Repository navigation
test: drop blocktank-dependent tests - #1817
Conversation
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 530c07b680
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
[Low impact] The changes affect test setup and staging automation, not shipped wallet behavior. The PR appears safe to merge, with a non-blocking update needed to the local test instructions. Findings
SummaryMoves instrumented test funding and mining from Blocktank to a local bitcoind.
Diagram%%{init: {'theme': 'neutral'}}%%
flowchart LR
CI[Integration workflow] --> Stack[Local bitcoind and electrs]
Tests[Instrumented tests] --> Helper[LocalRegtest]
Helper -->|Fund and mine| Bitcoin[bitcoind]
Bitcoin --> Electrs[electrs]
Wallet[Wallet built with E2E=true] -->|Sync| Electrs
Reviews (1) · Last reviewed commit: "test: harden local regtest helper and ga..." · Reviewed by Greptile |
Regtest APKDownload bitkit-dev-debug universal APK (expires in 30 days). |
talosmachina
left a comment
There was a problem hiding this comment.
No findings. Moves the instrumented integration tests (UtxoSelectionTests, TxBumpingTests, PreparedOnchainRecoveryFixtureTest) from Blocktank's staging regtest deposit and mine calls to a local bitcoind through LocalRegtest, starts the bitkit-docker regtest stack in ui-tests.yml for non-ComposeUi suites, removes BlocktankTest and drops the two staging nightly shards whose tests leave in synonymdev/bitkit-e2e-tests#302. Reviewed 530c07b, full tier (CI workflow files), reasoned from the code and from CI: on this track the auto review does not build the branch.
What I checked, and 6 candidates I ruled out
Read in full: ui-tests.yml, run-ui-tests.sh, LocalRegtest.kt, the three migrated test files, Env.kt (electrumServerUrl, electrumUrlForNetwork, homegateUrl, BackupTestEndpoints, BlocktankTestEndpoint), the E2E* build config in app/build.gradle.kts.
CI: run 42616169750 (ui-tests, integration suite) on 530c07b is green, 42 tests finished; the skips are OnchainBackupRestoreDeviceTest, UsdtNativeRecoveryIntegrationTest, SendPendingObservationDeviceTest and verifyPersistedOriginalWinnerAfterUi, each behind an assumeTrue on an instrumentation argument the run does not pass, so they skip the same way without this PR. On the PR head, build-and-test, build-local, lint, detekt and Greptile passed; compose-ui and the e2e-tests-local shards were still running.
Ruled out
- ComposeUi suite on every PR picks up
E2E=trueor the regtest stack:ci.ymlcalls the workflow withsuites: ComposeUi, soNEEDS_REGTESTandE2Eevaluate to'false', andenvFlagmaps that tofalse. E2E=truesilently moves other integration tests to local backends they cannot reach: the build flag changes electrum (totcp://10.0.2.2:60001, which the stack serves), homegate and the release URL; backup and Blocktank fixture endpoints stay off unless theirE2E_*URLs are set, and none is. The green run covers the rest of the integration set under that build.sendtoaddressfails on regtest without fee estimation, or with no wallet loaded: the run funds and confirms every deposit the three classes make.- Five
sendtoaddresscalls before mining chain unconfirmed change in the bitcoind wallet: a chain of five is well under bitcoind's 25-descendant limit, the order is the same as it was against Blocktank, and the run passes it. - Removing the injected
CoreServiceleaves a dangling use:compileDevDebugAndroidTestKotlinran in that job; no other reference remains in those files. - New workflow steps expose a secret: the added steps use no secret;
bitkit-dockeris checked out with the job's read-only token and onlydocker composeandbitcoin-clirun from it.
Merge confidence: 4/5, no findings, but compose-ui and the e2e-tests-local shards on this head were still running at review time.
jvsena42
left a comment
There was a problem hiding this comment.
Code scan at 857f3a197: no findings.
Checked and clean:
- The dropped
BlocktankTestheld onlytestGetInfo, which calls Blocktank. That matches iOS #1246. UtxoSelectionTests,TxBumpingTestsandPreparedOnchainRecoveryFixtureTestmove one-for-one fromregtestDeposit/regtestMinetoLocalRegtest, with assertions unchanged. No Blocktank callers are left in androidTest.ui-tests.yml: the per-PRComposeUilane is unchanged, becauseNEEDS_REGTEST/E2Eare false for exactly that filter.E2E_BACKEND=localreaches both the assemble and the connected-test Gradle calls.assert-android-tests-ran.pystill catches a zero-test filter.- e2e-staging: the
multi_addressandtransfershards dropped are the same on both platforms. Their staging tags mark only Blocktank purchases. Themulti_address_localshard is unaffected.
Device gate: N/A, test and CI changes only, no journeys touched.
Reviewer model: Claude Opus 5.5 (claude-opus-5-5)
Reasoning effort: not exposed
coreyphillips
left a comment
There was a problem hiding this comment.
Two independent reviews, nothing blocking a merge.
worth doing, does not block
- Instrumented lane checks out bitkit-docker at unpinned main (.github/workflows/ui-tests.yml:50-56)
- CPFP skip guard still describes a shared regtest chain (app/src/androidTest/java/to/bitkit/services/TxBumpingTests.kt:280-283).
TxBumpingTests.kt:280-283says 'The shared regtest chain can mine the parent before the child is built' andassumeFalses on 'Transaction already confirmed'. That guarded against other users mining on stag0. On the private bitkit-docker chain nothing mines between the deposit and the CPFP call, so this path would now hide a real regression as a skip instead of failing. The PR already rewrote the comment at line 232 in the same test. Updating this comment, or dropping the guard, belongs with that change.
nits
- LocalRegtest redefines the existing SATS_IN_BTC constant (app/src/androidTest/java/to/bitkit/test/LocalRegtest.kt:18)
@coreyphillips dropped the guard in dba02d0: on the private chain nothing mines the parent first, so |
jvsena42
left a comment
There was a problem hiding this comment.
Delta review 857f3a197..dba02d028: no findings.
- Pinning bitkit-docker to
f5eae182dmakes the lane reproducible. - Dropping the CPFP
assumeFalseis right now that the chain is local to the run: nothing else mines the parent. LocalRegtestreusesSATS_IN_BTCand adds a 30 s RPC timeout.
Device gate: N/A.
Reviewer model: Claude Opus 5.5 (claude-opus-5-5)
Reasoning effort: low
coreyphillips
left a comment
There was a problem hiding this comment.
Rechecked after 1 new commit (dba02d0). Nothing blocking remains.
addressed
- Instrumented lane checks out bitkit-docker at unpinned main (.github/workflows/ui-tests.yml:50-56). Fixed:
ui-tests.yml:55now usesref: f5eae182db9ddecd61a0f1b5e9f48aee283bb3dc, anddocs/android-test-lanes.md:44points readers at that pin. I confirmed the SHA exists in bitkit-docker (merge of PR #34, 2026-11-02). Itsdocker-compose.ymlstill defines what this branch relies on: containerbitcoin, rpcport 43782, electrs on 60001, thepolaruserrpcauth,fallbackfee=0.00002, and the bitcoind/bitcoinsetup/electrs services. Run 42625191534 on dba02d0 has passed the checkout and stack-startup steps. - CPFP skip guard still describes a shared regtest chain (app/src/androidTest/java/to/bitkit/services/TxBumpingTests.kt:280-283). Fixed:
TxBumpingTests.kt:271-275now callsaccelerateByCpfpdirectly, and theassumeFalseandrunSuspendCatchingimports are gone. I checked that nothing can mine the parent first. At the pinned commit,bitcoinsetupmines one block and exits (restart: 'no').run-ui-tests.shrunsconnectedDevDebugAndroidTestwithout sharding. So nothing mines between the deposit at line 232 and the CPFP call. - LocalRegtest redefines the existing SATS_IN_BTC constant (app/src/androidTest/java/to/bitkit/test/LocalRegtest.kt:18). Fixed:
LocalRegtest.kt:9,57-58imports and usesto.bitkit.models.SATS_IN_BTC, a publicconst valinmodels/Currency.kt:15. Long divided by Int stays Long, so%d.%08dstill formats correctly. The connect and read timeouts now come fromrpcTimeout = 30.seconds(line 20), used at lines 30-31.
Twin: synonymdev/bitkit-ios#1246
Companion: synonymdev/bitkit-e2e-tests#302
This PR moves the instrumented integration tests off Blocktank's staging regtest API. That API is being sunset, and its deposit and mine calls take tens of seconds each on stag0.
Description
BlocktankTest, whose one test calls Blocktank'sinfoendpoint and cannot work after the sunset.UtxoSelectionTests,TxBumpingTestsandPreparedOnchainRecoveryFixtureTestfrom Blocktank's regtest deposit and mine endpoints to a local bitcoind, through a newLocalRegtesttest helper. These tests only used Blocktank to fund and mine, so their coverage stays.ui-tests.ymland builds withE2E=trueso the wallet syncs from the local electrs. The ComposeUi suite, whichci.ymlruns on every PR, skips both.testAccelerateByCpfpskip for an already-confirmed parent: it guarded against other users mining on the shared staging chain, which the private chain does not have.docs/android-test-lanes.mdwith the steps to run the integration lane locally on the bitkit-docker stack.multi_addressandtransfershards. Their only tests (@multi_address_2,@transfer_1and@transfer_max, all Blocktank channel purchases) are removed in test: drop blocktank-dependent tests bitkit-e2e-tests#302.Out of Scope
CoreServiceregtest helpers, regtest dev screens): app code, left as is.BlocktankRepo: they test app code that still ships and never reach Blocktank.journeys/, thelspwrapper and theblocktank-apiagent plugin: some walkthroughs still fund through Blocktank's regtest API.Design
N/A — no UI changes.
Preview
N/A — no user-visible changes.
QA Notes
Setup
N/A — no device testing required; test and CI changes only.
Journeys
N/A — no user-visible behaviour change.
Manual Tests
N/A
Automated Checks
LocalRegtest.kt— funds and mines on the local bitcoind over JSON-RPCUtxoSelectionTests.kt,TxBumpingTests.kt,PreparedOnchainRecoveryFixtureTest.kt— fund and mine on local regtest instead of BlocktankBlocktankTest.kt— its only test called Blocktank'sinfoendpoint, which is being sunsetintegrationsuite throughui-tests.ymlon this branch: run 42616169750Evidence
530c07bpassed:UtxoSelectionTests(2),TxBumpingTests(2) andPreparedOnchainRecoveryFixtureTest(seedPreparedOriginalForRecoveryUi,verifySharedVssBindingVectors) passed against the local stack.verifyPersistedOriginalWinnerAfterUiskipped, as it does without its fixture argument.compileDevDebugAndroidTestKotlin,detektandjust lintpassed locally.Gaps
just testwas not run locally: no JVM unit test or main source changed. CI runs it.Models used
claude-opus-5-5(reasoning:medium)claude-opus-5-5(reasoning:medium)claude-sonnet-5-5(reasoning:Not exposed)The reviewer's run recorded a numeric effort tag, not a named level.