fix(ccip/sui): link upgraded offramp against upgraded ccip pkg#22016
Draft
fix(ccip/sui): link upgraded offramp against upgraded ccip pkg#22016
Conversation
upgradeSuiOffRamp compiled offramp v2 with the original ccip pkg as its `ccip` dep, so the published linkage pointed at fee_quoter v2. After BlockVersion(fee_quoter, v=2), commit reports with price_updates reverted inside offramp.commit → fee_quoter::update_prices → verify_function_allowed, producing zero CommitReportAccepted events and flaking Test_CCIP_Upgrade_EVM2Sui on roughly a price-update-cadence basis.
Contributor
|
✅ No conflicts with other open PRs targeting |
…v3-routed dep Previous attempt set ccip's published-at to CCIPMockV2PackageId without overriding original_ccip_pkg, so managePackage wrote Published.toml with original-id = v3 = published-at. Sui's PublishUpgradeMissingDependency check rejects because the on-chain original-id of CCIPMockV2PackageId is actually CCIPAddress (v1). Wire original-id = CCIPAddress while keeping published-at = CCIPMockV2PackageId so offramp v2's linkage routes fee_quoter::update_prices into v3 (unblocked) but the bytecode's linkage table references the correct on-chain original-id. Also drops the dead latest_ccip_pkg named-address — it has no consumer in ccip_offramp/Move.toml.
This was referenced Apr 14, 2026
…er upgrade Two independent failures cascade through the same chainaccessor batch-refresh path and zero-out the offramp OCR3 config digest, causing the CCIP commit plugin to self-reject its own reports (commit/report.go:222 "my config digest doesn't match offramp's"), making Test_CCIP_Upgrade_EVM2Sui and Test_CCIP_Upgrade_CommonPkg_EVM2Sui flaky on Sui. X. RMNProxy.GetARM → FunctionNotFound Sui chainreader config mapped MethodNameGetARM to Move function rmn_remote::get_arm, which doesn't exist in chainlink-sui's rmn_remote.move. Every refresh cycle hit FunctionNotFound in the RMN leg of the batch. On Sui the rmn_remote module lives inside the CCIP package, so the correct "ARM address" is the latest CCIP package id. Use ResponseFromInputs:["package_id"] to synthesize that response without touching chainlink-sui or Move source. Y. Sui fee_quoter has no dest-chain gas / token price state after upgrade AddLane(EVM→Sui) seeds the EVM-side fee_quoter only. No helper writes usd_per_unit_gas_by_dest_chain[EVM selector] or usd_per_token[Sui LINK] on the Sui side, so the commit plugin's observation reads abort EUnknownDestChainSelector(3) / EUnknownToken(4) from fee_quoter::*_internal. Seed those tables via FeeQuoterUpdatePricesWithOwnerCapOp against the post-upgrade (unblocked) CCIP package immediately after upgradeCCIP in both failing tests. Both failures share the batch-refresh fail-fast behavior in chainaccessor/config_processors.go: a single read error returns a zero-valued OfframpConfig, poisoning the config-digest cache. Fixing either alone still leaves the other to trigger the zero-digest cascade, so both must land together. (cherry picked from commit ba5feb2)
Contributor
|
I see you updated files related to
|
|
Theory: X (ResponseFromInputs:"package_id" for GetARM) is the decisive reader-side fix. Y was defense-in-depth that might be inert. Temporary commit for empirical validation. If A+X proves sufficient over 3 CI runs, this commit stays and Y stays dropped. If flakes return, revert.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Fixes intermittent CI flake in
Test_CCIP_Upgrade_EVM2Sui/Test_CCIP_Upgrade_CommonPkg_EVM2Sui: 20-min wait forCommitReportAcceptedyields zero events, then timeout. Three distinct but cascading failures — all three must land together for the tests to go green.Root causes (A + X + Y)
A. Offramp v2 linked against v1 ccip pkg
upgradeSuiOffRamp(integration-tests/smoke/ccip/ccip_sui_upgrade_test.go) compiled the v2 offramp with"ccip": state.CCIPAddress(original v1 ccip pkg) in the named-address map. Offramp v2's linkage table resolvedccip::fee_quoter::update_pricesto the v1 ccip pkg's fee_quoter (VERSION=2). AfterBlockVersion(fee_quoter, v=2), commit reports carrying non-empty price updates dispatch into fee_quoter v2 →verify_function_allowed(..., 2)fails → whole tx reverts → noCommitReportAcceptedevent emitted. Fix: point theccipdep's published-at atCCIPMockV2PackageId(upgraded pkg, fee_quoter v3). Supplyoriginal_ccip_pkg = CCIPAddressso Sui's upgrade validator accepts the dep swap.X.
RMNProxy.GetARM→FunctionNotFoundSui chainreader config (
core/capabilities/ccip/configs/sui/contract_reader.go) mappedMethodNameGetARMto Move functionrmn_remote::get_arm, which does not exist in chainlink-sui'srmn_remote.move(grep confirms zero definitions). Every config-refresh cycle failed on the RMN leg of the batch inchainaccessor/default_accessor.go→ fail-fast inconfig_processors.go→ zero-valuedOfframpConfig{}→ cached zeroConfigDigest→ commit plugin self-rejects own reports atcommit/report.go:222("my config digest doesn't match offramp's"). On Suirmn_remotelives inside the CCIP package, so the correct "ARM address" is the latest CCIP package id. Fix:ResponseFromInputs:["package_id"]synthesizes that value via chainlink-sui's existingexecuteFunctionplumbing — no Move source change, no dep bump.Y. Sui fee_quoter price tables empty for EVM dest
AddLane(from=EVM, to=Sui)populates EVM-side fee_quoter only (test_helpers_solana_v0_1_0.go:943-1012— nocase chainsel.FamilySui:on thetoFamilyswitch). Sui-sideusd_per_unit_gas_by_dest_chain[EVM selector]andusd_per_token[LINK]stay empty. Commit plugin observation aborts every cycle withEUnknownDestChainSelector(3)(fee_quoter.move:1578-1581) andEUnknownToken(4)(fee_quoter.move:1569-1571). In isolation these are tolerated warnings, but in the same batch as X they compound into the same zero-digest cascade. Fix: seed viaFeeQuoterUpdatePricesWithOwnerCapOpimmediately afterupgradeCCIP(so the write targets the post-upgrade, unblocked package). Values mirrorSendSuiCCIPRequestintest_sui_helpers.go:145-168.Why A, X, and Y cascade together
Both X and Y reads live in the same batch in
chainaccessor.prepareDestChainRequest.processOfframpResults/processStandardChainConfigResultsare fail-fast: one bad sibling zeros the wholeOfframpConfig→ cached zero digest → commit plugin refuses to transmit → even if reports ARE produced, they never land. Once X+Y let reports reach transmission, A (commit-tx linkage) is what finally lets the commit tx succeed on-chain.update_prices(when report carries price updates); works when report is empty — the original intermittencyEvidence
Failing job 71358465085 (latest CI on this branch before X+Y), log at
integration-tests/smoke/ccip/logs/2026-04-14T22-18-10_Test_CCIP_Upgrade_EVM2Sui.log:000acd0c35…) — proves read path isn't broken at the contract level.offRampConfigDigest: 0x0000…0000from poller cache → rejects own report.default_accessor.go:115errors repeat for full 22 min:get_arm: FunctionNotFoundandget_dest_chain_gas_price: EUnknownDestChainSelector.SubmitTransaction/signAndExecutecalls on the Sui chain writer through the test lifetime.Earlier failing job 71329559585 (pre-A-fix) shows
fee_quoter::update_pricesMoveAbort on commit-tx submission path — the A symptom that PR's original commit targets.Scope
integration-tests/smoke/ccip/ccip_sui_upgrade_test.go—upgradeSuiOffRampnamed-address fix (A);seedSuiFeeQuoterForRemotehelper called afterupgradeCCIPin both failing tests (Y)core/capabilities/ccip/configs/sui/contract_reader.go— 3-lineResponseFromInputsaddition onGetARM(X)go vet ./...andgo test -run=XXX -count=1 ./smoke/ccip/...clean.Test plan
Test_CCIP_Upgrade_EVM2SuipassesTest_CCIP_Upgrade_CommonPkg_EVM2SuipassesTest_CCIP_Messaging_EVM2Sui,Test_CCIP_Upgrade_Sui2EVM,Test_CCIP_Upgrade_NoBlock_EVM2Sui, EVM2Sui token-transfer testsResponseFromInputs:["package_id"]return the correct address for downstream RMNRemote binding (expected: yes — RMNRemote reads bind to the same CCIP package)Closes CORE-2414.