Fix/bal selfdestruct netzero by mh0lt · Pull Request #19528 · erigontech/erigon

mh0lt · 2026-02-27T20:19:49Z

The PR reinstates the BlockAssembler and the move of BAL processing to the versionedio code rather than as an auxiliary set of functions in the sync process.

The main advantage of this change is that it consolidates behaviour which makes agent understanding more focussed and makes the bal code re-usable in tests. This makes the codebase more streamlined avoids breaks and confusion in test analysis.

It contains fixes merged by the claude agent from #19434 and #19525, along with fixes that get to a running bal-devnet-2 node.

This is ready for merging.

…actor - Track selfDestructedAt access index so balance writes from later txns are not incorrectly skipped in BAL computation - Fix net-zero storage write detection in BAL - Extract BlockAssembler from builderstages/exec.go into exec package - Add merge.NewFaker and Prague pre-deploy support in exec module tester - Fix EIP-7702 committed flag for BAL tracking in state_object - Integrate versionedStateReader and parallel exec determinism fixes Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Reorder AccountPath enum so SelfDestructPath precedes BalancePath, ensuring updateWrite always sets the selfDestructed flag before evaluating balance skip logic. Add sortVersionedWrites to AsBlockAccessList to guarantee deterministic write processing order regardless of Go map iteration. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

This PR reinstates the BlockAssembler abstraction and consolidates Block Access List (BAL / EIP-7928) generation/normalization into the versioned I/O pipeline, while applying a set of fixes for determinism and fork-specific behavior (Amsterdam/Prague/Cancun) across execution, block building, and engine API surfaces.

Changes:

Reintroduce a reusable BlockAssembler and route BAL capture/computation through state.VersionedIO.AsBlockAccessList() rather than stagedsync helper code.
Fix/adjust BAL encoding/decoding, determinism (sorted writes), and selfdestruct/net-zero semantics in versioned execution paths.
Update block/test generation and Engine API plumbing to support newer fork requirements (e.g., Cancun beacon root availability, transition configuration exchange).

Reviewed changes

Copilot reviewed 26 out of 26 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
node/ethconfig/config.go	Adds default `ExperimentalBAL` config value.
node/cli/flags.go	Sets explicit default value for `--experimental.bal`.
execution/types/block_access_list_test.go	Clarifies RLP encoding expectations in test comments.
execution/types/block_access_list.go	Switches BAL RLP to minimal integer encoding + matching decode helpers.
execution/tests/blockgen/chain_makers.go	Records system-call I/O for BAL and sets Cancun beacon root prior to init.
execution/state/versionmap.go	Reorders `AccountPath` enum to ensure deterministic/selfdestruct-correct processing.
execution/state/versionedio.go	Adds deterministic write sorting and improves BAL construction rules (net-zero, selfdestruct).
execution/state/state_object.go	Adjusts committed-write tracking for versioned reads when setting storage.
execution/state/journal.go	Preserves read-set entries when reverting a touch to keep validation correct.
execution/state/intra_block_state_test.go	Updates test to clear stale caches before simulating re-execution.
execution/state/intra_block_state.go	Fixes versioned read tracking behavior and codehash sync during versionmap code reads.
execution/stagedsync/exec3_serial.go	Removes/adjusts Amsterdam+serial-exec messaging and minor whitespace cleanup.
execution/stagedsync/exec3_parallel.go	Adds BAL mismatch dumping and changes BAL mismatch handling behavior.
execution/stagedsync/bal_create.go	Simplifies BAL creation to use `VersionedIO.AsBlockAccessList()`.
execution/protocol/state_transition.go	Names state refund intermediate for clarity.
execution/protocol/block_exec.go	Expands gas-used mismatch logging with per-tx diagnostics.
execution/execmodule/execmoduletester/exec_module_tester.go	Sets default `experimentalBAL` for module tester options.
execution/execmodule/block_building.go	Only bumps payload version / includes BAL fields when header carries BAL hash.
execution/exec/txtask.go	Avoids genesis conversion when `genesis == nil`.
execution/exec/block_assembler.go	Reintroduces `BlockAssembler` with BAL capture during init/tx/finalize.
execution/engineapi/interface.go	Extends EngineAPI interface with transition configuration exchange method.
execution/engineapi/engine_server_test.go	Updates tests to assert BAL presence rather than exact “empty BAL” encoding.
execution/engineapi/engine_server.go	Returns UnknownPayload when assembled response has nil ExecutionPayload.
execution/engineapi/engine_api_methods.go	Implements `ExchangeTransitionConfigurationV1`.
execution/builder/builderstages/finish.go	Only embeds BAL hash into header for Amsterdam+ blocks.
execution/builder/builderstages/exec.go	Switches builder execution to use `BlockAssembler` and re-enables forced parallel on Amsterdam.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-03T09:16:28Z

execution/stagedsync/exec3_parallel.go

+										// TEMPORARY: warn instead of error to allow sync to continue for debugging
+										log.Warn("BAL mismatch (continuing)", "block", applyResult.BlockNum, "computed", bal.Hash(), "expected", headerBALHash, "storedBAL", dbBALBytes != nil)
 									}


Continuing sync when the computed BAL hash doesn’t match the header hash breaks block validity rules for Amsterdam blocks. This should return an error (and mark the block invalid) unless explicitly gated behind a dedicated debug/devnet flag; otherwise the node may accept/serve an invalid chain state.

Fixed by: d816bea

Copilot · 2026-03-03T09:16:28Z

execution/exec/block_assembler.go

+func (ba *BlockAssembler) Initialize(ibs *state.IntraBlockState, tx kv.TemporalTx, logger log.Logger) {
+	protocol.InitializeBlockExecution(ba.cfg.Engine,
+		NewChainReader(ba.cfg.ChainConfig, tx, ba.cfg.BlockReader, logger), ba.Header, ba.cfg.ChainConfig, ibs, &state.NoopWriter{}, logger, nil)
+	if ba.HasBAL() {
+		ba.balIO = ba.balIO.Merge(ibs.TxIO())
+		ibs.ResetVersionedIO()
+	}


BlockAssembler.Initialize calls protocol.InitializeBlockExecution (which returns an error) but ignores the return value. Initialization failures (e.g., system contract init/predeploy issues) would be silently swallowed and could produce invalid blocks/state. Change Initialize to return an error and propagate it to callers (e.g., SpawnBuilderExecStage) so block building fails fast on init errors.

Fixed by: 12b1b2c

Copilot · 2026-03-03T09:16:28Z

execution/tests/blockgen/chain_makers.go

+		// Set ParentBeaconBlockRoot for Cancun+ blocks before InitializeBlockExecution
+		// so that EIP-4788 can store it during initialization.
+		if config.IsCancun(b.header.Time) {
+			var beaconBlockRoot common.Hash
+			if _, err := rand.Read(beaconBlockRoot[:]); err != nil {
+				return nil, nil, nil, fmt.Errorf("can't create beacon block root: %w", err)
+			}
+			b.header.ParentBeaconBlockRoot = &beaconBlockRoot
+		}


ParentBeaconBlockRoot is being set using crypto/rand, which makes GenerateChain non-deterministic and can cause flaky tests. Also, this file already sets ParentBeaconBlockRoot again later in block generation; adding a value here risks it being overwritten after EIP-4788 initialization, making the header inconsistent with the state. Suggest deriving a deterministic root (e.g., from block number/parent hash) and ensuring it’s assigned exactly once per block (before InitializeBlockExecution).

I don't agree with this reasoning this makes a deterministic link for non deterministic data - this causes assumed determinism bugs. The double assigment is a bug, which should be tested and reviewed.

The duplication was an error its fixed by: 3628943

yperbasis · 2026-03-03T09:34:27Z

execution/engineapi/engine_api_methods.go

 	return nil, err
 }
+
+func (e *EngineServer) ExchangeTransitionConfigurationV1(ctx context.Context, beaconConfig *engine_types.TransitionConfiguration) (*engine_types.TransitionConfiguration, error) {


exchangeTransitionConfiguration was removed a long time ago (see #12430). Why restore it?

It seems to cause test fails when hive testing - I'll investigate the issue.

Fixed by: b07260a the testing agent was using an out of data hive instance.

taratorio · 2026-03-04T01:59:07Z

.claude/skills/launch-bal-devnet-2/skill.md

@@ -0,0 +1,164 @@
+# Launch bal-devnet-2 (Erigon + Lighthouse)


this looks like a duplicated file.... a commit adds this to main already - 7c13b94

Fixed by: 6b7d81a

Claude adds 'skills.md' files locally while making skills changes, these should never be checked into ci, so I've added a .gitignore entry

The PR reinstates the BlockAssembler and the move of BAL processing to the versionedio code rather than as an auxiliary set of functions in the sync process. The main advantage of this change is that it consolidates behaviour which makes agent understanding more focussed and makes the bal code re-usable in tests. This makes the codebase more streamlined avoids breaks and confusion in test analysis. It contains fixes merged by the claude agent from #19434 and #19525, along with fixes that get to a running bal-devnet-2 node. This is ready for merging. --------- Co-authored-by: taratorio <94537774+taratorio@users.noreply.github.com> Co-authored-by: yperbasis <andrey.ashikhmin@gmail.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Andrew Ashikhmin <34320705+yperbasis@users.noreply.github.com>

…fter post block validation and before commitment validation (#19595) DO NOT MERGE BEFORE #19528 - adds the new bal-devnet-3 test fixtures and applies the skips for failing eips - moves the bals hash check after post block validation and before commitment verification so that we can clearly see the real errors (e.g. gas mismatches/receipt mismatches/etc) instead of always seeing `block access list mismatch`

The PR reinstates the BlockAssembler and the move of BAL processing to the versionedio code rather than as an auxiliary set of functions in the sync process. The main advantage of this change is that it consolidates behaviour which makes agent understanding more focussed and makes the bal code re-usable in tests. This makes the codebase more streamlined avoids breaks and confusion in test analysis. It contains fixes merged by the claude agent from #19434 and #19525, along with fixes that get to a running bal-devnet-2 node. This is ready for merging. --------- Co-authored-by: taratorio <94537774+taratorio@users.noreply.github.com> Co-authored-by: yperbasis <andrey.ashikhmin@gmail.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Andrew Ashikhmin <34320705+yperbasis@users.noreply.github.com>

…fter post block validation and before commitment validation (#19595) DO NOT MERGE BEFORE #19528 - adds the new bal-devnet-3 test fixtures and applies the skips for failing eips - moves the bals hash check after post block validation and before commitment verification so that we can clearly see the real errors (e.g. gas mismatches/receipt mismatches/etc) instead of always seeing `block access list mismatch`

taratorio and others added 15 commits February 17, 2026 14:17

enable ExperimentalBAL

abf349c

enable Exec3Parallel

63fa956

Merge branch 'main' into bal-devnet-2

adee8db

Merge branch 'main' of github.com:erigontech/erigon into bal-devnet-2

1aa09ab

Merge branch 'main' into bal-devnet-2

c798852

Merge branch 'main' into bal-devnet-2

88c6ea8

Merge branch 'main' into bal-devnet-2

ef8f3f6

Merge branch 'main' into bal-devnet-2

205806a

Merge branch 'main' into bal-devnet-2

f8ecee4

Merge branch 'main' into bal-devnet-2

f321a5f

Merge branch 'main' into bal-devnet-2

a5e976e

align test and binary defaults

af19df6

remove unwanted flags & defaults

ef26c6c

mh0lt requested review from AskAlexSharov, Giulio2002 and yperbasis as code owners February 27, 2026 20:19

mh0lt requested review from shohamc1, taratorio and yperbasis and removed request for AskAlexSharov, Giulio2002, shohamc1, taratorio and yperbasis February 27, 2026 21:58

mh0lt added the Glamsterdam https://eips.ethereum.org/EIPS/eip-7773 label Feb 27, 2026

Merge branch 'main' into fix/bal-selfdestruct-netzero

9b8cd81

yperbasis and others added 5 commits March 3, 2026 09:49

Merge branch 'main' into fix/bal-selfdestruct-netzero

364d41c

rm skill.md

560cdbe

restore SKILL.md

953adef

revert changes to SKILL.md

c0e3ec9

newline

2f98717

yperbasis requested a review from Copilot March 3, 2026 09:11

Copilot started reviewing on behalf of yperbasis March 3, 2026 09:11 View session

Copilot AI reviewed Mar 3, 2026

View reviewed changes

yperbasis requested changes Mar 3, 2026

View reviewed changes

mh0lt and others added 7 commits March 3, 2026 16:24

fix test fails

33a3623

skill updated

e2da64f

remove depricated method

b07260a

remove parent ParentBeaconBlockRoot setting

3628943

reinsert bal check

d816bea

add missing err return

12b1b2c

Merge branch 'main' into fix/bal-selfdestruct-netzero

41b35fb

taratorio reviewed Mar 4, 2026

View reviewed changes

taratorio mentioned this pull request Mar 4, 2026

execution/tests: add bal-devnet-3 fixtures and move bals hash check after post block validation and before commitment validation #19595

Merged

mh0lt and others added 3 commits March 4, 2026 07:44

Merge branch 'main' into fix/bal-selfdestruct-netzero

d1825ac

fix tests

815962a

remove redundant skill file

6b7d81a

mh0lt requested a review from yperbasis March 4, 2026 09:00

yperbasis approved these changes Mar 4, 2026

View reviewed changes

mh0lt merged commit 58dba7c into main Mar 4, 2026
24 checks passed

mh0lt deleted the fix/bal-selfdestruct-netzero branch March 4, 2026 09:06

yperbasis mentioned this pull request Mar 4, 2026

bal-devnet-2 implementation #18680

Closed

		@@ -0,0 +1,164 @@
		# Launch bal-devnet-2 (Erigon + Lighthouse)

Conversation

mh0lt commented Feb 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mh0lt Mar 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

mh0lt commented Feb 27, 2026 •

edited

Loading

mh0lt Mar 4, 2026 •

edited

Loading