fix(rpc): decode EVM logs from tx data + align receipt/nonce fields with cosmos/evm v0.6.0 by puneet2019 · Pull Request #292 · mocachain/moca

puneet2019 · 2026-06-29T19:04:31Z

What & why

Part of the cosmos/evm v0.6.0 migration review (#278). Fixes BLOCKER 2 plus the three grouped RPC concerns, all verified against cosmos/evm@v0.6.0 rpc/backend/.

BLOCKER 2 — `eth_getLogs` / receipt `logs` / `logsBloom` / WS log subscriptions always empty

In cosmos/evm v0.6.0, AttributeKeyTxLog is defined but emitted nowhere — EVM logs now live in the tx response Data (a marshalled MsgEthereumTxResponse). The migration kept the old event-parsing extraction and merely repointed EventTypeTxLog → EventTypeEthereumTx, so every log consumer saw empty results (breaking wallets, indexers, The Graph, DeFi).

Fix (mirrors upstream rpc/backend/utils.go + tx_info.go):

GetLogsFromBlockResults → evmtypes.DecodeTxLogs(txResult.Data, height) (one entry per tx).
Receipt log path (tx_info.go) and eth_getTransactionLogs (eth namespace) → evmtypes.DecodeMsgLogs(txResult.Data, msgIndex, height).
Removed the now-dead helpers AllTxLogsFromEvents / TxLogsFromEvents / ParseTxLogsFromEvent.

RPC concerns (grouped here per the review's PR plan)

Receipt signer (tx_info.go): branch like upstream — protected → LatestSignerForChainID(ethTx.ChainId()), unprotected → ethtypes.FrontierSigner{} (their ChainId() is 0, so a chain-id signer fails to recover the sender). Removes the prior unconditional LatestSignerForChainID(b.ChainID()).
effectiveGasPrice (tx_info.go): now set on all receipt types (legacy/access-list = tx.GasPrice(); dynamic-fee = baseFee + min(gasTipCap, gasFeeCap - baseFee)), matching geth/upstream RPCMarshalReceipt. Previously only dynamic-fee receipts carried it. (Also fixes the empty-logs placeholder type [][]*Log{} → []*Log{}.)
getAccountNonce pending signer (utils.go): use the precomputed b.chainID instead of the redundant b.ChainID() RPC, which — before the EIP-155 fork height — returns an error and makes us silently skip counting pending txs. Mirrors upstream, which signs with b.EvmChainID here.

Note: b.chainID is moca's de-facto signing chain id throughout the RPC layer (call_tx / sign_tx / tracing). The separate cosmos-vs-EVM chain-id default question is tracked in #288 and is out of scope here.

Added types.SafeUint64 (mirrors the existing types.SafeInt64) for the int64 → uint64 height conversion that upstream does via utils.SafeUint64.

Tests

Unit (rpc/backend, rpc/types):

TestGetLogsFromBlockResults (utils_test.go): crafts a ResultBlockResults whose TxsResults[i].Data is a marshalled MsgEthereumTxResponse with a log; asserts the decoded logs are non-empty and correct (address / data / blockNumber / txHash), and that empty Data yields empty logs without error.
TestGetTransactionReceiptFields (tx_info_test.go): directly asserts the receipt fields this PR fixes — populated logs / logsBloom, recovered from, effectiveGasPrice on a legacy receipt (== gasPrice), status, and type.
DecodeMsgLogs index-selection coverage for the Cosmos-msg-index → EVM-response-index mapping (the mixed [non-EVM, MsgEthereumTx] case from review); the all-EVM-tx invariant is documented.
Updated TestGetLogs + mock (RegisterBlockResultsWithEventLog) to carry logs in tx Data (the cosmos/evm v0.6.0 form) instead of events.

E2E (e2e/kind/tests/test_rpc_suite.sh, against a live kind chain):

eth_newFilter / eth_getFilterChanges rehydration + eth_getLogs.
eth_subscribe("logs") over the WebSocket transport (first-party go-ethereum client, e2e/kind/tools/wslogs).
Both log paths assert the returned log carries the finalized block context this PR restores — transactionHash (== the emitting tx), blockHash (== the finalized block), and a non-zero blockTimestamp — so the tests fail on the pre-fix behavior, not just on address/topics.

Scope

Fixes BLOCKER 2 + the RPC concerns from #278. Does not touch BLOCKER 1 (precompile value guard), BLOCKER 3 (feemarket min_gas_price), or the storage/nit cleanups — those are separate stacked PRs.

🤖 Generated with Claude Code

phenix3443

One blocker remains in the new log-decoding path.

DecodeMsgLogs is being called with res.MsgIndex, which is the message index in the full Cosmos tx. The decoder, however, indexes into the filtered MsgEthereumTxResponse list stored in TxMsgData. Those only line up when the EVM message is also at Cosmos message position 0. For mixed Cosmos txs with a non-EVM message before the EVM message, this now errors and receipt / eth_getTransactionLogs logs become empty again.

Please remap that index to the EVM-response position (or copy the exact upstream logic) and add a regression test for a tx shaped like [non-EVM msg, MsgEthereumTx].

phenix3443

The previous MsgIndex concern is resolved, but I still see one blocking issue and one remaining gap.

Blocking:

golangci is still failing on new issues introduced in rpc/backend/utils_test.go: two gosec G115 findings for the int64 -> uint64 conversions and one misspell (marshalled). This needs to be fixed before the PR is merge-ready.

Non-blocking but still worth tightening:

The new decoder/unit tests are good, but TestGetTransactionReceipt still does not assert the receipt fields this PR claims to fix (logs, logsBloom, signer recovery, effectiveGasPrice). Adding one direct receipt-path assertion would make this much safer to maintain.

…mos/evm v0.6.0 cosmos/evm v0.6.0 no longer emits per-log cosmos events (AttributeKeyTxLog is defined but emitted nowhere); EVM logs are carried in the tx response Data as a marshalled MsgEthereumTxResponse. The migration kept the old event-parsing extraction (just repointing EventTypeTxLog -> EventTypeEthereumTx), so eth_getLogs, receipt logs/logsBloom, and WS log subscriptions were always empty. - GetLogsFromBlockResults: decode via evmtypes.DecodeTxLogs(txResult.Data, height) - Receipt + eth_getTransactionLogs per-msg path: evmtypes.DecodeMsgLogs(Data, msgIndex, height) - Remove now-dead event-parsing helpers (AllTxLogsFromEvents/TxLogsFromEvents/ParseTxLogsFromEvent) Also align three RPC concerns with cosmos/evm v0.6.0: - receipt signer: chain-id signer for protected txs, FrontierSigner for unprotected - effectiveGasPrice: set on all receipt types (legacy/access-list = GasPrice()) - getAccountNonce: use the precomputed b.chainID instead of the redundant b.ChainID() RPC that silently skips pending-tx counting pre-EIP155 Add types.SafeUint64 and a unit test for GetLogsFromBlockResults; update the TestGetLogs mock to carry logs in tx Data. Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

…nvariant Addresses review feedback on #292. A MsgEthereumTx is only ever included in an all-EVM tx: routing into the EVM ante requires the ExtensionOptionsEthereumTx option, the EVM ante decorators reject any non-EVM message, and the cosmos ante's RejectMessagesDecorator rejects MsgEthereumTx elsewhere. So res.MsgIndex equals the EVM message's position in the filtered MsgEthereumTxResponse slice that DecodeMsgLogs indexes into -- matching cosmos/evm v0.6.0 rpc/backend. - Add TestDecodeMsgLogsByMsgIndex covering the only multi-message shape moca permits (two MsgEthereumTx): MsgIndex 0 and 1 each select the correct response's logs. - Document the invariant at both decode sites (receipt + eth_getTransactionLogs). Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

The buildLogTxResultData helper cast the height int64 to uint64 directly (twice), tripping gosec G115, and its doc comment misspelled "marshalled". Convert height once via mocatypes.SafeUint64 — mirroring the production rpc/backend/utils.go convention — and fix the spelling, so golangci is green.

eth_subscribe("logs") and eth_getFilterChanges delivered logs with correct content (address/topics/data) but zeroed TxHash, BlockHash and BlockTimestamp: they decoded logs from the live CometBFT tx event Data, which does not carry the tx/block context (BlockHash is not known at execution time — it is backfilled into the finalized block result). Source each tx's logs from the finalized block result instead — via GetLogsByHeight (filter API) / GetLogsFromBlockResults (websockets), indexed by the tx's block position — matching the eth_getLogs path. Subscribers and pollers now receive fully-populated logs that correlate to their tx and block, mirroring cosmos/evm v0.6.0 subscription semantics. Verified live: eth_subscribe and eth_getFilterChanges return the canonical txHash and blockHash (matches eth_getLogs and the block header).

…langci gosec G115)

phenix3443 · 2026-07-02T02:08:29Z

Re-reviewed this against the current diff and the matching cosmos/evm v0.6.0 paths. I did not find a blocking issue in the tx-data log decoding, receipt field alignment, or the MsgIndex-based log selection.

One residual risk remains around end-to-end coverage for the live subscription paths (eth_subscribe("logs") / eth_getFilterChanges), since they now rehydrate logs from finalized block results instead of the tx event payload. I do not see a concrete bug there in the current implementation, but that is still the area I would watch most closely.

staticcheck QF1008 flagged the promoted-field accesses added with the EVM log-rehydration handlers (eth_subscribe logs, eth_getFilterChanges): dataTx is a tmtypes.EventDataTx which embeds abci.TxResult, so use the promoted dataTx.Height / dataTx.Index directly. No behavior change. Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

Deploy TestERC20, open a logs filter, emit a Transfer, and assert the receipt logs, eth_getFilterChanges (the live NewFilter poller) and eth_getLogs all surface the decoded log from the finalized-block-result rehydration (GetLogsByHeight / GetLogsFromBlockResults) -- the path called out in review for eth_subscribe("logs") / eth_getFilterChanges. Full RPC suite green locally (9/9). Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

Resolve CHANGELOG Unreleased > Bug Fixes conflict: keep both the #292 (rpc) and #287 (payment) entries. Code (rpc/*, types/*) merged cleanly; build + rpc/types tests green. Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

phenix3443 · 2026-07-02T07:34:10Z

Re-reviewed the current diff in an isolated worktree and re-ran go test ./rpc/... ./types/... successfully.

I did not find a new blocking correctness issue in the tx-data log decoding, receipt field alignment, or the MsgIndex-based selection after the follow-up fixes.

One remaining gap I still see is test coverage:

The new e2e added in e2e/kind/tests/test_rpc_suite.sh validates eth_newFilter / eth_getFilterChanges and eth_getLogs, but it does not actually exercise the eth_subscribe("logs") websocket path in rpc/websockets.go that this PR also changed.
rpc/backend/tx_info_test.go still does not directly assert the receipt fields this PR fixes (logs, logsBloom, signer recovery, effectiveGasPrice), so the receipt-path behavior is still lightly guarded.

I’m treating those as follow-up coverage gaps rather than blockers for this PR.

Add TestGetTransactionReceiptFields, a passing receipt test that directly asserts the fields reworked in the cosmos/evm v0.6.0 migration: logs decoded from the tx response Data, the derived logsBloom, GetSenderLegacy sender recovery, effectiveGasPrice (now populated on legacy receipts), and status/type. Previously only a failing NotEqual case existed. Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

Add e2e/kind/tools/wslogs (a first-party go-ethereum SubscribeFilterLogs client; no external ws CLI) and test_evm_ws_log_subscription: subscribe over WS, emit a Transfer, assert the pushed log -- exercising the rpc/websockets.go subscribeLogs push path that the eth_getFilterChanges/eth_getLogs HTTP tests don't reach. Adds a gated Go setup step to the RPC-suite CI job for the host-side tool build. Validated 10/10 in a local kind run. Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

@v5

…-tag) Semgrep's 'Scan' gate blocks newly-introduced mutable action tags on the PR diff. Pin the RPC-suite job's setup-go to the v5.6.0 commit SHA (identical to what @v5 resolves to today) to clear the finding while keeping the explicit Go setup the WS e2e's host-side build relies on. Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

fixed all issues

Delete moca's ~16k-LOC forked rpc/ tree and serve the EVM JSON-RPC from github.com/cosmos/evm/rpc (v0.6.0): backend, namespaces, and the newHeads/logs/pendingTx subscriptions now come from cosmos/evm, fed by an in-process CometBFT event stream and an ante PendingTxListener hook on *Moca. The indexer moves to cosmos/evm's server/types.TxResult. The only retained moca RPC code is a thin server/websockets.go shim: a copy of cosmos/evm's WS server with newHeads emitting the canonical CometBFT block hash (moca #232 / issue #229) instead of the derived eth-header hash, completing the half-fix in cosmos/evm#725 (the canonical hash is carried on the stream but unused in newHeads upstream). Removable once that lands and we bump. Also drop now-unused config keys (query-timeout, getlogs-rate-limit, getlogs-burst-limit, fix-revert-gas-refund-height) and the dead ethermint.types.v1.TxResult / EVMTxIndexer types; add json-rpc.ws-origins. Depends on #292 (WS/getFilterChanges e2e gate); merge after #292. Verified: go build ./... green; e2e RPC suite 10/10 (incl. eth_subscribe logs). Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

phenix3443 · 2026-07-02T17:01:30Z

Re-reviewed the latest head (c8f9c352) and re-ran go test ./rpc/... ./types/... plus go build ./e2e/kind/tools/wslogs in an isolated worktree.

I do not see a new blocking correctness issue in the current code, but I do still see one important coverage gap in the new e2e additions:

test_evm_ws_log_subscription now exercises the real eth_subscribe("logs") transport, which is good, but it only asserts address and topics[0].
test_evm_log_subscription does the same for eth_getFilterChanges.

The bug this PR fixed for the subscription/filter paths was not that address/topics/data were wrong; it was that the pushed logs were missing finalized block context (txHash, blockHash, blockTimestamp) because they were sourced from the live tx event payload instead of the finalized block result.

So with the current assertions, the pre-fix broken behavior could still satisfy both new tests. I would strongly prefer at least one assertion in each path that the returned log carries canonical finalized context, e.g. non-zero / expected txHash and blockHash (and optionally blockTimestamp).

phenix3443 · 2026-07-03T04:10:05Z

Re-reviewed the PR again on the current head (c8f9c352); there do not appear to be any new commits since the last pass.

I still do not see a new correctness bug in the implementation itself, but the same coverage gap remains in the new e2e tests:

test_evm_ws_log_subscription only asserts address and topics[0].
test_evm_log_subscription / eth_getFilterChanges does the same.

That is weaker than the bug this PR actually fixed. The regression here was not that address/topics/data were wrong; it was that the subscription/filter logs were missing finalized tx/block context (txHash, blockHash, blockTimestamp) because they were being sourced from the live tx event payload instead of the finalized block result.

So the current tests could still stay green against the old broken behavior. I still strongly recommend asserting at least canonical txHash and blockHash in both paths (and optionally blockTimestamp) so the new coverage matches the actual fix.

The WS (eth_subscribe logs) and getFilterChanges e2e tests asserted only address + topics[0], which the pre-fix behavior also satisfied. Assert the returned log carries the finalized block context this PR restores — transactionHash (== the mint tx), blockHash (== the finalized block), and a non-zero blockTimestamp — so the tests now fail on the pre-fix code. Closes the review coverage gap. Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

Delete moca's ~16k-LOC forked rpc/ tree and serve the EVM JSON-RPC from github.com/cosmos/evm/rpc (v0.6.0): backend, namespaces, and the newHeads/logs/pendingTx subscriptions now come from cosmos/evm, fed by an in-process CometBFT event stream and an ante PendingTxListener hook on *Moca. The indexer moves to cosmos/evm's server/types.TxResult. The only retained moca RPC code is a thin server/websockets.go shim: a copy of cosmos/evm's WS server with newHeads emitting the canonical CometBFT block hash (moca #232 / issue #229) instead of the derived eth-header hash, completing the half-fix in cosmos/evm#725 (the canonical hash is carried on the stream but unused in newHeads upstream). Removable once that lands and we bump. Also drop now-unused config keys (query-timeout, getlogs-rate-limit, getlogs-burst-limit, fix-revert-gas-refund-height) and the dead ethermint.types.v1.TxResult / EVMTxIndexer types; add json-rpc.ws-origins. Depends on #292 (WS/getFilterChanges e2e gate); merge after #292. Verified: go build ./... green; e2e RPC suite 10/10 (incl. eth_subscribe logs). Signed-off-by: puneetmahajan <59960662+puneet2019@users.noreply.github.com>

github-actions Bot added tests types labels Jun 29, 2026

puneet2019 force-pushed the fix/rpc-evm-logs branch 2 times, most recently from 82616a8 to 041dd8f Compare June 30, 2026 04:24

puneet2019 requested a review from phenix3443 June 30, 2026 04:55

Base automatically changed from chore/cosmos-evm-add to main June 30, 2026 05:10

puneet2019 requested a review from a team June 30, 2026 05:10

phenix3443 requested changes Jun 30, 2026

View reviewed changes

Comment thread rpc/backend/tx_info.go

phenix3443 temporarily deployed to release June 30, 2026 07:27 — with GitHub Actions Inactive

puneet2019 force-pushed the fix/rpc-evm-logs branch from 4257ae3 to ca2108c Compare June 30, 2026 08:17

puneet2019 temporarily deployed to release June 30, 2026 08:17 — with GitHub Actions Inactive

puneet2019 temporarily deployed to release June 30, 2026 13:11 — with GitHub Actions Inactive

phenix3443 previously requested changes Jun 30, 2026

View reviewed changes

Comment thread rpc/backend/utils_test.go Outdated

puneet2019 requested a review from phenix3443 July 1, 2026 05:21

puneet2019 force-pushed the fix/rpc-evm-logs branch from bc993aa to 606cfd8 Compare July 1, 2026 05:24

puneet2019 temporarily deployed to release July 1, 2026 05:24 — with GitHub Actions Inactive

puneet2019 temporarily deployed to release July 1, 2026 06:06 — with GitHub Actions Inactive

puneet2019 temporarily deployed to release July 1, 2026 10:30 — with GitHub Actions Inactive

puneet2019 added 5 commits July 1, 2026 22:28

test(rpc): assert expected BlockNumber to drop int64->uint64 cast (go…

d3230ef

…langci gosec G115)

puneet2019 force-pushed the fix/rpc-evm-logs branch from 9c09c1c to d3230ef Compare July 1, 2026 18:34

puneet2019 temporarily deployed to release July 1, 2026 18:34 — with GitHub Actions Inactive

puneet2019 added 2 commits July 2, 2026 09:07

puneet2019 temporarily deployed to release July 2, 2026 06:30 — with GitHub Actions Inactive

puneet2019 added 2 commits July 2, 2026 12:39

puneet2019 temporarily deployed to release July 2, 2026 08:39 — with GitHub Actions Inactive

github-actions Bot added the CI label Jul 2, 2026

puneet2019 temporarily deployed to release July 2, 2026 10:33 — with GitHub Actions Inactive

puneet2019 mentioned this pull request Jul 2, 2026

refactor(rpc)!: replace in-tree JSON-RPC with cosmos/evm/rpc #309

Merged

puneet2019 temporarily deployed to release July 3, 2026 06:17 — with GitHub Actions Inactive

phenix3443 approved these changes Jul 3, 2026

View reviewed changes

puneet2019 merged commit 8ad01d8 into main Jul 3, 2026
31 checks passed

puneet2019 deleted the fix/rpc-evm-logs branch July 3, 2026 08:29

puneet2019 mentioned this pull request Jul 8, 2026

chore(types,utils): delete dead ethermint-era helpers #320

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(rpc): decode EVM logs from tx data + align receipt/nonce fields with cosmos/evm v0.6.0#292

fix(rpc): decode EVM logs from tx data + align receipt/nonce fields with cosmos/evm v0.6.0#292
puneet2019 merged 12 commits into
mainfrom
fix/rpc-evm-logs

puneet2019 commented Jun 29, 2026 •

edited

Loading

Uh oh!

phenix3443 left a comment

Uh oh!

Uh oh!

phenix3443 left a comment

Uh oh!

Uh oh!

phenix3443 commented Jul 2, 2026

Uh oh!

phenix3443 commented Jul 2, 2026

Uh oh!

phenix3443 commented Jul 2, 2026

Uh oh!

phenix3443 commented Jul 3, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

puneet2019 commented Jun 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What & why

BLOCKER 2 — eth_getLogs / receipt logs / logsBloom / WS log subscriptions always empty

RPC concerns (grouped here per the review's PR plan)

Tests

Scope

Uh oh!

phenix3443 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

phenix3443 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

phenix3443 commented Jul 2, 2026

Uh oh!

phenix3443 commented Jul 2, 2026

Uh oh!

phenix3443 commented Jul 2, 2026

Uh oh!

phenix3443 commented Jul 3, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

puneet2019 commented Jun 29, 2026 •

edited

Loading

BLOCKER 2 — `eth_getLogs` / receipt `logs` / `logsBloom` / WS log subscriptions always empty