packages: 30 of 39 published packages declare no sideEffects, so consumers cannot tree-shake them; declare per package now, make registration explicit later so the 1,380-line array gate can go

Filing gate ③: a maintainer-directed task. Maintainer, 2026-10-24, verbatim: 「同意：建议新建：与合并无关、独立成卡的产品/契约项（可现在建）」, approving the review's list of independent product / contract cards. Batch-of: 13 (this review batch; the per-fire quota is overridden by the maintainer's direction).

维护者速读

39 个发布包里 30 个没有 sideEffects 字段（false 5、true 2、数组 2）。字段缺失在 webpack / Rollup / Vite 消费者那里等于「每个模块都可能有副作用」，整个包不能摇树——这正是 check-side-effects-array.mjs 自己文件头量过的代价（app-shell 漏写时 console 首屏多出约 3.3 MB gzip）。而那个 1,380 行的门禁只校验两个包的数组。短期逐包声明；长期把「导入即注册」改成显式 register*()，每个包都能写 false，门禁随之删除。

Measured (2026-10-24, HEAD `7981384`; manifest census)

sideEffects present in 9 of 39 published manifests: false in core, i18n, react-runtime, sdui-parser, types; true in components, fields; arrays in app-shell (18 entries) and layout. Absent in the other 30: auth, cli, collaboration, create-plugin, data-objectstack, mobile, permissions, providers, react, runner, every plugin-* except none, console.
scripts/check-side-effects-array.mjs (1,380 lines) validates exactly the two array packages; its header records the measured cost of an omitted field on app-shell.
The plugin packages register on import (ComponentRegistry.register(…) at module top level; apps/console/src/register-plugins.ts imports five of them for the side effect), so for them an honest value today is an array or true; pure libraries (react, providers, permissions, mobile, auth, collaboration, data-objectstack) can say false after a read of each entry.
Business reason: every external host that bundles @object-ui/* pays for code it never calls; the console's own first-load budget (objectui#13447, console: first authenticated load pulls 336 eager JS chunks (~11.7 MB decoded) objectui#11798) is the in-house symptom of the same mechanism.
Dedup (sideEffects): 12 hits, 2 open — objectui#13410 (app-shell pulling the chatbot root; the consumer-side symptom) and a Dependabot bump; closed: feat(app-shell): the designer registries are observable, and their readers re-render on registration (objectui#11939 step 1) objectui#11963, feat(app-shell): register record:approval_decision, armed in @object-ui/types (objectui#12342) objectui#12362 (registration observability). No card asks for the field on the 30 packages.

Task

Per package: false for the pure libraries, an array (the existing gate in --list mode helps derive it) or true for packages that register on import. sideEffects is one of the eight contract fields, so each change carries a changeset.
Longer term (decision in its own card when ready): explicit registration (register*() exports, no import-time mutation) so every package can declare false and check-side-effects-array.mjs is deleted.
未实测: consumer bundle delta — measure with check-eager-closure-budget on apps/console before and after.

Who acts: objectui domain:ui seat, one PR for the declarations. Effort M (30 manifests, each read). Risk: a wrong false drops a live registration (the objectui#6535 failure class) — run the gate per package before landing.

Filed by the read-only whole-repo review session https://claude.ai/code/session_01BDtwYUWhLGyWLzYQjZ95Cf on 2026-10-24, written through the maintainer's own account token (the session holds no fleet-bot credential). Evidence files: that session's scratchpad (objectui-review-report.md, findings-*.md). ⛔ Not a claim; triage grades and routes; suggested labels are suggestions only.

Generated by Claude Code

Triage: first grade — `tech-debt` · `priority:p3` · `domain:ui` · `area:devpath` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-24T17:06Z

⛔ Not a claim, ⛔ not a dispatch.
finding comes off.

Declare sideEffects per package now, measured package by package. Explicit register*() comes later, on its own card, because it changes what an import does for consumers.

Generated by Claude Code

Claim: PM loop round 1 · 2026-10-25T02:19Z
Session: session_01Q4K9dtZAX6vw67etsxBeuP (the card session that works this card; claim written by the domain:ui seat 4 session session_01XWGTsmqrtrt1Zqh1ioWqsT)
Account: os-elon-musk
Branch: claude/project-thread-q1we3g (platform-assigned at spawn; the card session can push only this branch)
Worktree: objectui-issue-13464
Domain: domain:ui
Seat: domain:ui#4
File surface: the sideEffects field (that field only) in the published packages/*/package.json manifests that lack it; scripts/check-side-effects-array.mjs only if an array package's list must change and its own tests; one .changeset/13464-*.md naming every package it changes (stop on breach; explain in the report)
Container & model: M, mode:thread, model: default tier (dispatch-gates --tier --repo objectstack-ai/objectui: no path-derived mandate), effort: high
Clause-②: no
Metadata-protocol: none
Responsibility: n/a — not a defect card
Thread-read: 6387079537
Premise: holds as a census — on objectui main 23e912e the card's count (30 of 39 published manifests without sideEffects) is re-measured by the card session at its start; it works from its own list
Serial constraints cleared: seat 4's PR objectstack-ai/objectui#13481 (React peer narrowing, 31 manifests) and PR objectstack-ai/objectui#13486 (plugin-list / plugin-map / plugin-timeline manifests) are in the merge queue — merge main before opening and again before landing; seat 1's objectui#13442 / objectstack-ai/objectui#13449 / objectstack-ai/objectui#13450 declare no manifest field
Writes: all PR, comment and label writes go only through objectstack's relay (scripts/pm/fleet-write); never write with GitHub MCP.

Ruling section: filed under the maintainer's 「同意：建议新建：与合并无关、独立成卡的产品/契约项（可现在建）」; triage 6387079537: declare per package now, measured package by package; explicit register*() later, on its own card.
Per package, read before writing: false only after reading every entry module and finding no import-time mutation (registry register(…), global CSS, polyfills, window writes); a package that registers on import gets an array of exactly those modules (derive it with check-side-effects-array.mjs --list) or true. A wrong false drops a live registration (the objectui#6535 class): for each false, the PR names the read; for each array, the gate passes on it.
Measure: the console eager closure before and after (apps/console/dist/eager-closure.json, real production build), stated in the body. Lower no ceiling in this PR unless the drop is measured; the console first load: @objectstack/spec arrives as self-contained subpath bundles that duplicate each other; three tiny imports pull /system, /kernel and /api whole into the eager closure #25925 card owns the ceiling line concurrently — if both move it, the second to land re-measures.
⛔ Out: explicit register*() and deleting check-side-effects-array.mjs (the card's longer-term item); apps/console (not published).

Claim amendment · domain:ui#4 · 2026-10-25T02:32Z · amends the claim 6392391793 (append-only; the claim's other lines stand)

File surface, added on the card session's request: scripts/__tests__/side-effects-declaration-consistency.test.ts, for two things only:

grow its census of declaring packages by name;
teach its impureNode scan the carve-out the array gate already has (local-binding-write): a member write whose root binding is declared in the same module and whose right side calls nothing, e.g. X.displayName = '…'. Pin both directions: that shape passes, and a write to an imported or global binding still fails.

This keeps false honest for react, auth, mobile and permissions, whose only top-level statements are displayName writes on local bindings.

Readings accepted: the executables (cli, create-plugin, runner) and plugin-grid get true. plugin-grid's reachability chain admits no array under the gate, which the PR states.

Generated by Claude Code

os-dev-report

{
  "issue": 13464,
  "status": "done",
  "branch": "claude/project-thread-q1we3g",
  "pr": "https://github.com/objectstack-ai/objectui/pull/13495",
  "session": "session_01Q4K9dtZAX6vw67etsxBeuP",
  "premise_still_valid": true,
  "summary": "29 published manifests now declare sideEffects (apps/console out of scope). false for react, auth, collaboration, data-objectstack, mobile, permissions, providers after a full-src scan (only same-module displayName writes). Gate-derived arrays for 15 plugins. true for plugin-ai, plugin-chatbot, plugin-dashboard, plugin-grid (no array satisfies every guard: no bundler alias / pure barrels via export star / loop-registered keys / reachability chain) and for the executables cli, create-plugin, runner. Per the amended claim 6392496025 the consistency test gains the gate local-binding-write carve-out (pinned both directions) and its census grows.",
  "tests": "head fcfc372: vitest run side-effects-declaration-consistency + check-side-effects-array + check-sdui-registration-pins tests: Test Files 3 passed (3), Tests 172 passed (172). c16787a: vitest run --maxWorkers=2 scripts/ = 1 failed of 5857 (sdui-registration-pins could not read plugin-dashboard loop keys; fixed in fcfc372 by declaring true, re-run green above); export-to-react-preamble + tenancyPostureWall.parity: 9 passed. Reverse check on c16787a: carve-out line removed -> 2 failed / 113 passed (the false-package case and the pure-direction pin); restored from HEAD, git diff HEAD empty.",
  "gates": "local: check-side-effects-array exit 0 (17 packages); check-eager-closure-budget exit 0; check-sdui-registration-pins exit 0 (68 registrations present); check-changeset-presence exit 0; check-changeset-no-major exit 0; check-changeset-fixed exit 0; check:new-line-citations exit 0; eslint on the changed test exit 0. objectui has no dispatch-gates.mjs, list hand-derived. CI families: in_progress at report time (PR just opened).",
  "measurement": "console eager closure, real production build (turbo build packages + console build): before 6f629a7 3,106,386 B gzip 85/2291 chunks; after fcfc372 3,106,204 B gzip 85/2291 chunks (-182 B). No ceiling moved.",
  "dogfood_battery": "n/a: no ! in the PR title and no (narrowing) arm",
  "mcp_calls": "0",
  "api_writes": "2 relay strokes: pr_create objectui#13495 (assignee os-elon-musk); this os-dev-report comment + pr_request_reviewers",
  "open_questions": [],
  "out_of_scope_findings": [
    "carrier: none. The two sideEffects guards disagree on a pure barrel (array gate requires every entry form; consistency test refuses a named source entry with no load-time effect), so plugin-chatbot is true. Noted in PR Acceptance notes, not filed: a reading, no consumer-visible defect."
  ]
}

Generated by Claude Code

ACCEPT → PR objectstack-ai/objectui#13495 comment 6392817892 · domain:ui#4 · 2026-10-25T03:13Z. Landing under the maintainer's standing word; Part of, so after merge this card returns to pm:queue with the explicit register*() item and the check-side-effects-array.mjs retirement.

domain:devx#1 · os-elon-musk · session_01LKHDniRQSWjRJyDtacwpEd

Claim: PM loop round 1 · 2026-10-29T14:51Z
Session: pending (card thread, dispatched by seat session session_01LKHDniRQSWjRJyDtacwpEd; amended when it starts)
Account: os-elon-musk
Branch: pending (the card thread's harness-pinned branch; amended)
Worktree: objectstack-issue-25913
Domain: domain:devx
Seat: domain:devx#1
File surface: the sideEffects field of the published ui/packages/*/package.json manifests that lack one, one changeset per touched package, and scripts/check-side-effects-array.mjs only if an array package is added (stop on breach; explain in the report)
Container & model: M, mode:thread, model: default, effort: high
Clause-②: no — a manifest bundling hint, no packages/spec/src/** schema and no accepted set changes; the dev says so in the PR, and if a declaration would narrow what a consumer can import, it stops and reports
Metadata-protocol: none
Responsibility: 30 of 39 published ui packages declare no sideEffects, so no consumer can tree-shake them | check-side-effects-array.mjs covers only the two array packages | every external host that bundles the ui packages
Thread-read: 6430916285
Premise: measured 2026-10-24 at 7981384; the dev re-runs the manifest census on origin/main first
Serial constraints cleared: the dev lists open PRs touching any ui/packages/*/package.json before editing; changesets live in the root .changeset/ (standing constraint)

Seat pick on the full order with seat 2 (#20163), alternation (knock 6499301324). Scope = the card's first box only (per-package declarations). The explicit-registration rewrite stays a later decision card. Contract review: not triggered unless the dev declares Clause-②: yes.

Generated by Claude Code

domain:devx#1 · os-elon-musk · session_01LKHDniRQSWjRJyDtacwpEd

Release: session_01LKHDniRQSWjRJyDtacwpEd (devx seat 1) · the seat dispatched the stale first box, not the re-scope 6416621475; the card thread found that out on measurement and changed nothing · next: pm:queue (the re-scoped item waits on the root ci.yml constraint, 6430916285)

Seat correction. The dispatch order was written from the card body's first box. It should have been written from triage's re-scope 6416621475: only the retirement of check-side-effects-array.mjs remains on this card. The card thread (session_016WV4cyd9xXusapQ3CGmV1w, branch claude/project-thread-3hxw8u) measured before editing and stopped. Nothing was pushed.

Census at origin/main 77e993ea (every ui/packages/* and ui/apps/* manifest): every published ui/packages/* manifest already declares sideEffects.

false: 11 (auth, collaboration, core, data-objectstack, i18n, mobile, permissions, providers, react, react-runtime, types).
Arrays: 17 (layout + 16 plugins).
true: 7 (cli, components, create-plugin, fields, plugin-chatbot, plugin-dashboard, plugin-grid).
Absent only on @object-ui/test-support (private) and @object-ui/console (ui/apps). For the console, absent is the honest value: Vite applies a package's sideEffects to its own relative imports, so false would tree-shake the console's own CSS and register-plugins imports.

That matches objectui#13495 (landed 2026-10-25). The remaining item, retiring the gate, keeps the two pitfalls in 6430916285: the root ci.yml wiring is under the maintainer's ci.yml constraint, and check-sdui-registration-pins.mjs reads the script as a library.

Generated by Claude Code

domain:devx#2 · os-justin · session_01QSeKYmTYfaPbLQD4BJfomW

Maintainer ruling recorded · 2026-10-30T12:50Z. The question was put to the maintainer (os-justin) in seat session session_01QSeKYmTYfaPbLQD4BJfomW. It named the remaining item (retire ui/scripts/check-side-effects-array.mjs), the root ci.yml wiring and the library reader. The maintainer first said, verbatim, 「p3 的任务也可以处理」, then chose, verbatim: 「A 授权改 ci.yml (Recommended)」.

Option A as put to the maintainer:

The dev first measures whether side-effects-declaration-consistency.test.ts fully supersedes the gate.
If it does: delete the CLI, move the shared derivation into a library module, and drop the ci.yml step. The PR is Tier H and waits on maintainer approval.
If it does not: report back; the card closes not planned.

This lifts the standing ci.yml constraint for this card's one step only.

Claim: PM loop round 12 · 2026-10-30T12:50Z
Session: session_01QSeKYmTYfaPbLQD4BJfomW (mode:subagent: the dev shares this session ID; its identity is the branch)
Account: os-justin (the seat's linked user, as GET /user answers it; the card's assignee)
Branch: claude/issue-25913-retire-side-effects-array-gate
Worktree: objectstack-issue-25913
Domain: domain:devx
Seat: domain:devx#2
File surface (stop on breach; explain in the report):

ui/scripts/check-side-effects-array.mjs and its test ui/scripts/__tests__/check-side-effects-array.test.ts;
the readers that import or name it: ui/scripts/check-sdui-registration-pins.mjs (+ its test), ui/scripts/check-eager-closure-budget.mjs, ui/scripts/vite-declared-lazy-views.ts, and ui/scripts/__tests__/side-effects-declaration-consistency.test.ts (reference text only);
a new library module under ui/scripts/ if the shared derivation moves;
the check:side-effects-array line in ui/package.json;
.github/workflows/ci.yml L3562 (that one step only);
any lint.yml run_self_test line the self-test wiring gate forces. Code comments in ui/packages/app-shell/** that name the script are text only.
Container & model: M, mode:subagent, model: opus, effort: high; the contract review runs at CONTRACT_REVIEW_TIER (Tier H via .github/workflows/**), and the seat waits for that tier and never downgrades.
Clause-②: no (internal tooling; no published package changes)
Metadata-protocol: none
Responsibility: n/a — not a defect card (gate retirement, tooling)
Thread-read: 6500172459
Premise: holds at origin/main 6410029d1b. The gate is still wired at .github/workflows/ci.yml:3562 (pnpm --dir ui run check:side-effects-array). ui/package.json:71 defines the script. check-sdui-registration-pins.mjs still imports from it. Whether the consistency test fully supersedes it is the dev's first measurement.
Serial constraints cleared: the dev lists open PRs touching .github/workflows/ci.yml or ui/scripts/** before editing, and merges main before opening and again before landing.

Generated by Claude Code

domain:devx#2 · os-justin · session_01QSeKYmTYfaPbLQD4BJfomW

Dispatched under claim 6523253073 · 2026-10-30T12:51Z: an os-dev subagent (mode:subagent, effort high) on branch claude/issue-25913-retire-side-effects-array-gate.

Measurement comes first. The dev builds a table with one row per gate assertion, each showing where that assertion lives after retirement.

Any row it cannot cover: it stops, changes nothing, and reports premise_still_valid: false. The card then closes not planned, as option A said.
Every row covered: it moves the shared derivation into a library module, repoints the readers, migrates the library tests, and removes the one ci.yml step.

That PR is Tier H. It waits on the at-tier contract review and on maintainer approval. The report comes back here as os-dev-report.

Generated by Claude Code

domain:devx#2 · os-justin · session_01QSeKYmTYfaPbLQD4BJfomW
os-dev-report

{
  "issue": 25913,
  "status": "done",
  "branch": "claude/issue-25913-retire-side-effects-array-gate",
  "pr": null,
  "session": "session_01QSeKYmTYfaPbLQD4BJfomW",
  "premise_still_valid": false,
  "summary": "H1 FAILS: the consistency test does not supersede ui/scripts/check-side-effects-array.mjs, so per the maintainer's ruling (Option A, 「A 授权改 ci.yml (Recommended)」, 'if it does not: report back; the card closes not planned') nothing was changed and no PR was opened. Read at origin/main 59547058ae. The decisive LOST row is the gate's core MISSING-registrar check, and it was measured, not only read: deleting app-shell's './dist/console/diagnostics/CloudAiModelStatus.js' (a ruled control) from its sideEffects array makes the gate exit 1 (MISSING) while side-effects-declaration-consistency.test.ts stays 109/109 green; the same holds when the './src/...CloudAiModelStatus.tsx' line is deleted instead. Both texts already say this: the gate header (L224-235) calls the two guards 'deliberately independent', and the consistency test (L636-641, L1369-1370) hands 'is the set EXACTLY right, missing or stale' to the gate. H1 TABLE (gate assertion @ gate line → asserted where after retirement): R1 at least one array package, else exit 2 @L1291 → COVERED consistency L1083 'discovers the declaring packages'. R2 a non-string entry throws @L325 → COVERED L209. R3 an unsupported workspace glob throws @L302 → COVERED L184. R4 every published entry form is named @L1226 → COVERED L1320 'every sideEffects array covers all of its package's entry forms'. R5 the barrel's source spelling is named @L1227 → COVERED by L1258 + L1320 (the alias src form must exist and be declared). R6 MISSING: every registering module in the entry graph is named in both spellings @L1228-1235 → LOST (probes above; check-sdui-registration-pins derives its keys FROM the array, so it cannot see an unnamed registrar either). R7 STALE source spelling @L1236 → PARTIAL, L1377 'every declared source entry really does have a load-time side effect'; that classifier counts any call, write or bare import, where the gate counts registrations only. R8 STALE published (dist) spelling of a deep module @L1236 → LOST (the L1348 phantom check accepts any module form). R9 reachability, every registrar reached through covered modules @L1247 checkReachability → LOST. R10 an unknown top-level effect in an array package is refused with exit 2 @L1204 → LOST (the consistency test reds on effects only in false packages, L1291). R11 initializer registration, export const X = f() where f writes an escaping binding @L620/L705 → LOST (the consistency header L126-132 says it does not prove this). R12 every published subpath is a walk root @L1187-1192 → LOST (the consistency test walks the barrel only, L1027). R13 an unclassifiable published form is refused @L1109 → LOST. R14 the spelling map round-trips on the barrel @L974 → LOST. R15 a published barrel and a source barrel exist @L923/L937 → PARTIAL: source COVERED at L1242, published LOST. R16 an unresolved relative specifier is refused @L1194 → COVERED for the barrel graph at L1264. R17 a glob entry is refused @L1238 → COVERED L1428. R18 the per-package graph has at least 2 modules @L1255 → PARTIAL, only an aggregate floor of 150 at L1287. H2 reading (no edit): two importers. ui/scripts/check-sdui-registration-pins.mjs:64 imports deriveSpellingMap and readArrayPackages; ui/scripts/__tests__/check-side-effects-array.test.ts:10-20 imports EXIT_*, classifyEffect, evaluate, evaluatePackage, main, readArrayPackages and walkEntryGraph. Readers by name only: check-eager-closure-budget.mjs:162; vite-declared-lazy-views.ts:64, :362 and :592 (the last is author-facing remedy text naming 'pnpm check:side-effects-array'); pins :16/:103/:122/:134 and its test :39/:196; consistency :639/:736/:1370; app-shell register-builtin-designers.ts:39. ui/scripts has no lib/ dir (shared modules sit flat, e.g. component-registrations.mjs). Today the exact-enumeration condition and the spelling-map exit-2 condition have exactly one owner each: evaluatePackage (L1170-1276) and deriveSpellingMap (L921) in this gate, and pins L134 defers to it. Retirement would leave both with zero owners. H3 reading: holds but moot. The only wiring is .github/workflows/ci.yml:3561-3562. The step name and 'check:side-effects-array' appear nowhere else in .github/, scripts/ or ui/scripts (no lint.yml run_self_test line, no roster, no required-check name).",
  "tests": "No code changed. Baseline at 59547058ae under os-verify-lock: vitest run of side-effects-declaration-consistency + check-side-effects-array + check-sdui-registration-pins: Test Files 3 passed (3), Tests 170 passed (170), VERDICT command-exit 0. Baseline gate: node scripts/check-side-effects-array.mjs exit 0, 17 packages ok. Mutation probe 1, via scripts/ablation-replace.mjs (anchor x1 → x0, blob 31e89666 → a019a722), deleting app-shell './dist/console/diagnostics/CloudAiModelStatus.js': GATE_EXIT=1 'MISSING ./dist/console/diagnostics/CloudAiModelStatus.js'; consistency test TEST_EXIT=0, Tests 109 passed (109). Restored: blob == HEAD 31e89666, git diff HEAD empty. Mutation probe 2, the same with './src/console/diagnostics/CloudAiModelStatus.tsx' (blob → ee2a41de): GATE_EXIT=1 MISSING; TEST_EXIT=0, 109 passed. Restored: blob == HEAD, git diff HEAD empty. Each part's exit was captured on its own line, not taken from the batch verdict.",
  "gates": "dispatch-gates --commands --repo objectstack-ai/objectstack: exit 2, 'this branch changes nothing against origin/main (merge base 59547058a) — nothing to derive' (empty diff, so --ran has nothing to reconcile). No PR, so no check-runs.",
  "dogfood_battery": "n/a: no PR, no ! title and no (narrowing) arm",
  "mcp_calls": "0",
  "api_writes": "1: POST issues/25913/comments (this os-dev-report) via scripts/pm/post-stamped.mjs. Also one git push (an empty branch at 59547058ae, the landing marker), which is not a REST write.",
  "line_budget": "n/a: no file changed",
  "files_changed": [],
  "deviations": [
    "The worktree was cut from origin/main 59547058ae, not the claim's premise sha 6410029d1b, because main moved. Both premise lines still hold at the new base: ci.yml:3562 and ui/package.json:71.",
    "Mutation probe 1 joined its parts with ';', so os-verify-lock printed batch-last-exit. Per-part exits (GATE_EXIT, TEST_EXIT) were echoed and are what is cited. Probe 2 used '&&'-joined groups.",
    "The empty branch claude/issue-25913-retire-side-effects-array-gate remains on origin at 59547058ae with no commits. It was not deleted; deleting it is a PM call."
  ],
  "open_questions": [],
  "out_of_scope_findings": [
    "carrier: none. Noted, not filed: the card's remaining item contradicts the gate's own recorded design (two independent guards, header L224-235). Closing not-planned per Option A is the ruled outcome. Dedupe words: side-effects-array, sideEffects gate retirement, consistency test supersede, check-sdui-registration-pins."
  ]
}

Generated by Claude Code

domain:devx#2 · os-justin · session_01QSeKYmTYfaPbLQD4BJfomW

Closed not planned · 2026-10-30T12:58Z. This follows the maintainer's ruling recorded in 6523253073. The maintainer, os-justin, chose 「A 授权改 ci.yml (Recommended)」 in seat session session_01QSeKYmTYfaPbLQD4BJfomW. That option's second arm reads: "if it does not [fully supersede the gate]: report back; the card closes not planned."

Why the arm fired (os-dev-report 6523384102; the seat re-read the two texts at origin/main):

Measured. Deleting a registering module from app-shell's sideEffects array makes check-side-effects-array.mjs exit 1 (MISSING), while side-effects-declaration-consistency.test.ts stays green at 109/109. The probe was run twice, once with the dist spelling and once with the src spelling, and the tree was restored to HEAD each time.
By design. The gate's header (L224-235) says the two are "deliberately independent ... because two guards that share a derivation fail together". The consistency test (L636-641, L1369-1370) hands "is the set exactly right" to the gate. The dev's 18-row table shows 9 gate assertions LOST and 3 only PARTIAL after a retirement. They include missing-registrar enumeration, reachability, stale published spellings, initializer registrations and published subpath roots.
Library reader. check-sdui-registration-pins.mjs derives its keys from the arrays and imports deriveSpellingMap / readArrayPackages from the gate, so it cannot stand in for the missing-registrar check.

State:

Nothing changed and no PR was opened. The ci.yml step at L3562 stays.
The sideEffects declarations themselves landed in objectui#13495. Explicit register*() stays uncarded per triage 6416621475.
pm:dispatched comes off in this stroke.
The empty branch claude/issue-25913-retire-side-effects-array-gate (no commits) is left as it is.
Reopening is free; the maintainer can override.

Generated by Claude Code

Navigation Menu

Repository navigation

packages: 30 of 39 published packages declare no sideEffects, so consumers cannot tree-shake them; declare per package now, make registration explicit later so the 1,380-line array gate can go #25913

维护者速读

Measured (2026-10-24, HEAD `7981384`; manifest census)

Task

Triage: first grade — `tech-debt` · `priority:p3` · `domain:ui` · `area:devpath` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-24T17:06Z

22 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

packages: 30 of 39 published packages declare no sideEffects, so consumers cannot tree-shake them; declare per package now, make registration explicit later so the 1,380-line array gate can go #25913

Description

维护者速读

Measured (2026-10-24, HEAD 7981384; manifest census)

Task

Activity

objectstack-fleet commented on Oct 24, 2026

Triage: first grade — tech-debt · priority:p3 · domain:ui · area:devpath · pm:queue · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-24T17:06Z

objectstack-fleet commented on Oct 25, 2026

objectstack-fleet commented on Oct 25, 2026

objectstack-fleet commented on Oct 25, 2026

objectstack-fleet commented on Oct 25, 2026

22 remaining items

objectstack-fleet commented on Oct 29, 2026

objectstack-fleet commented on Oct 29, 2026

objectstack-fleet commented on Oct 30, 2026

objectstack-fleet commented on Oct 30, 2026

objectstack-fleet commented on Oct 30, 2026

objectstack-fleet commented on Oct 30, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Measured (2026-10-24, HEAD `7981384`; manifest census)

Triage: first grade — `tech-debt` · `priority:p3` · `domain:ui` · `area:devpath` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-24T17:06Z