execution/tests: update eest fixtures to stable v5.4.0 by taratorio · Pull Request #20541 · erigontech/erigon

taratorio · 2026-04-14T08:23:00Z

Four commits, all in service of moving Erigon's EEST integration from v5.3.0 to v5.4.0:

execution/tests/execution-spec-tests (submodule bump)

Bumps the pinned fixtures submodule from v5.3.0 to v5.4.0. Pure submodule SHA change — the fixtures themselves are generated by ethereum/execution-spec-tests.

.github/workflows/test-hive-eest.yml

Sets EEST_VERSION: v5.4.0 so the Hive EEST CI job downloads the v5.4.0 fixtures tarball that matches the submodule.

execution/protocol/rules/ethash/rules.go (header-validation fix)

Removes the early return nil short-circuit in Ethash.VerifyHeader that skipped validation when the header was already present in the chain reader:

     // Short circuit if the header is known, or its parent not
     if chain.GetHeader(header.Hash(), number) != nil {
             return nil
     }

This skip was safe under the old stage_headers pipeline, which performed header validation before persisting. Now that stage_headers is gone and the live flow is insertPosBlocks, blocks are persisted
first and validated later in ValidateChain. With the old skip, already-persisted pre-merge headers would be treated as "known" and return nil without being validated, causing v5.4.0 frontier/berlin
tests to accept otherwise-invalid blocks.

execution/protocol/txn_executor.go (tx validation ordering fix)

      if st.gp != nil && st.msg.Gas() > st.gp.Gas() {
              return ErrGasLimitReached
      }

Added in preCheck immediately before the London fee-cap block. The new v5.4.0 test_tx_gas_limit case (PR ethereum/execution-spec-tests#1731) builds a tx whose gas_limit exceeds the block gas_limit and
has gas_price < base_fee, and asserts the client surfaces GAS_ALLOWANCE_EXCEEDED. The EELS reference (cancun/fork.py::check_transaction) checks tx.gas > gas_available before the fee-cap check; Erigon
did the fee-cap check first, so the 5 post-London fork variants were failing with INSUFFICIENT_MAX_FEE_PER_GAS. The new early probe fixes the ordering. The authoritative SubGas deduction in buyGas is
unchanged.

Net result

Fixtures and CI now target v5.4.0.
Two consensus-ordering bugs that v5.4.0's expanded frontier-validation coverage exposed are fixed.
Diff stat: 4 files changed, 6 insertions, 6 deletions.

…pre-merge

…fixtures_stable_v5.4.0

taratorio · 2026-04-21T05:36:42Z

@yperbasis this is good to go

green hive eest run: https://github.com/erigontech/erigon/actions/runs/24702808441

Copilot

Pull request overview

Updates Erigon’s EEST integration from v5.3.0 to v5.4.0, aligning fixtures/CI inputs and fixing two consensus/validation ordering issues exposed by the newer EEST coverage.

Changes:

Bump EEST fixtures submodule to v5.4.0 and update the Hive EEST CI workflow to download v5.4.0 fixtures.
Ethash: remove the “known header” short-circuit so persisted headers still get validated during later validation passes.
Tx execution: check tx gas vs remaining block gas earlier to ensure “gas allowance exceeded” classification wins over London fee-cap errors when both would apply.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
execution/tests/execution-spec-tests	Submodule bump to EEST v5.4.0 fixtures.
.github/workflows/test-hive-eest.yml	Update `EEST_VERSION` to v5.4.0 for CI fixture download parity.
execution/protocol/rules/ethash/rules.go	Ensure header validation runs even if the header is already persisted/known.
execution/protocol/txn_executor.go	Adjust pre-check ordering to surface `ErrGasLimitReached` before EIP-1559 fee-cap checks when appropriate.

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

Copilot · 2026-04-21T07:47:00Z

@@ -311,6 +311,10 @@ func (st *TxnExecutor) preCheck(gasBailout bool, intrinsicGasResult mdgas.Intrin
 		}
 	}



The new early gas-pool check is intentionally redundant with buyGas()->gp.SubGas(), but the reason (error ordering vs London fee-cap checks) isn’t obvious from the code. Consider adding a short comment explaining that this check must happen before the EIP-1559 fee-cap validation to match EEST/reference-client error classification (gas allowance exceeded should win over fee-cap errors when tx.gas > gasAvailable).

Suggested change

// This early gas-pool check is intentionally redundant with the later

// buyGas()->gp.SubGas() path. It must happen before the EIP-1559 fee-cap

// validation so error classification matches EEST/reference clients:

// when tx.gas > gasAvailable, ErrGasLimitReached should win over fee-cap errors.

Copilot · 2026-04-21T07:47:01Z

+	if st.gp != nil && st.msg.Gas() > st.gp.Gas() {
+		return ErrGasLimitReached
+	}


There’s no regression unit test covering the new error-ordering behavior (tx gas > remaining block gas should return ErrGasLimitReached even if London fee-cap/baseFee checks would also fail). Since this file already has TxnExecutor-focused tests, please add a test that sets a non-zero baseFee (NoBaseFee=false), constructs a London tx/message with maxFeePerGas < baseFee and gas > gp.Gas(), and asserts ErrGasLimitReached.

yperbasis

Suggestion:

Add a one-line comment to the new probe in txn_executor.go:preCheck. The probe returns the same ErrGasLimitReached that buyGas → gp.SubGas would eventually return, so a future
reader will likely flag it as dead code and remove it. A comment like // Early probe: EELS check_transaction expects GAS_ALLOWANCE_EXCEEDED to surface before
INSUFFICIENT_MAX_FEE_PER_GAS (or similar pointing to cancun/fork.py::check_transaction) would preserve the ordering intent. Same story for rules/ethash/rules.go — optionally a brief
comment explaining why the short-circuit is gone (post-stage_headers, headers may be persisted pre-validation) would save the next archaeologist a git-blame trip.

The EIP-8037 inclusion check at TxnExecutor.preCheck (#21207) verifies that a transaction fits in the remaining block-level gas pool per dimension: regular: min(tx.gas, MaxTxnGasLimit if Amsterdam+) <= gp.RegularGasAvailable() state: tx.gas <= gp.StateGasAvailable() (Amsterdam+) This is gated on st.gp != nil. In serial exec st.gp is the block-level pool reused across all txs (exec3_serial.go:372, see #20541), so the check fires correctly. In parallel exec each worker constructs its own per-tx gas pool sized to txn.GetGasLimit() (trace_worker.go:121), so the same call from preCheck compares tx.gas against itself and never fires -- the check is effectively bypassed under parallel. Reported symptoms: EEST blocks crafted with tx.gas > block.gasLimit (GAS_ALLOWANCE_EXCEEDED expected) were not rejected by the parallel matrix in #21017. The bad tx either failed later in preCheck with the wrong sentinel (ErrFeeCapTooLow) or executed and tripped the gas-used mismatch -- in either case, EEST/Hive ExceptionMappers mis-classified the failure. Fix: 1. Extract the per-dimension check into protocol.CheckBlockGasInclusion (execution/protocol/gaspool.go). Same logic, one source of truth. A nil pool returns nil so simulation paths (eth_call / eth_simulateV1 / trace_call) -- which construct executors without a block-level pool -- are unaffected. 2. Replace the inline check in TxnExecutor.preCheck with a call to the helper. Serial behaviour is unchanged. 3. Add the parallel equivalent in exec3_parallel.go finalize, where be.gasPool (the block-level pool) is in scope. The call runs in tx-completion order against be.gasPool, immediately before the existing ConsumeRegular / ConsumeState / SubBlobGas block, so the check sees the correctly-decremented pool for each tx. Wraps rules.ErrInvalidBlock for the parallel reject, so the existing bad-block handling (POSSync ReportBadHeaderPoS, unwind, BadBlock(hash)) fires identically to other parallel-side rejects. The earlier revision of this PR added a static tx.gas > header.GasLimit check at the block-iteration layer of both exec3.go (parallel) and exec3_serial.go (serial). That approach worked but duplicated logic across the two paths and didn't follow the EIP-8037 per-dimension semantics. Replaced by the helper-based version here. Local validation: - TestSimulatedBackend_CallContractRevert PASS (regression check) - TestSimulatedBackend_PendingAndCallContract PASS - TestGeneratedTraceApiCollision PASS - go test -short ./execution/{protocol,vm,stagedsync,tests} PASS

taratorio requested review from mh0lt and yperbasis as code owners April 14, 2026 08:23

taratorio requested review from AskAlexSharov and mriccobene as code owners April 16, 2026 03:05

taratorio added 3 commits April 16, 2026 13:33

execution/tests: update eest fixtures to stable v5.4.0

a3b3748

github/workflows: update EEST_VERSION to v5.4.0 for Hive EEST CI

0b51a64

fix failing frontier/berlin tests due to header validation skips for …

a88ea19

…pre-merge

taratorio force-pushed the update_eest_fixtures_stable_v5.4.0 branch from 15ada4b to a88ea19 Compare April 16, 2026 03:34

AskAlexSharov and others added 2 commits April 16, 2026 15:00

Merge branch 'main' into update_eest_fixtures_stable_v5.4.0

5e86be3

Merge branch 'main' of github.com:erigontech/erigon into update_eest_…

ee427a3

…fixtures_stable_v5.4.0

taratorio mentioned this pull request Apr 20, 2026

[DO-NOT-MERGE] ci: use hive eest stable fixtures instead of develop #20689

Closed

taratorio added 4 commits April 20, 2026 15:16

fix eest consume-engine expecting GAS_ALLOWANCE_EXCEEDED

644f831

Merge branch 'main' of github.com:erigontech/erigon into update_eest_…

42fc4b7

…fixtures_stable_v5.4.0

point eest fixtures to v5.4.0 from main

93c472e

fix for amsterdam gaspool=nil

5f5288b

taratorio changed the title ~~[DO-NOT-MERGE] execution/tests: update eest fixtures to stable v5.4.0~~ execution/tests: update eest fixtures to stable v5.4.0 Apr 21, 2026

Merge branch 'main' into update_eest_fixtures_stable_v5.4.0

cfbd756

taratorio enabled auto-merge April 21, 2026 05:42

yperbasis requested a review from Copilot April 21, 2026 07:42

Copilot started reviewing on behalf of yperbasis April 21, 2026 07:43 View session

Copilot AI reviewed Apr 21, 2026

View reviewed changes

yperbasis approved these changes Apr 21, 2026

View reviewed changes

taratorio added this pull request to the merge queue Apr 21, 2026

Merged via the queue into main with commit adad10a Apr 21, 2026
46 checks passed

taratorio deleted the update_eest_fixtures_stable_v5.4.0 branch April 21, 2026 08:54

taratorio mentioned this pull request May 18, 2026

execution: extract EIP-8037 inclusion check, run it for parallel exec #21237

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

execution/tests: update eest fixtures to stable v5.4.0#20541

execution/tests: update eest fixtures to stable v5.4.0#20541
taratorio merged 10 commits into
mainfrom
update_eest_fixtures_stable_v5.4.0

taratorio commented Apr 14, 2026 •

edited

Loading

Uh oh!

taratorio commented Apr 21, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

yperbasis left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

		@@ -311,6 +311,10 @@ func (st *TxnExecutor) preCheck(gasBailout bool, intrinsicGasResult mdgas.Intrin
		}
		}

+	// This early gas-pool check is intentionally redundant with the later
+	// buyGas()->gp.SubGas() path. It must happen before the EIP-1559 fee-cap
+	// validation so error classification matches EEST/reference clients:
+	// when tx.gas > gasAvailable, ErrGasLimitReached should win over fee-cap errors.

Conversation

taratorio commented Apr 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

taratorio commented Apr 21, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

yperbasis left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

taratorio commented Apr 14, 2026 •

edited

Loading