc = kvt.get(contractHashKey(acc.statement.codeHash).toOpenArray)
-      if rc.isErr:
-        warn logTxt "getCode()", codeHash=acc.statement.codeHash, error=($$rc.error)
-      else:
-        acc.code = move(rc.value)
-        acc.flags.incl CodeLoaded
-  else:
-    acc.flags.incl CodeLoaded # avoid hash comparisons
-
-  acc.code
-
-proc getCode*(ac: AccountsLedgerRef, address: EthAddress): seq[byte] =
   let acc = ac.getAccount(address, false)
   if acc.isNil:
-    return
+    return CodeBytesRef()
 
-  acc.getCode(ac.kvt)
+  if acc.code == nil:
+    acc.code =
+      if acc.statement.codeHash != EMPTY_CODE_HASH:
+        ac.code.lruFetch(acc.statement.codeHash).valueOr:
+          var rc = ac.kvt.get(contractHashKey(acc.statement.codeHash).toOpenArray)
+          if rc.isErr:
+            warn logTxt "getCode()", codeHash=acc.statement.codeHash, error=($$rc.error)
+            CodeBytesRef()
+          else:
+            let newCode = CodeBytesRef.init(move(rc.value))
+            ac.code.lruAppend(acc.statement.codeHash, newCode, codeLruSize)
+      else:
+        CodeBytesRef()
+
+  acc.code
 
 proc getCodeSize*(ac: AccountsLedgerRef, address: EthAddress): int =
   let acc = ac.getAccount(address, false)
   if acc.isNil:
-    return
+    return 0
 
-  acc.getCode(ac.kvt).len
+  if acc.code == nil:
+    if acc.statement.codeHash == EMPTY_CODE_HASH:
+      return 0
+    acc.code = ac.code.lruFetch(acc.statement.codeHash).valueOr:
+      # On a cache miss, we don't fetch the code - instead, we fetch just the
+      # length - should the code itself be needed, it will typically remain
+      # cached and easily accessible in the database layer - this is to prevent
+      # EXTCODESIZE calls from messing up the code cache and thus causing
+      # recomputation of the jump destination table
+      var rc = ac.kvt.len(contractHashKey(acc.statement.codeHash).toOpenArray)
+
+      return rc.valueOr:
+        warn logTxt "getCodeSize()", codeHash=acc.statement.codeHash, error=($$rc.error)
+        0
+
+  acc.code.len()
 
 proc getCommittedStorage*(ac: AccountsLedgerRef, address: EthAddress, slot: UInt256): UInt256 =
   let acc = ac.getAccount(address, false)
@@ -521,7 +557,9 @@ proc setCode*(ac: AccountsLedgerRef, address: EthAddress, code: seq[byte]) =
   if acc.statement.codeHash != codeHash:
     var acc = ac.makeDirty(address)
     acc.statement.codeHash = codeHash
-    acc.code = code
+    # Try to reuse cache entry if it exists, but don't save the code - it's not
+    # a given that it will be executed within LRU range
+    acc.code = ac.code.lruFetch(codeHash).valueOr(CodeBytesRef.init(code))
     acc.flags.incl CodeChanged
 
 proc setStorage*(ac: AccountsLedgerRef, address: EthAddress, slot, value: UInt256) =
@@ -701,7 +739,7 @@ proc getStorageRoot*(ac: AccountsLedgerRef, address: EthAddress): Hash256 =
 proc update(wd: var WitnessData, acc: AccountRef) =
   # once the code is touched make sure it doesn't get reset back to false in another update
   if not wd.codeTouched:
-    wd.codeTouched = CodeChanged in acc.flags or CodeLoaded in acc.flags
+    wd.codeTouched = CodeChanged in acc.flags or acc.code != nil
 
   if not acc.originalStorage.isNil:
     for k, v in acc.originalStorage:
@@ -801,7 +839,7 @@ proc getStorageRoot*(db: ReadOnlyStateDB, address: EthAddress): Hash256 {.borrow
 proc getBalance*(db: ReadOnlyStateDB, address: EthAddress): UInt256 {.borrow.}
 proc getStorage*(db: ReadOnlyStateDB, address: EthAddress, slot: UInt256): UInt256 {.borrow.}
 proc getNonce*(db: ReadOnlyStateDB, address: EthAddress): AccountNonce {.borrow.}
-proc getCode*(db: ReadOnlyStateDB, address: EthAddress): seq[byte] {.borrow.}
+proc getCode*(db: ReadOnlyStateDB, address: EthAddress): CodeBytesRef {.borrow.}
 proc getCodeSize*(db: ReadOnlyStateDB, address: EthAddress): int {.borrow.}
 proc contractCollision*(db: ReadOnlyStateDB, address: EthAddress): bool {.borrow.}
 proc accountExists*(db: ReadOnlyStateDB, address: EthAddress): bool {.borrow.}

nimbus/db/ledger/base.nim

@@ -14,6 +14,7 @@
 
 import
   eth/common,
+  ../../evm/code_bytes,
   ../../stateless/multi_keys,
   ../core_db,
   ./base/[api_tracking, base_desc]
@@ -33,6 +34,7 @@ type
   ReadOnlyStateDB* = distinct LedgerRef
 
 export
+  code_bytes,
   LedgerFnInx,
   LedgerProfListRef,
   LedgerRef,
@@ -175,7 +177,7 @@ proc getBalance*(ldg: LedgerRef, eAddr: EthAddress): UInt256 =
   result = ldg.ac.getBalance(eAddr)
   ldg.ifTrackApi: debug apiTxt, api, elapsed, eAddr, result
 
-proc getCode*(ldg: LedgerRef, eAddr: EthAddress): Blob =
+proc getCode*(ldg: LedgerRef, eAddr: EthAddress): CodeBytesRef =
   ldg.beginTrackApi LdgGetCodeFn
   result = ldg.ac.getCode(eAddr)
   ldg.ifTrackApi: debug apiTxt, api, elapsed, eAddr, result=result.toStr
@@ -371,7 +373,7 @@ proc getStorageRoot*(db: ReadOnlyStateDB, eAddr: EthAddress): Hash256 {.borrow.}
 proc getBalance*(db: ReadOnlyStateDB, eAddr: EthAddress): UInt256 {.borrow.}
 proc getStorage*(db: ReadOnlyStateDB, eAddr: EthAddress, slot: UInt256): UInt256 {.borrow.}
 proc getNonce*(db: ReadOnlyStateDB, eAddr: EthAddress): AccountNonce {.borrow.}
-proc getCode*(db: ReadOnlyStateDB, eAddr: EthAddress): seq[byte] {.borrow.}
+proc getCode*(db: ReadOnlyStateDB, eAddr: EthAddress): CodeBytesRef {.borrow.}
 proc getCodeSize*(db: ReadOnlyStateDB, eAddr: EthAddress): int {.borrow.}
 proc contractCollision*(db: ReadOnlyStateDB, eAddr: EthAddress): bool {.borrow.}
 proc accountExists*(db: ReadOnlyStateDB, eAddr: EthAddress): bool {.borrow.}

nimbus/evm/code_bytes.nim

@@ -0,0 +1,74 @@
+# Nimbus
+# Copyright (c) 2018-2024 Status Research & Development GmbH
+# Licensed under either of
+#  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or http://www.apache.org/licenses/LICENSE-2.0)
+#  * MIT license ([LICENSE-MIT](LICENSE-MIT) or http://opensource.org/licenses/MIT)
+# at your option. This file may not be copied, modified, or distributed except according to those terms.
+
+import stew/byteutils, results, ./interpreter/op_codes
+
+export results
+
+type CodeBytesRef* = ref object
+  ## Code buffer that caches invalid jump positions used for verifying jump
+  ## destinations - `bytes` is immutable once instances is created while
+  ## `invalidPositions` will be built up on demand
+  bytes: seq[byte]
+  invalidPositions: seq[byte] # bit seq of invalid jump positions
+  processed: int
+
+template bitpos(pos: int): (int, byte) =
+  (pos shr 3, 1'u8 shl (pos and 0x07))
+
+func init*(T: type CodeBytesRef, bytes: sink seq[byte]): CodeBytesRef =
+  let ip = newSeq[byte]((bytes.len + 7) div 8)
+  CodeBytesRef(bytes: move(bytes), invalidPositions: ip)
+
+func init*(T: type CodeBytesRef, bytes: openArray[byte]): CodeBytesRef =
+  CodeBytesRef.init(@bytes)
+
+func init*(T: type CodeBytesRef, bytes: openArray[char]): CodeBytesRef =
+  CodeBytesRef.init(bytes.toOpenArrayByte(0, bytes.high()))
+
+func fromHex*(T: type CodeBytesRef, hex: string): Opt[CodeBytesRef] =
+  try:
+    Opt.some(CodeBytesRef.init(hexToSeqByte(hex)))
+  except ValueError:
+    Opt.none(CodeBytesRef)
+
+func invalidPosition(c: CodeBytesRef, pos: int): bool =
+  let (bpos, bbit) = bitpos(pos)
+  (c.invalidPositions[bpos] and bbit) > 0
+
+func bytes*(c: CodeBytesRef): lent seq[byte] =
+  c[].bytes
+
+func len*(c: CodeBytesRef): int =
+  len(c.bytes)
+
+func isValidOpcode*(c: CodeBytesRef, position: int): bool =
+  if position >= len(c):
+    false
+  elif c.invalidPosition(position):
+    false
+  elif position <= c.processed:
+    true
+  else:
+    var i = c.processed
+    while i <= position:
+      var opcode = Op(c.bytes[i])
+      if opcode >= Op.Push1 and opcode <= Op.Push32:
+        var leftBound = (i + 1)
+        var rightBound = leftBound + (opcode.int - 95)
+        for z in leftBound ..< rightBound:
+          let (bpos, bbit) = bitpos(z)
+          c.invalidPositions[bpos] = c.invalidPositions[bpos] or bbit
+        i = rightBound
+      else:
+        i += 1
+    c.processed = i - 1
+
+    not c.invalidPosition(position)
+
+func `==`*(a: CodeBytesRef, b: openArray[byte]): bool =
+  a.bytes == b

nimbus/evm/code_stream.nim

@@ -6,76 +6,52 @@
 # at your option. This file may not be copied, modified, or distributed except according to those terms.
 
 import
-  chronicles, strformat, strutils, sequtils, parseutils,
+  std/[sequtils, strutils], chronicles, eth/common, ./interpreter/op_codes, ./code_bytes
-  eth/common,
-  ./interpreter/op_codes
 
-logScope:
+export code_bytes
-  topics = "vm code_stream"
 
-type
+type CodeStream* = object
-  CodeStream* = ref object
+  code: CodeBytesRef
-    bytes*: seq[byte]
-    depthProcessed: int
-    invalidPositions: seq[byte] # bit seq of invalid jump positions
   pc*: int
 
-proc `$`*(b: byte): string =
+func init*(T: type CodeStream, code: CodeBytesRef): T =
-  $(b.int)
+  T(code: code)
 
-template bitpos(pos: int): (int, byte) =
+func init*(T: type CodeStream, code: sink seq[byte]): T =
-  (pos shr 3, 1'u8 shl (pos and 0x07))
+  T(code: CodeBytesRef.init(move(code)))
 
-proc newCodeStream*(codeBytes: sink seq[byte]): CodeStream =
+func init*(T: type CodeStream, code: openArray[byte]): T =
-  new(result)
+  T(code: CodeBytesRef.init(code))
-  result.bytes = system.move(codeBytes)
-  result.pc = 0
-  result.invalidPositions = newSeq[byte]((result.bytes.len + 7) div 8)
-  result.depthProcessed = 0
 
-proc invalidPosition(c: CodeStream, pos: int): bool =
+func init*(T: type CodeStream, code: openArray[char]): T =
-  let (bpos, bbit) = bitpos(pos)
+  T(code: CodeBytesRef.init(code))
-  (c.invalidPositions[bpos] and bbit) > 0
 
-proc newCodeStream*(codeBytes: string): CodeStream =
+template read*(c: var CodeStream, size: int): openArray[byte] =
-  newCodeStream(codeBytes.mapIt(it.byte))
-
-proc newCodeStreamFromUnescaped*(code: string): CodeStream =
-  # from 0xunescaped
-  var codeBytes: seq[byte] = @[]
-  for z, c in code[2..^1]:
-    if z mod 2 == 1:
-      var value: int
-      discard parseHex(&"0x{code[z+1..z+2]}", value)
-      codeBytes.add(value.byte)
-  newCodeStream(codeBytes)
-
-template read*(c: CodeStream, size: int): openArray[byte] =
-  # TODO: use openArray[bytes]
   if c.pc + size - 1 < c.bytes.len:
     let pos = c.pc
     c.pc += size
-    c.bytes.toOpenArray(pos, pos + size - 1)
+    c.code.bytes.toOpenArray(pos, pos + size - 1)
   else:
     c.pc = c.bytes.len
-    c.bytes.toOpenArray(0, -1)
+    c.code.bytes.toOpenArray(0, -1)
 
-proc readVmWord*(c: var CodeStream, n: static int): UInt256 =
+func readVmWord*(c: var CodeStream, n: static int): UInt256 =
   ## Reads `n` bytes from the code stream and pads
   ## the remaining bytes with zeros.
   let result_bytes = cast[ptr array[32, byte]](addr result)
 
-  let last = min(c.pc + n, c.bytes.len)
+  let last = min(c.pc + n, c.code.bytes.len)
   let toWrite = last - c.pc
-  for i in 0 ..< toWrite : result_bytes[i] = c.bytes[last - i - 1]
+  for i in 0 ..< toWrite:
+    result_bytes[i] = c.code.bytes[last - i - 1]
   c.pc = last
 
-proc len*(c: CodeStream): int =
+func len*(c: CodeStream): int =
-  len(c.bytes)
+  len(c.code)
 
-proc next*(c: var CodeStream): Op =
+func next*(c: var CodeStream): Op =
-  if c.pc != c.bytes.len:
+  if c.pc != c.code.len:
-    result = Op(c.bytes[c.pc])
+    result = Op(c.code.bytes[c.pc])
     inc c.pc
   else:
     result = Op.Stop
@@ -86,55 +62,42 @@ iterator items*(c: var CodeStream): Op =
     yield nextOpcode
     nextOpcode = c.next()
 
-proc `[]`*(c: CodeStream, offset: int): Op =
+func `[]`*(c: CodeStream, offset: int): Op =
-  Op(c.bytes[offset])
+  Op(c.code.bytes[offset])
 
-proc peek*(c: var CodeStream): Op =
+func peek*(c: var CodeStream): Op =
-  if c.pc < c.bytes.len:
+  if c.pc < c.code.bytes.len:
-    Op(c.bytes[c.pc])
+    Op(c.code.bytes[c.pc])
   else:
     Op.Stop
 
-proc updatePc*(c: var CodeStream, value: int) =
+func updatePc*(c: var CodeStream, value: int) =
   c.pc = min(value, len(c))
 
-proc isValidOpcode*(c: CodeStream, position: int): bool =
+func isValidOpcode*(c: CodeStream, position: int): bool =
-  if position >= len(c):
+  c.code.isValidOpcode(position)
-    false
-  elif c.invalidPosition(position):
-    false
-  elif position <= c.depthProcessed:
-    true
-  else:
-    var i = c.depthProcessed
-    while i <= position:
-      var opcode = Op(c[i])
-      if opcode >= Op.Push1 and opcode <= Op.Push32:
-        var leftBound = (i + 1)
-        var rightBound = leftBound + (opcode.int - 95)
-        for z in leftBound ..< rightBound:
-          let (bpos, bbit) = bitpos(z)
-          c.invalidPositions[bpos] = c.invalidPositions[bpos] or bbit
-        i = rightBound
-      else:
-        i += 1
-    c.depthProcessed = i - 1
 
-    not c.invalidPosition(position)
+func bytes*(c: CodeStream): lent seq[byte] =
+  c.code.bytes()
 
 proc decompile*(original: CodeStream): seq[(int, Op, string)] =
   # behave as https://etherscan.io/opcode-tool
-  var c = newCodeStream(original.bytes)
+  var c = CodeStream.init(original.bytes)
   while true:
     var op = c.next
     if op >= Push1 and op <= Push32:
       result.add(
-        (c.pc - 1, op, "0x" & c.read(op.int - 95).mapIt($(it.BiggestInt.toHex(2))).join("")))
+        (
+          c.pc - 1,
+          op,
+          "0x" & c.read(op.int - 95).mapIt($(it.BiggestInt.toHex(2))).join(""),
+        )
+      )
     elif op != Op.Stop:
       result.add((c.pc - 1, op, ""))
     else:
       result.add((-1, Op.Stop, ""))
       break
 
-proc atEnd*(c: CodeStream): bool =
+func atEnd*(c: CodeStream): bool =
-  c.pc >= c.bytes.len
+  c.pc >= c.code.bytes.len

nimbus/evm/computation.nim

@@ -16,6 +16,7 @@ import
   "."/[types],
   ./interpreter/[gas_meter, gas_costs, op_codes],
   ./evm_errors,
+  ./code_bytes,
   ../common/[common, evmforks],
   ../utils/utils,
   stew/byteutils,
@@ -203,9 +204,9 @@ template selfDestruct*(c: Computation, address: EthAddress) =
   else:
     c.execSelfDestruct(address)
 
-template getCode*(c: Computation, address: EthAddress): seq[byte] =
+template getCode*(c: Computation, address: EthAddress): CodeBytesRef =
   when evmc_enabled:
-    c.host.copyCode(address)
+    CodeBytesRef.init(c.host.copyCode(address))
   else:
     c.vmState.readOnlyStateDB.getCode(address)
 
@@ -236,14 +237,14 @@ proc newComputation*(vmState: BaseVMState, sysCall: bool, message: Message,
 
   if result.msg.isCreate():
     result.msg.contractAddress = result.generateContractAddress(salt)
-    result.code = newCodeStream(message.data)
+    result.code = CodeStream.init(message.data)
     message.data = @[]
   else:
-    result.code = newCodeStream(
+    result.code = CodeStream.init(
       vmState.readOnlyStateDB.getCode(message.codeAddress))
 
 func newComputation*(vmState: BaseVMState, sysCall: bool,
-                     message: Message, code: seq[byte]): Computation =
+                     message: Message, code: CodeBytesRef): Computation =
   new result
   result.vmState = vmState
   result.msg = message
@@ -251,7 +252,7 @@ func newComputation*(vmState: BaseVMState, sysCall: bool,
   result.stack = EvmStackRef.new()
   result.returnStack = @[]
   result.gasMeter.init(message.gas)
-  result.code = newCodeStream(code)
+  result.code = CodeStream.init(code)
   result.sysCall = sysCall
 
 template gasCosts*(c: Computation): untyped =

nimbus/evm/interpreter/op_handlers/oph_envinfo.nim

@@ -176,8 +176,8 @@ proc extCodeCopyOp (k: var VmCtx): EvmResultVoid =
       cpt.gasCosts[ExtCodeCopy].m_handler(cpt.memory.len, memPos, len),
       reason = "ExtCodeCopy fee")
 
-  let codeBytes = cpt.getCode(address)
+  let code = cpt.getCode(address)
-  cpt.memory.writePadded(codeBytes, memPos, codePos, len)
+  cpt.memory.writePadded(code.bytes, memPos, codePos, len)
   ok()
 
 
@@ -194,8 +194,8 @@ proc extCodeCopyEIP2929Op (k: var VmCtx): EvmResultVoid =
                     cpt.gasEip2929AccountCheck(address)
   ? cpt.opcodeGastCost(ExtCodeCopy, gasCost, reason = "ExtCodeCopy EIP2929")
 
-  let codeBytes = cpt.getCode(address)
+  let code = cpt.getCode(address)
-  cpt.memory.writePadded(codeBytes, memPos, codePos, len)
+  cpt.memory.writePadded(code.bytes(), memPos, codePos, len)
   ok()
 
 # -----------

nimbus/graphql/ethapi.nim

@@ -543,7 +543,7 @@ proc accountCode(ud: RootRef, params: Args, parent: Node): RespResult {.apiPragm
   let acc = AccountNode(parent)
   try:
     let code = acc.db.getCode(acc.address)
-    resp(code)
+    resp(code.bytes())
   except RlpError as ex:
     err(ex.msg)
 

nimbus/rpc/p2p.nim

@@ -246,7 +246,7 @@ proc setupEthRpc*(
     let
       accDB   = stateDBFromTag(quantityTag)
       address = data.ethAddr
-    result = accDB.getCode(address)
+    result = accDB.getCode(address).bytes()
 
   template sign(privateKey: PrivateKey, message: string): seq[byte] =
     # message length encoded as ASCII representation of decimal

nimbus/transaction/call_common.nim

@@ -166,7 +166,7 @@ proc setupHost(call: CallParams): TransactionHost =
   # with the contract address.  This differs from the previous Nimbus EVM API.
   # Guarded under `evmc_enabled` for now so it doesn't break vm2.
   when defined(evmc_enabled):
-    var code: seq[byte]
+    var code: CodeBytesRef
     if call.isCreate:
       let sender = call.sender
       let contractAddress =
@@ -174,7 +174,7 @@ proc setupHost(call: CallParams): TransactionHost =
       host.msg.recipient = contractAddress.toEvmc
       host.msg.input_size = 0
       host.msg.input_data = nil
-      code = call.input
+      code = CodeBytesRef.init(call.input)
     else:
       # TODO: Share the underlying data, but only after checking this does not
       # cause problems with the database.
@@ -189,7 +189,7 @@ proc setupHost(call: CallParams): TransactionHost =
     let cMsg = hostToComputationMessage(host.msg)
     host.computation = newComputation(vmState, call.sysCall, cMsg, code)
 
-    host.code = system.move(code)
+    host.code = code
 
   else:
     if call.input.len > 0:

nimbus/transaction/evmc_dynamic_loader.nim

@@ -14,7 +14,7 @@ import
   evmc/evmc, ../config
 
 # The built-in Nimbus EVM, via imported C function.
-proc evmc_create_nimbus_evm(): ptr evmc_vm {.cdecl, importc, raises: [].}
+proc evmc_create_nimbus_evm(): ptr evmc_vm {.cdecl, importc, raises: [], gcsafe.}
 
 # Import this module to link in the definition of `evmc_create_nimbus_evm`.
 # Nim thinks the module is unused because the function is only called via

nimbus/transaction/evmc_host_glue.nim

@@ -123,26 +123,9 @@ proc evmcExecComputation*(host: TransactionHost): EvmcResult =
   let hostContext = cast[evmc_host_context](host)
   host.hostInterface = hostInterface.unsafeAddr
 
-  # Without `{.gcsafe.}:` here, the call via `vm.execute` results in a Nim
-  # compile-time error in a far away function.  Starting here, a cascade of
-  # warnings takes place: "Warning: '...' is not GC-safe as it performs an
-  # indirect call here [GCUnsafe2]", then a list of "Warning: '...' is not
-  # GC-safe as it calls '...'" at each function up the call stack, to a high
-  # level function `persistBlocks` where it terminates compilation as an error
-  # instead of a warning.
-  #
-  # It is tempting to annotate all EVMC API functions with `{.cdecl, gcsafe.}`,
-  # overriding the function signatures from the Nim EVMC module.  Perhaps we
-  # will do that, though it's conceptually dubious, as the two sides of the
-  # EVMC ABI live in different GC worlds (when loaded as a shared library with
-  # its own Nim runtime), very similar to calling between threads.
-  #
-  # TODO: But wait: Why does the Nim EVMC test program compile fine without
-  # any `gcsafe`, even with `--threads:on`?
-  {.gcsafe.}:
   result = vm.execute(vm, hostInterface.unsafeAddr, hostContext,
                         evmc_revision(host.vmState.fork.ord), host.msg,
-                          if host.code.len > 0: host.code[0].unsafeAddr
+                        if host.code.len > 0: host.code.bytes[0].unsafeAddr
                         else: nil,
                         host.code.len.csize_t)
 

nimbus/transaction/host_services.nim

@@ -218,7 +218,7 @@ proc copyCode(host: TransactionHost, address: HostAddress,
   #
   # Note, when there is no code, `getCode` result is empty `seq`.  It was `nil`
   # when the DB was first implemented, due to Nim language changes since then.
-  var code: seq[byte] = host.vmState.readOnlyStateDB.getCode(address)
+  let code = host.vmState.readOnlyStateDB.getCode(address)
   var safe_len: int = code.len # It's safe to assume >= 0.
 
   if code_offset >= safe_len.HostSize:
@@ -230,7 +230,7 @@ proc copyCode(host: TransactionHost, address: HostAddress,
     safe_len = buffer_size.int
 
   if safe_len > 0:
-    copyMem(buffer_data, code[safe_offset].addr, safe_len)
+    copyMem(buffer_data, code.bytes()[safe_offset].addr, safe_len)
   return safe_len.HostSize
 
 proc selfDestruct(host: TransactionHost, address, beneficiary: HostAddress) {.show.} =

nimbus/transaction/host_types.nim

@@ -10,7 +10,7 @@ import
   std/sets,
   stint, evmc/evmc,
   eth/common/eth_types,
-  ../evm/types
+  ../evm/[code_bytes, types]
 
 # Object `TransactionHost` represents "EVMC host" to the EVM.  "Host services"
 # manage account state outside EVM such as balance transfers, storage, logs and
@@ -59,7 +59,7 @@ type
     computation*:     Computation
     msg*:             EvmcMessage
     input*:           seq[byte]
-    code*:            seq[byte]
+    code*:            CodeBytesRef
     cachedTxContext*: bool
     txContext*:       EvmcTxContext
     depth*:           int

nimbus/utils/state_dump.nim

@@ -1,5 +1,5 @@
 # Nimbus
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or
 #    http://www.apache.org/licenses/LICENSE-2.0)
@@ -75,7 +75,7 @@ proc dumpAccount*(db: LedgerRef, acc: EthAddress): DumpAccount =
     nonce   : db.getNonce(acc),
     root    : db.getStorageRoot(acc),
     codeHash: db.getCodeHash(acc),
-    code    : db.getCode(acc),
+    code    : db.getCode(acc).bytes(),
     key     : keccakHash(acc)
   )
   for k, v in db.cachedStorage(acc):

premix/downloader.nim

@@ -11,7 +11,7 @@
 import
   std/[json, strutils],
   json_rpc/[rpcclient], httputils,
-  eth/[common, rlp], chronicles,
+  eth/common, chronicles,
   ../nimbus/utils/utils,
   ./parser
 

tests/all_tests_macro.nim

@@ -6,15 +6,12 @@
 # at your option. This file may not be copied, modified, or distributed except according to those terms.
 
 import
-  std/[times, macros, strutils, os, osproc, threadpool],
+  std/[times, macros, strutils, os, osproc],
   unittest2,
   ../nimbus/compile_info,
   ../nimbus/utils/utils
 
-export strutils, os, unittest2, osproc, threadpool
+export strutils, os, unittest2, osproc
-
-# AppVeyor may go out of memory with the default of 4
-setMinPoolSize(2)
 
 proc executeMyself(numModules: int, names: openArray[string]): int =
   let appName = getAppFilename()

tests/test_code_stream.nim

@@ -11,7 +11,7 @@ import  unittest2, sequtils,
 proc codeStreamMain*() =
   suite "parse bytecode":
     test "accepts bytes":
-      let codeStream = newCodeStream("\x01")
+      let codeStream = CodeStream.init("\x01")
       check(codeStream.len == 1)
 
 
@@ -22,14 +22,14 @@ proc codeStreamMain*() =
     #         CodeStream(code_bytes)
 
     test "next returns the correct opcode":
-      var codeStream = newCodeStream("\x01\x02\x30")
+      var codeStream = CodeStream.init("\x01\x02\x30")
       check(codeStream.next == Op.ADD)
       check(codeStream.next == Op.MUL)
       check(codeStream.next == Op.ADDRESS)
 
 
     test "peek returns next opcode without changing location":
-      var codeStream = newCodeStream("\x01\x02\x30")
+      var codeStream = CodeStream.init("\x01\x02\x30")
       check(codeStream.pc == 0)
       check(codeStream.peek == Op.ADD)
       check(codeStream.pc == 0)
@@ -40,14 +40,14 @@ proc codeStreamMain*() =
 
 
     test "stop opcode is returned when end reached":
-      var codeStream = newCodeStream("\x01\x02")
+      var codeStream = CodeStream.init("\x01\x02")
       discard codeStream.next
       discard codeStream.next
       check(codeStream.next == Op.STOP)
 
     # Seek has been dommented out for future deletion
     # test "seek reverts to original position on exit":
-    #   var codeStream = newCodeStream("\x01\x02\x30")
+    #   var codeStream = CodeStream.init("\x01\x02\x30")
     #   check(codeStream.pc == 0)
     #   codeStream.seek(1):
     #     check(codeStream.pc == 1)
@@ -56,13 +56,13 @@ proc codeStreamMain*() =
     #   check(codeStream.peek == Op.ADD)
 
     test "[] returns opcode":
-      let codeStream = newCodeStream("\x01\x02\x30")
+      let codeStream = CodeStream.init("\x01\x02\x30")
       check(codeStream[0] == Op.ADD)
       check(codeStream[1] == Op.MUL)
       check(codeStream[2] == Op.ADDRESS)
 
     test "isValidOpcode invalidates after PUSHXX":
-      var codeStream = newCodeStream("\x02\x60\x02\x04")
+      var codeStream = CodeStream.init("\x02\x60\x02\x04")
       check(codeStream.isValidOpcode(0))
       check(codeStream.isValidOpcode(1))
       check(not codeStream.isValidOpcode(2))
@@ -71,7 +71,7 @@ proc codeStreamMain*() =
 
 
     test "isValidOpcode 0":
-      var codeStream = newCodeStream(@[2.byte, 3.byte, 0x72.byte].concat(repeat(4.byte, 32)).concat(@[5.byte]))
+      var codeStream = CodeStream.init(@[2.byte, 3.byte, 0x72.byte].concat(repeat(4.byte, 32)).concat(@[5.byte]))
       # valid: 0 - 2 :: 22 - 35
       # invalid: 3-21 (PUSH19) :: 36+ (too long)
       check(codeStream.isValidOpcode(0))
@@ -86,7 +86,7 @@ proc codeStreamMain*() =
 
     test "isValidOpcode 1":
       let test = @[2.byte, 3.byte, 0x7d.byte].concat(repeat(4.byte, 32)).concat(@[5.byte, 0x7e.byte]).concat(repeat(4.byte, 35)).concat(@[1.byte, 0x61.byte, 1.byte, 1.byte, 1.byte])
-      var codeStream = newCodeStream(test)
+      var codeStream = CodeStream.init(test)
       # valid: 0 - 2 :: 33 - 36 :: 68 - 73 :: 76
       # invalid: 3 - 32 (PUSH30) :: 37 - 67 (PUSH31) :: 74, 75 (PUSH2) :: 77+ (too long)
       check(codeStream.isValidOpcode(0))
@@ -109,7 +109,7 @@ proc codeStreamMain*() =
 
 
     test "right number of bytes invalidates":
-      var codeStream = newCodeStream("\x02\x03\x60\x02\x02")
+      var codeStream = CodeStream.init("\x02\x03\x60\x02\x02")
       check(codeStream.isValidOpcode(0))
       check(codeStream.isValidOpcode(1))
       check(codeStream.isValidOpcode(2))

tests/test_helpers.nim

@@ -128,7 +128,7 @@ proc verifyStateDB*(wantedState: JsonNode, stateDB: ReadOnlyStateDB) =
       wantedBalance = UInt256.fromHex accountData{"balance"}.getStr
       wantedNonce = accountData{"nonce"}.getHexadecimalInt.AccountNonce
 
-      actualCode = stateDB.getCode(account)
+      actualCode = stateDB.getCode(account).bytes()
       actualBalance = stateDB.getBalance(account)
       actualNonce = stateDB.getNonce(account)
 

tools/t8n/transition.nim

@@ -106,7 +106,7 @@ proc envToHeader(env: EnvStruct): BlockHeader =
 proc postState(db: LedgerRef, alloc: var GenesisAlloc) =
   for accAddr in db.addresses():
     var acc = GenesisAccount(
-      code: db.getCode(accAddr),
+      code: db.getCode(accAddr).bytes(),
       balance: db.getBalance(accAddr),
       nonce: db.getNonce(accAddr)
     )

vendor/nim-evmc

@@ -1 +1 @@
-Subproject commit 86d22a026b0aa07c07b3afd7d91ca475e0eae12a
+Subproject commit 6e261148565a311536b1a29f1568e8c4470baf9d