Repository navigation
test(dogfood): the census re-anchors rest-server.ts's enforceAuth control 61 -> 64 for three comment-only mentions - #23580
Conversation
…trol 61 -> 64 for three comment-only mentions Commit 1d79556 (the API-key lifecycle) added three prose mentions of `enforceAuth` to packages/rest/src/rest-server.ts: two in the docblock on `apiKeyLifecycleRefusals` and one in `computeExecCtx`'s outer catch. It added and removed no call site. The census counts the bare term, comments included, so the control moved 61 -> 64 while population, reach, blind spot and every other control on the row stayed as pinned. Re-anchored with a provenance note, as the row's earlier moves were; the control stays an exact count. Claude-Session: https://claude.ai/code/session_01YRp5z4kK6vgYkrdH97ztsy Co-authored-by: Claude <noreply@anthropic.com>
📓 Docs Drift CheckNothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs. What this run could not see
Coarse fallback — 0 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): |
席位核对ACCEPT · No contract face, so no review is owed beyond this one: a private test-census pin ( 1. Shape
2. Premises measured
3. Gates
4. Deviations accepted
5. Out of scope
6. Landing: Tier S, ordinary queue, once the head is all green. Generated by Claude Code |
Fixes #23563
Clause-②: no
The hourly full run's
Dogfood Regression Gate (3/3)is red onmainatpackages/qa/dogfood/test/authz-probe-blind-spot.test.ts:88:control "enforceAuth" moved in packages/rest/src/rest-server.ts: expected 64 to be 61. This PR re-anchors that one control by the census's own rule for a moved positive control, with a provenance note. It does not drop, narrow or relax it.What changed
One file:
packages/qa/dogfood/test/authz-probe-blind-spot.census.ts, on therest-server.tsrow only.controls.enforceAuth:61->64. It stays an exact count, not a range or a floor.[#23050 slice 1]one, in the same voice as the earlier moves ([commit cc837dbfe]61 -> 64,[#20102]64 -> 56,[#23050 slice 1]57 -> 61). It says the control was RE-ANCHORED, not relaxed, and names the three comment-only mentions by line.Not touched:
packages/rest/**(the comments #23541 added stay as written), the test runner, every other row and control, and the census's counting rule. The census states thatoccurrencescounts the bare term, comments included. Its earlier moves were re-anchored under that rule, so this is a re-anchor, not a rule change.Measurements (H1 to H4 from the dispatch)
Every count below comes from the census's own
deriveProbeFileCensus(), imported from the checkout named and run withtsx. None is a grep of my own.H1: only comments moved. Confirmed.
enforceAuth(derived / recorded)8226755ee9(last green hourly run; parent of1d79556766)1d79556766(PR #23541, #23367)fa6a6085c(this branch's base)git show 1d79556766 -- packages/rest/src/rest-server.tsadds exactly three lines that spellenforceAuthand removes none.data-action-routes.ts, the row's second file, is not in that commit. The three lines, atfa6a6085c:rest-server.ts:2395, docblock onapiKeyLifecycleRefusals: "by {@link computeExecCtx}'s catch so {@link enforceAuth} can answer it."rest-server.ts:2397, the same docblock: "identity fault: every door that never reachesenforceAuthkeeps the"rest-server.ts:3471, an inline//comment incomputeExecCtx's outercatch(declared at:3044): "forenforceAuth, which answers its403and code."All three are comments. The same commit also changed
enforceAuth's own body (the lifecycle-refusal answer) and its docblock, but none of those lines spells the term. It added no call site, forwarder or interface member. No commit between8226755ee9andfa6a6085cother than1d79556766touches either file the row reads.H2: nothing else on the row moved. Confirmed. At
fa6a6085c,population70 /reachable19 /blindSpot51 match the record. So do the other six controls:private register*Endpoints(16,this.routeManager.register(63,data-action-routes.ts: this.routeManager.register(9,registerPerItemRoute(8,const registerPerItemRoute =1,forwarder slice: this.routeManager.register(1. Every other census row also reads as pinned (0 rows moved).H3: red before, green after, locally. The dogfood shards skip on
pull_requestandmerge_group(ci.yml, thedogfoodjob'sif:), so this PR's CI will not run this test. These two local runs are the measurement of record. Both ran throughscripts/pm/os-verify-lock.sh, withpnpm --filter @objectstack/dogfood exec vitest run --maxWorkers=2 test/authz-probe-blind-spot.test.ts.Before, on
fa6a6085c(clean tree):After, on
11d842198(this PR's head):H4: the next move in flight. Measured, and it does not move this control. At PR #23571's head
3dd76b2a8(branchclaude/issue-23491-patch-data-request-gate, merge base086fa4640f, which already contains1d79556766), the census readsenforceAuth64, the same as the value re-anchored here. Itsrest-server.tsdiff (+41 / -17) adds and removes no line that spells the term. Every other control and figure on the row also reads as pinned there. That PR was not touched.Gates
node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstackat11d842198derives 47 commands for the one changed path. All 47 ran locally, each recorded with its exit code captured before any pipe:dist/across the workspace, and this worktree has no workspace build). These are NOT MEASURED locally:check:dts-closure,check:dual-build-cjs-loads,check:lean-entry-closureandcheck:sourcemap-no-sources-content. CI'sLint & Repo Gatesruns them.--ranreconciliation:✓ dispatch-gates --ran: 47 derived famil(ies) accounted for — 43 run, 4 NOT-MEASURED (4 DERIVED from a recorded exit 3).Declared narrowings:
pnpm test(the three dogfood shards) was not run locally. The census module is imported by exactly one file,test/authz-probe-blind-spot.test.ts(git grep). That file was run in full, so the single-file run covers the whole reach of this diff.pnpm --filter @objectstack/dogfood typecheckis left to CI'sTypeScript Type Check, whose PR selection includes the changed package. The edit is a numeric literal in the row'scontrolsrecord (typed as a read-only record of string to number), plus comments. Building the dependency closure would mean 67 packages, 65 of them a turbo cache MISS on a dry run.Changeset
None.
@objectstack/dogfoodisprivate: true, so this diff publishes nothing from any released package. Under the repo's convention the PR carries theskip-changesetlabel.Acceptance notes
enforceAuthtextually, comments included, by its own stated rule. Any prose mention added torest-server.tsordata-action-routes.tstherefore moves the control. Because the dogfood shards skip on PRs and in the merge queue, only the hourly full run onmainsees such a move. That is the two-speed design working as declared. Changing the counting rule is a census redesign, out of this card's scope. Carrier: none.Generated by Claude Code