Repository navigation
test(cron): pin cronLgtm's merge count and its no-read short circuits - #573
Closed
hivecommons-hive[bot] wants to merge 1 commit into
Closed
hivecommons-hive[bot] wants to merge 1 commit into
hivecommons-hive[bot] wants to merge 1 commit into
Conversation
cronLgtm resolves to the number of pull requests it merged or enqueued,
and a pull request it skips before the shared merge path - gate failure
or locked - costs no further request. Neither was asserted: the count is
discarded by handleCronJob, and tide's own blockedReason also refuses a
locked or gated PR, so the existing merge observers could not see the
guards. A mutation run over src/cronJobs/lgtm.ts left 13 survivors; the
nine behavioural ones (jobsDone++, return false, the !gate.ok and locked
guards, { required: true } on github-token, state: 'open') are now
killed by direct cronLgtm calls that watch GET /pulls/:number. The four
left are log text.
Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com>
hivecommons-hive
Bot
added
the
hold
Contributor
Author
|
Important Held for human review by the hive's ACMM level gate. This PR was opened by the "quality" agent while Hive policy required a human checkpoint for that agent. Agent PRs are held at ACMM L3–L5; an Hive will keep the |
![hivecommons-hive[bot]](https://avatars.githubusercontent.com/in/3568013?s=60&v=4){kind=small}
Contributor
Author
There was a problem hiding this comment.
Read the test file against the tree at the reviewed head, not just the diff. This looks correct to me.
Verified against src/cronJobs/lgtm.ts:
- Count contract —
cronLgtmaccumulatesprogress.jobsDone++per merged/enqueued PR (lgtm.ts:98-100) and returns it from the empty-page arm (lgtm.ts:84). The three count tests (2 across two pages, 0 all-skipped, carry-in{jobsDone: 3}→ 4) pin exactly that, and the carry-in test correctly starts at page 2 so the emptypages[0]never terminates it early. - No-read short circuits — a gate failure returns from
tryMergePrat lgtm.ts:162-165 beforeevaluateMerge, andpr.lockedreturns at lgtm.ts:93-95 beforetryMergePr; neither path can issueGET /pulls/:number. TheObserveRequest.notCalled()assertions (testUtils.ts:200) observe that directly, andfailOnUnhandledRequestkeeps the fixtures honest. state: 'open'is passed ingetOpenPrs(lgtm.ts:129) and the listing test reads it off the captured request rather than trusting the stub.- Token guard —
core.getInput('github-token', { required: true })at lgtm.ts:60 runs before any listing; the test deletes the env var and assertslisting.refstays null, matching the order. - The custom
do-not-merge/*gate case matchesmeetsMergeGate'smissing_labelsarm (src/utils/mergeGate.ts:26-28).
One note, not a defect: the PR body says handleCronJob.ts:33 discards the count, but that line is return await cronLgtm(1, context)... — the value is propagated, presumably dropped further up. Doesn't change the finding that no prior test observed it.
Additive only (one new test file, no source or existing-test edits), disjoint from #571/#572 as the body claims, CI green. No findings.
— hive: agent=reviewer backend=copilot model=claude-fable-5 copilot=1.0.88
This was referenced Oct 25, 2026
Contributor
Author
|
/kind cleanup 🐝 Hive Agent: — hive: agent=quality backend=copilot model=claude-fable-5.1 copilot=1.0.88 |
github-actions
Bot
added
kind/cleanup
Merged
11 tasks done
Member
|
Consolidated into #588 together with the other eight open mutation-survivor pin PRs (commit cherry-picked unchanged, DCO preserved). Closing in favour of that PR. |
mrbobbytables
added a commit
that referenced
this pull request
Oct 26, 2026
, #578-#583, #585) (#588) * test(cron): pin cronLgtm's merge count and its no-read short circuits cronLgtm resolves to the number of pull requests it merged or enqueued, and a pull request it skips before the shared merge path - gate failure or locked - costs no further request. Neither was asserted: the count is discarded by handleCronJob, and tide's own blockedReason also refuses a locked or gated PR, so the existing merge observers could not see the guards. A mutation run over src/cronJobs/lgtm.ts left 13 survivors; the nine behavioural ones (jobsDone++, return false, the !gate.ok and locked guards, { required: true } on github-token, state: 'open') are now killed by direct cronLgtm calls that watch GET /pulls/:number. The four left are log text. Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test(owners): pin the per-team group cache key, parseGroup anchor, decode, defaultBranch and isNotFound guards Running the bounded Stryker check against src/utils/owners.ts left 15 survivors at 100% line coverage. Thirteen are behavioural: with one groupCache key every @org/team entry resolves to the first team's members; parseGroup without its leading anchor reads alice@org/team as a group; a bare `options:` key, a blob without content, a payload without a repository, a string or null tree rejection and a path-less tree entry each hit a TypeError instead of the handled path; and repoHasOwners would answer a second repository from the first. Each is pinned with a direct call in a new file so the harness reports 98.49% on the module, leaving only the warning text held #564 asserts and a debug string. Closes #574 Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test(tide): pin mergeOnce's status guard, the 403-only fork diagnosis, sorted workflow list, queue-path retryIf, entry-less dequeue and isUnknown's state arm Each case fails when the guard it names is removed or inverted; line coverage already reached them all (Stryker on src/plugins/tide.ts: 477/517 killed before, 488/517 after). Refs #570 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test(auth): pin the repository-less org check, the gone-user comment skip, the non-RequestError and half-formed root OWNERS reads, and the OWNERS path in role errors A Stryker probe of src/utils/auth.ts (--mutate src/utils/auth.ts on the harness of #572) left 22 of 273 mutants alive. Eleven of them flip a guard the suite reaches but never distinguishes; this file pins each one: - checkOrgMember on a payload without a repository returns false by the early return, not by a TypeError caught as an unexpected error, so no warning annotation is emitted - checkIssueComments skips a comment whose user is null (e.user?.login) instead of aborting the scan - retrieveOwnersFile reports a string or null rejection of repos.getContent as the OWNERS lookup error rather than tripping on 'status' in e - a contents response with content but no encoding, or an encoding but no content, is rejected as an invalid OWNERS file rather than decoded - a root OWNERS role that is not a list is reported as 'OWNERS at OWNERS:' from both assertRootOwner and rootOwnersIncludes The seven remaining survivors are core.debug strings and the contents === '' short-circuit of rootOwnersIncludes, which is equivalent: parseOwners('OWNERS', '') yields empty lists and expandGroups of an empty list returns an empty set without a request. Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test(approve): pin the github-actions[bot] login-only skip, the uncoverable-file suggestion, the notifier whitespace and marker checks, the unlabeled other-label skip, the base-branch OWNERS probe and the required token A Stryker probe of src/plugins/approve.ts (--mutate src/plugins/approve.ts on the harness of #572, main@e3119ad) left 59 of 515 mutants alive. Twenty-six of them flip a guard the suite reaches but never distinguishes; this file pins each one in __tests__/plugins/approveGuards.test.ts: - approvalEvents skips a github-actions[bot] comment by its login alone, without relying on the Bot user type - suggestApprovers still names an approver when another changed file has no OWNERS entry at all, instead of dereferencing the missing per-file set - computeApproval logs the approver who covers none of the changed files and says nothing for a pull request without an author - renderNotifier's no-files body is exactly the verdict, the explanation and the marker - approveOnPullRequest returns on `unlabeled` of another label without a single api call, probes the OWNERS of the event's base branch rather than the default branch, and fails on a missing github-token input before reading anything - upsertNotifier recognises a stale notifier by a github-actions[bot] user without a Bot type, skips a comment whose user is gone, treats a notifier that differs only by surrounding whitespace as up to date, and never mistakes a bot comment without the literal marker for the notifier - approveOnReview on a payload without a review or a pull request fails on the missing pull request, not on the missing review Re-running the probe with this file kills all twenty-six; the 33 that remain are equivalent (log strings, optional chains on values already checked, switch defaults, an excluded.add after the covered files were removed). Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test(mergeQueue): pin errorMessage's null-rejection guards, the once-per-run warning's initial state and the [bot] suffix anchor Adds __tests__/utils/mergeQueueGuards.test.ts. Every line and branch of src/utils/mergeQueue.ts is already covered, but five of its mutants survive the Stryker harness: `npx stryker run --mutate src/utils/mergeQueue.ts` (main@e3119ad) kills 106 of 111 (score 95.50%). The survivors, each re-applied by hand at its reported span and confirmed to pass the existing mergeQueue.test.ts: - mergeQueue.ts:184 `e !== null` -> `true`, the whole `typeof e === 'object' && e !== null && 'errors' in e` conjunction -> `true`, and the final `&&` -> `||`: a null or undefined rejection of octokit.graphql would throw a TypeError out of errorMessage, so queueState would reject instead of warning and falling back to the REST merge, and enqueue would throw instead of returning `{ ok: false }`. - mergeQueue.ts:64 `let warnedUnavailable = false` -> `true`: every existing test calls resetMergeQueueWarnings() (directly or via setupActionsEnv), so the first real run of a job never warning went unnoticed. Pinned on a fresh module instance (vi.resetModules) that nothing has reset. - mergeQueue.ts:153 `/^github-actions$|\[bot\]$/i` -> `/^github-actions$|\[bot\]/i`: a login merely containing `[bot]` would count as an automation whose queue entry the gate may undo. With this file all five are killed (manual re-application of each mutant fails 1-3 of the new cases; 111/111 killed). No production code changes. Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test(lgtm): pin the refusal error log, the no-label cancel debug, the token guard, the missing-head error and isForbidden's non-object arms A Stryker probe of src/labels/lgtm.ts and src/plugins/lgtmBinding.ts (npx stryker run --mutate src/labels/lgtm.ts,src/plugins/lgtmBinding.ts on the harness of #572, main @ e3119ad, Stryker 10.0.0 / Vitest 5.0.4, 8 workers, 6 m 36 s) left 24 of 202 mutants alive (88.12 %). Fourteen flip a guard or drop a log line the suite reaches but never distinguishes from its mutation. Two new test files kill all fourteen; the re-run reports 192 killed / 10 survived (95.05 %). __tests__/label/lgtmGuards.test.ts (src/labels/lgtm.ts): - 29 `getInput('github-token', { required: true })` (2 mutants): the handler fails fast with "Input required and not supplied" and posts nothing - 140 `core.error(msg)` in refuse: the refusal is logged as an error with the same text it replies with - 96 the "could not find lgtm to remove" debug line of a cancel on an issue without the label, and that no DELETE follows __tests__/plugins/lgtmBindingGuards.test.ts (src/plugins/lgtmBinding.ts): - 171 `getInput('github-token', { required: true })` (2 mutants) in lgtmOnPullRequest - 166 `pull_request?.head?.sha` -> `head.sha`: a pull_request without a head throws the "missing pull request head" TypeError, not a property read failure - 173 the "lgtm: bind_to_commit is false" debug line - 182 isForbidden `typeof error === 'object' && error !== null && 'status' in error` (6 mutants: `true` x4, `||` x2): bindLgtm and isLgtmBound wrap a null, string or number rejection in their own error instead of throwing from `'status' in <primitive>` / `null.status` Reviewed survivors, not pinned (10): lgtm.ts 33, 34 (x2), 35, 65 and lgtmBinding.ts 162, 177, 178 are optional chains on fields every issue_comment / pull_request payload carries (comment.body, comment.user, issue.user, pull_request after the head check, sender once isBotUser is true); lgtmBinding.ts 66 `targetUrl === undefined` -> false spreads `target_url: undefined`, which JSON serialisation drops, so the request body is identical; lgtmBinding.ts 156 `label?.name ?? ''` -> any other string still fails the `!== 'lgtm'` comparison and returns early. Refs #554 Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test(trigger): pin the guards of /retest, /test, /ok-to-test and okToTestOnPullRequest that a Stryker probe left unasserted A Stryker probe of src/issueComment/trigger.ts (346 mutants) left 38 alive. Twenty-three of them flip a guard the suite reaches but never distinguishes from its mutation: the react() comment-id guard, the refusal error log, the gone-commenter and gone-author optional chains, the required github-token input, the .yaml and full-path workflow matches, the no-match argument echo, the ok-to-test label equality, the approved-and-labeled reaction, the status-only action_required candidate, the null-rejection statusOf guard, the synchronize/reopened action filter, the pull-less and head-less payloads and the null label entry. This file adds one test per guard; the probe re-run reports 331 killed / 15 survived (95.66 %). The 15 survivors are equivalent or unreachable and are listed in the pull request. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test: pin the milestone list separator, isBotUser's login arm, createComment's message and skipReadOnlyForkRun's payload guards A Stryker probe over the issue-comment, label and utility modules left nine behavioural survivors the unit suite reaches but never distinguishes from their mutations. Three new files pin them: - __tests__/issueCommentTest/milestoneAvailableList.test.ts: the not-found error joins the open milestones with ', ' (milestone.ts:66). - __tests__/utils/commentsGuards.test.ts: isBotUser accepts the github-actions[bot] login without a Bot type (comments.ts:67), createComment rewraps a refused post as 'could not add comment: ...' (comments.ts:34), and createCommentOnce honours a login-only bot marker. - __tests__/pullReqTest/readOnlyForkPayloads.test.ts: skipReadOnlyForkRun returns false rather than throwing on a payload without pull_request, head, repository or a string full_name (events.ts:23-25). Closes #584 Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> --------- Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This was referenced Oct 27, 2026
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
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.
Test Improvement
Adds
__tests__/cronJobTest/lgtmShortCircuits.test.ts(new file; nothing else changes), callingcronLgtminsrc/cronJobs/lgtm.tsdirectly to pin two contracts its docstring states but no test observed:cronLgtmreturns the number of PRs merged or enqueued across every page.handleCronJob.ts:33discards it, sojobsDone++ → --andreturn false → trueintryMergePrpassed every existing test.meetsMergeGateor islockedis skipped beforeevaluateMerge. Because tide's ownblockedReasonalso refuses such a PR, the existing merge observers could not distinguish the guard from its absence; the new cases watchGET /pulls/:numberand assert it is never made.state: 'open'on the listing and{ required: true }on thegithub-tokeninput.Evidence
Stryker (harness from #572,
--mutate src/cronJobs/lgtm.ts, main @ e3119ad, local): 13 survivors → 4, score 80.60% → 94.03%. The four left are log-text literals (lines 57, 87, 125, 133). Unit/e2e line and branch coverage of the file were already 100% (npm run test:coverage,npm run test:coverage:e2e, same revision), so this is a mutation finding, not a coverage gap.Occupied-ground check
Disjoint from every open hold-gated PR: #571 touches
__tests__/cronJobTest/lgtm.test.ts(harness-wide edit) — this PR adds a separate file and does not edit it; #572's defaultmutateset is mergeGate/labelMatch/hold/duration, notcronJobs/lgtm.ts.Checks run
npx vitest run __tests__/cronJobTest— 98 passednpx eslint __tests__/cronJobTest/lgtmShortCircuits.test.ts,npx tsc --noEmitFiled by quality agent (hold-gated mode). Human review required.
— hive: agent=quality backend=copilot model=claude-fable-5.1 copilot=1.0.88