fix(giga): write receipt for state-transition errors that bump the nonce (CON-256) by wen-coding · Pull Request #3383 · sei-protocol/sei-chain

wen-coding · 2026-05-04T23:19:08Z

Invariant

A receipt is written for an EVM tx iff that tx bumped the sender's nonce. Equivalently: any tx the chain treats as having advanced the sender's account state — and only those — is observable via eth_getTransactionByHash / eth_getTransactionReceipt. Txs included in a block but rejected before the nonce bump (e.g. ante-handler failures, pre-execution state-transition errors) leave no receipt and are invisible to standard EVM clients, matching go-ethereum semantics.

This PR fixes a place where the Giga executor violated the invariant: a state-transition error inside Execute() (notably EIP-7623 floor-data-gas underflow, post-Pectra) bumped the sender's nonce but produced no receipt, so eth_getTransactionReceipt returned null forever and clients that polled for it hung. The invariant is now enforced symmetrically across V2 and Giga.

How the invariant is enforced

V2 path (no change in this PR — already correct):

BasicDecorator.AnteHandle registers a WithDeliverTxCallback (x/evm/ante/basic.go:33-39) that fires after DeliverTx regardless of success/failure. The callback bumps the sender's nonce and calls SetNonceBumped.
After msgServer.EVMTransaction returns err, the SDK populates ExecTxResult.Code != 0 and ExecTxResult.Log = err.Error().
At EndBlock, GetAllEVMTxDeferredInfo (x/evm/keeper/deferred.go:24) synthesizes a DeferredInfo{TxHash, Error: txRes.Log} for each tx that didn't append its own.
The EndBlock loop (x/evm/keeper/abci.go:100-113) writes a synthetic receipt for entries with Error != "" — gated on GetNonceBumped, which is exactly the invariant. Existing coverage: TestEndBlock_NoReceiptForNonceMismatch and TestEndBlock_ReceiptCreatedWhenNonceBumped in x/evm/keeper/abci_test.go.

Giga path (fixed in this PR):

executeEVMTxWithGigaExecutor already explicitly bumps the sender's nonce in its execErr != nil branch (app/app.go:1887).
It also calls AppendToEvmTxDeferredInfo for surplus tracking — but with Error == "", so the EndBlock synthetic-receipt loop above skips it.
Fix: call WriteReceipt directly in the err-branch, gated on the explicit nonce bump it already does. EIP-1559 effective gas price is used (not GasFeeCap) so the receipt's EffectiveGasPrice matches go-ethereum semantics for dynamic-fee txs.

Tests

TestGiga_FailedExecution_ProducesReceipt (giga/tests/giga_test.go) drives the EIP-7623 floor-data-gas-underflow case through the Giga executor, asserts the explicit nonce bump (nonceBefore+1 == nonceAfter) to lock in the invariant, and asserts the receipt was written with status=0, gasUsed=gasLimit, and the floor-data-gas error in VmError.

Things done

Giga path (app/app.go): write receipt in execErr != nil branch using EIP-1559 effective gas price
Giga test: nonce-bump assertion + receipt assertion
gofmt -s clean

🤖 Generated with Claude Code

Tx that's included in a block must produce an EVM receipt. The msg_server err-not-VM-error branch (V2: x/evm/keeper/msg_server.go; Giga: app/app.go executeEVMTxWithGigaExecutor) used to drop the receipt and log only — leaving eth_getTransactionByHash and eth_getTransactionReceipt returning null forever for included-but-failed txs. Write a status=0 receipt with gasUsed=gasLimit instead. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

github-actions · 2026-05-04T23:20:07Z

The latest Buf updates on your PR. Results from workflow Buf / buf (pull_request).

Build	Format	Lint	Breaking	Updated (UTC)
`✅ passed`	`✅ passed`	`✅ passed`	`✅ passed`	May 6, 2026, 10:26 PM

codecov · 2026-05-04T23:23:18Z

Codecov Report

❌ Patch coverage is 66.66667% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 59.06%. Comparing base (ab54587) to head (a6ad8e1).

Files with missing lines	Patch %	Lines
app/app.go	66.66%	4 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3383      +/-   ##
==========================================
+ Coverage   59.04%   59.06%   +0.01%     
==========================================
  Files        2105     2099       -6     
  Lines      173288   172779     -509     
==========================================
- Hits       102318   102049     -269     
+ Misses      62091    61873     -218     
+ Partials     8879     8857      -22

Flag	Coverage Δ
sei-chain-pr	`58.99% <66.66%> (?)`
sei-db	`70.41% <ø> (ø)`

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
app/app.go	`70.30% <66.66%> (+0.61%)`	⬆️

... and 66 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

Adds TestGiga_FailedExecution_ProducesReceipt — the Giga-path counterpart to TestEVMTransactionStateTransitionErrorProducesReceipt. Triggers the EIP-7623 floor-data-gas check inside go-ethereum's Execute() (intrinsic gas passes EvmStatelessChecks but floor data gas fails inside Execute()), then asserts the transient receipt store contains a status=0 receipt with gasUsed=gasLimit and a populated VmError. Without the app.go fix, the receipt is dropped and the assertion fails with "receipt not found"; with the fix it passes. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Mirror the success-branch evmMsg construction (PR #3384): use effectiveGasPrice (computed at line 1866) for receipt's EffectiveGasPrice field instead of ethTx.GasPrice() which returns GasFeeCap for dynamic-fee txs. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

LegacyTx had GasPrice == GasFeeCap == GasTipCap, so the EffectiveGasPrice assertion would pass equally whether the receipt stored maxFee or the EIP-1559 effective price. Switch to a DynamicFeeTx where tip < cap so the assertion discriminates: if anyone reverts the err-branch evmMsg to hand-roll with ethTx.GasPrice() (returns GasFeeCap for dynamic-fee txs), the assertion catches it. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

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

V2's existing EndBlock synthetic-receipt path (x/evm/keeper/abci.go:100) already writes a receipt for state-transition errors: the GetAllEVMTxDeferredInfo fallback synthesizes a DeferredInfo from txRes.Log when none was appended, and EndBlock's loop gates the synthetic-receipt write on GetNonceBumped. BasicDecorator's WithDeliverTxCallback bumps the nonce + calls SetNonceBumped on every DeliverTx (including failures), so the rule "receipt iff the tx bumped the sender's nonce" already holds for V2. The explicit WriteReceipt added to msg_server.go was redundant and also bypassed the GetNonceBumped gate, so it's removed. The Giga path is still missing the receipt because its AppendToEvmTxDeferredInfo call in the err-branch doesn't propagate the Error string, so EndBlock's `if deferredInfo.Error != ""` check fails and skips the synthetic-receipt write. Giga's explicit WriteReceipt in app.go is kept and is gated on the explicit nonce bump it already does at app.go:1887. Both tests are extended with nonce-before/after assertions to lock in the invariant. The V2 test is rewritten to drive the full production sequence (BasicDecorator -> msgServer -> deliverTxCallback -> SetMsgs/SetTxResults -> EndBlock) so it exercises the synthetic path end-to-end for an EIP-7623 floor-data-gas-underflow tx. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

V2's msg_server.go is unchanged on this branch (the EndBlock synthetic path already handles state-transition errors), so the V2 test added earlier — useful as the experiment that proved the V2 path works for floor-data-gas-underflow — is out of scope for a Giga-only PR. The existing tests in x/evm/keeper/abci_test.go (TestEndBlock_NoReceipt ForNonceMismatch, TestEndBlock_ReceiptCreatedWhenNonceBumped) already cover the EndBlock synthetic-receipt mechanism and its GetNonceBumped gate. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

arajasek · 2026-05-06T20:17:55Z

+// Without the fix in app.go's executeEVMTxWithGigaExecutor, the receipt was dropped
+// for these "should not happen" cases, so eth_getTransactionReceipt returned null
+// forever for an included tx, hanging any client that polls. This is the Giga-path
+// counterpart to TestEVMTransactionStateTransitionErrorProducesReceipt in


nit: where is TestEVMTransactionStateTransitionErrorProducesReceipt? i don't see it in the file mentioned.

Good point, changed.

PR review feedback (arajasek): the doc comment on TestGiga_FailedExecution_ProducesReceipt still pointed at TestEVMTransactionStateTransitionErrorProducesReceipt as the V2-side counterpart, but the prior commit (a889125) deleted that test along with the redundant V2 WriteReceipt. Replace the stale reference with the actual reason V2 doesn't need a test: EndBlock's synthetic-receipt path (GetAllEVMTxDeferredInfo + GetNonceBumped) already covers V2 state-transition errors. Giga doesn't because its execErr branch passes an empty Error string to AppendToEvmTxDeferredInfo, so the EndBlock fallback never fires for it — hence the explicit WriteReceipt this test guards. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

codchen · 2026-05-07T05:54:24Z

 		// Match V2 error handling: bump nonce, commit fee deduction, track surplus
 		stateDB.SetNonce(sender, stateDB.GetNonce(sender)+1, tracing.NonceChangeEoACall)
 		surplus, ferr := stateDB.Finalize()
 		if ferr != nil {


I don't think we want to create a receipt under this scenario, although it is practically impossible (or at least not expected) to have an error here.

wen-coding added the non-app-hash-breaking label May 4, 2026

wen-coding changed the title ~~evm: write status=0 receipt for state-transition errors~~ evm: write status=0 receipt for state-transition errors (CON-256) May 4, 2026

wen-coding mentioned this pull request May 5, 2026

giga: report EIP-1559 effective gas price on receipt (CON-256) #3384

Merged

3 tasks

wen-coding and others added 6 commits May 4, 2026 17:16

Drop PR cross-reference from inline comment

b540480

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

gofmt -s test file

118004f

Merge branch 'main' into wen/write_evm_receipt_for_failed_tx

e038c7a

wen-coding changed the title ~~evm: write status=0 receipt for state-transition errors (CON-256)~~ fix(giga): write receipt for state-transition errors that bump the nonce (CON-256) May 6, 2026

arajasek approved these changes May 6, 2026

View reviewed changes

wen-coding and others added 2 commits May 6, 2026 15:16

Merge branch 'main' into wen/write_evm_receipt_for_failed_tx

a6ad8e1

codchen approved these changes May 7, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(giga): write receipt for state-transition errors that bump the nonce (CON-256)#3383

fix(giga): write receipt for state-transition errors that bump the nonce (CON-256)#3383
wen-coding wants to merge 11 commits intomainfrom
wen/write_evm_receipt_for_failed_tx

wen-coding commented May 4, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented May 4, 2026 •

edited

Loading

Uh oh!

codecov Bot commented May 4, 2026 •

edited

Loading

Uh oh!

arajasek May 6, 2026

Uh oh!

wen-coding May 6, 2026

Uh oh!

codchen May 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

wen-coding commented May 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Invariant

How the invariant is enforced

Tests

Things done

Uh oh!

github-actions Bot commented May 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov Bot commented May 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

arajasek May 6, 2026

Choose a reason for hiding this comment

Uh oh!

wen-coding May 6, 2026

Choose a reason for hiding this comment

Uh oh!

codchen May 7, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

wen-coding commented May 4, 2026 •

edited

Loading

github-actions Bot commented May 4, 2026 •

edited

Loading

codecov Bot commented May 4, 2026 •

edited

Loading