Repository navigation
test doubles: a limit-blind ObjectQL find double in a plugin-sharing testkit sits outside both double gates, which walk *.test.ts only #28608
Description
domain:server#1 · os-project-manager · session_01R1q2xdtYgU6FKrZ7LgjPCC
Filing gate: ① out-of-scope finding of the #23181 slice 30 dev (os-dev-report 6555141723 on #23181), measured by a probe that was restored. Filed by domain:server seat 1. ⛔ Not a claim. No grade: triage sets the lane and grade.
What was measured. Both gates that hold test doubles to the real engine's contract discover doubles only in test files:
scripts/check-engine-double-contract.mjswalk()admits/\.(test|spec)\.(ts|tsx|mts)$/.scripts/check-objectql-double-limit.mjstestFilesUnder()admits/\.test\.ts$/.
The repo keeps 62 shared kits under __tests__/*.testkit.ts (and similar), which neither gate reads. With both walks widened to *.testkit.ts (probe only, not committed; output in pm-orders/server/evidence/23181-s30/odl-widen.log), check:objectql-double-limit reports:
packages/plugins/plugin-sharing/src/exec-context-seam.testkit.ts: NEW ObjectQLfinddouble that does not hold the caller's bound (1 blind, 0 breaking a shape rule, 0 unjudged). line 58: BLIND
So one limit-blind find double is live today outside the gate meant to catch it.
Two questions for triage.
- The plugin-sharing double: make it hold
limitby presence (or throw on a bound it does not implement), as the gate's own advice says. - Whether the two double gates should also walk testkit files. That widens what two existing gates enforce, so it is not a seat call (「新增门禁默认否」). The slice 30 report's option A measured the cost: one regex in
check-engine-double-contract(rows move cleanly with--write), and incheck-objectql-double-limitthe regex,ROOT_DIR_WATCH_HINTSand its--self-testDECLARED_TAILpin, plus paying any limit debt the widened walk surfaces (the baseline only shrinks).
Not affected by this card. Slice 30 is re-ordered to keep plugin-auth's shared double in its .test.ts file (seat ruling on #23181), so nothing there leaves either gate's sight.
Generated by Claude Code
Activity
objectstack-fleet commentedon Nov 1, 2026
domain:server#1 · os-project-manager · session_01R1q2xdtYgU6FKrZ7LgjPCC
Triage: p3 ·
domain:server·pm:queue· 2026-11-01T12:15ZRuling-ref: 「定级你可以自己决定」 (Leehom, 2026-11-01T12:11Z), which lets the server seats grade this lane's cards; the grading was split between seats 1, 2 and 3 in advance. ⛔ Not a claim.
makeSeamQl().findinpackages/plugins/plugin-sharing/src/exec-context-seam.testkit.ts(about line 58) filters onwhereand ignoreslimit. Five plugin-sharing files import it, so a caller's bound is silently dropped wherever a test runs through the seam. It is a test-double fidelity defect with no production reach, hence p3.limitby presence (slice the filtered rows), or throw on a bound it does not implement, as the gate's own advice says.limit: 1over two matching rows, on the base double.*.testkit.ts: graded out, not this card. Widening what two existing gates enforce is a gate addition in effect, and under 「新增门禁默认否」 that needs the maintainer to name it. Without that, the answer is "do not add". The probe log stays inpm-orders/server/evidence/23181-s30/odl-widen.logas the measured cost if the maintainer asks for it.bug,priority:p3,pm:queue.Generated by Claude Code
objectstack-fleet commentedon Nov 1, 2026
domain:server#1 · os-project-manager · session_01R1q2xdtYgU6FKrZ7LgjPCC
Claim: PM loop round 1 · card thread · 2026-11-01T12:20Z
Session:
session_01R1q2xdtYgU6FKrZ7LgjPCC(the seat); the card thread issession_01FEkQ5kA7pasGpR4Pf7mtNbAccount:
os-project-manager(the seat's linked user; the card's assignee)Branch:
claude/project-thread-qk6ezf(the card thread's harness-pinned branch; it names no issue, so this claim is the identity record)Worktree: the card thread's own container checkout
Domain:
domain:serverSeat:
domain:server#1Ruling-ref: triage record
6557807357(p3, question 1 only), graded under 「定级你可以自己决定」 (2026-11-01T12:11Z); p3 dispatch under 「server 车道现在就是你负责，任务有一大堆啊，都可以派，包括 p3」 (2026-11-01T03:46Z). Server seats 2 and 3 were told before this claim.File surface:
packages/plugins/plugin-sharing/src/exec-context-seam.testkit.ts(makeSeamQl().findholdslimit) and one plugin-sharing test for it. ⛔ No edit toscripts/check-engine-double-contract.mjsorscripts/check-objectql-double-limit.mjs.Container & model:
S,mode:thread, model and effort as the thread was openedClause-②: no (expected: a test double only)
Metadata-protocol: none (expected)
Responsibility: the code is plugin-sharing's shared test seam · no platform path (test-only) · reached by the five plugin-sharing test files that import the seam
Thread-read: none
Premise: holds (read 2026-11-01T12:20Z at
origin/main0b6a52e8a0:makeSeamQl().findfilterswhereand never readsopts.limit)Serial constraints cleared: no open PR touches
packages/plugins/plugin-sharing/src/Resume-from: none — fresh from
origin/main0b6a52e8a0State:
pm:queue→pm:dispatched, assigned toos-project-manager, recorded by this claim.Generated by Claude Code
objectstack-fleet commentedon Nov 1, 2026
domain:server#1 · os-project-manager · session_01FEkQ5kA7pasGpR4Pf7mtNb
os-dev-report
{
"issue": 28608,
"status": "done",
"branch": "claude/project-thread-qk6ezf",
"pr": "#28670",
"session": "session_01FEkQ5kA7pasGpR4Pf7mtNb",
"premise_still_valid": true,
"summary": "makeSeamQl().find in packages/plugins/plugin-sharing/src/exec-context-seam.testkit.ts now slices the where-filtered rows to the caller's limit by presence (typeof opts?.limit === 'number'), the first shape check-objectql-double-limit.mjs prescribes. Slicing was picked over throwing because resolveAuthzContext reads through this double with real bounds (limit 1, 2, 200 and others), so a throwing double would break every importer. where handling is unchanged; makeSeamQl is exported only for its new pin exec-context-seam.limit.test.ts. Neither gate script is touched (question 2 out). Importer count measured: 3 test files import the seam plus 1 type-only import (exec-context-annotation.pin.ts), not five; all pass unchanged. No PR assignee and no PR subscription, per the order.",
"tests": "Red first (export only, fix absent): exec-context-seam.limit.test.ts 'answers at most limit rows when more match' fails, expected [ Array(2) ] to deeply equal [ { id: 'a', …(1) } ], Tests 1 failed | 2 passed (3). Green on fix: Tests 3 passed (3). Ablation via scripts/ablation-replace.mjs: limit line replaced by 'return matched;' (anchor x1 -> x0, blob d27c845e10d6 -> 17d1ac7833f6), same case red (1 failed | 2 passed), restore: blob after restore == HEAD blob d27c845e10d6, git diff HEAD empty; testkit is source imported relatively, no dist leg. Whole suite on head e410091 under the lock: pnpm --filter @objectstack/plugin-sharing test, VERDICT command-exit 0, Test Files 59 passed (59), Tests 1299 passed (1299). pnpm --filter @objectstack/plugin-sharing typecheck VERDICT command-exit 0. eslint --no-inline-config packages/plugins/plugin-sharing exit 0. Gate-walk probe (copy of check-objectql-double-limit.mjs with testFilesUnder widened to (test|testkit).ts, never committed, deleted): head exit 0 'OK ... 643 double(s) graded', no plugin-sharing row; base testkit blob exit 1 'exec-context-seam.testkit.ts ... (1 blind ...) line 58: BLIND'; restore proven by blob equality and empty git diff HEAD.",
"gates": "Local subset on head e410091 from dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (52 derived). 51 exit 0, incl. check:objectql-double-limit, check:engine-double-contract, check:where-matcher, check:query-options-erasure, check:test-source-alias, check:type-check-debt, check:published-files, check:nul-bytes. check-plugin-teardown-shape.mjs --self-test exit 3 (shallow) then exit 0 after fetching fixture 621a487. check:i18n exit 3 (CLI unbuilt) then exit 0 after its prerequisite build. check:dual-build-cjs-loads exit 3 PREREQUISITE NOT MET = NOT MEASURED (needs every package dist), CI runs it. --ran: '52 derived famil(ies) accounted for — 51 run, 1 NOT-MEASURED', exit 0. Record: pm-orders/server/28608-gates-ran.txt. PR check-runs: in_progress at report time (not waited on).",
"dogfood_battery": "n/a — no ! in the PR title and no (narrowing) arm",
"mcp_calls": "4 — reads only: issue_read get, issue_read get_comments x2, pull_request_read get",
"api_writes": "2 — POST /repos/objectstack-ai/objectstack/dispatches (relay pr_create with labels skip-changeset, run 41308407869 success, PR #28670 body read back identical, assignees none), and this os-dev-report comment via post-stamped.mjs; plus git push (not REST)",
"open_questions": [],
"out_of_scope_findings": [
"none — the widened-walk probe on head shows no other testkit double red; the probe log row for plugin-auth memory-engine.testkit.ts is moot because that file does not exist on main"
]
}
Generated by Claude Code
objectstack-fleet commentedon Nov 1, 2026
domain:server#1 · os-project-manager · session_01R1q2xdtYgU6FKrZ7LgjPCC
Landing record · #28608 · PR #28670
PR #28670 merged by the merge queue as
49a6a5c372onmain(Fixes #28608; ACCEPT6558400493). The card closed on the merge.git ls-remote origin 'refs/heads/gh-readonly-queue/main/pr-28670-*'origin/mainby contentgit diff --quiet 49a6a5c372 origin/main --the PR's 2 filesgit merge-base --is-ancestor 49a6a5c372 origin/mainmakeSeamQl().findnow holds the caller'slimitafter itswherefilter, pinned byexec-context-seam.limit.test.ts. Question 2 (widening the double gates to*.testkit.ts) stays graded out. Test-only.Release: branch
claude/project-thread-qk6ezfmerged; nothing unpushed.pm:dispatchedremoved and the card unassigned.Generated by Claude Code