Skip to content
This repository was archived by the owner on Nov 4, 2026. It is now read-only.

Repository navigation
This repository was archived by the owner on Nov 4, 2026. It is now read-only.

@object-ui/plugin-dashboard imports react-grid-layout's stylesheet at module scope — the last load-leg blocker, worth 2 entries not 1 #5384

Closed
Task
#5455
Closed
@object-ui/plugin-dashboard imports react-grid-layout's stylesheet at module scope — the last load-leg blocker, worth 2 entries not 1#5384
Task
#5455
Assignees
os-support-ai
Labels
pm:dispatched

Description

@os-support-ai
os-support-ai
opened on Aug 20, 2026

Filed by the repo:objectui execution seat (PM round 19, session session_01RV6yuVCxymHYE16PL9vQkE) while accepting PR5383 (#5357). Filed unassigned and unlabelled — grading and routing belong to the triage seat. ⛔ Not a grading.

This is the follow-up the accepting seat owes, recorded now rather than left in a PR body.

Why it is worth a card of its own

pnpm check:node-esm-load stands at 34 of 39 published ESM entries. After #5214 and #5357 cleared the entire extensionless-specifier class, the remaining 5 break down as:

package reason ledger
@object-ui/plugin-dashboard imports react-grid-layout's stylesheet at module scope; Node cannot load .css at all LOAD_DEBT
@object-ui/app-shell reaches that same stylesheet through static @object-ui/plugin-dashboard imports in DashboardView.tsx and ReportView.tsx LOAD_DEBT
@object-ui/plugin-map imports maplibre-gl's stylesheet; same question LOAD_DEBT
create-plugin bin-only CLI not importable by design
runner built web app not importable by design

So this single change would clear two entries, not one — plugin-dashboard and app-shell together, taking the load leg 34 → 36. That is by a wide margin the largest remaining unlock, and per PR5383's dev it is "the only one left that is a code change rather than a design decision."

⚠️ The app-shell half was invisible until this week. While app-shell emitted extensionless specifiers its first error was ERR_MODULE_NOT_FOUND on its own file, which masked the .css failure entirely — the same first-failure shape #5214 measured when it predicted 5 red and observed 10. It only became legible once PR5383 removed the mask.

The question, which is not obviously one shape

Node cannot evaluate a .css import at all, so this is not a spelling fix. Plausible directions, not a recommendation — the choice is the graders':

  • Move the stylesheet import off module scope (a consumer-side import the host performs, or a documented CSS entry the package advertises rather than executes).
  • Decide that unbundled Node consumers are not supported for style-carrying plugin packages, and make LOAD_DEBT say that as a permanent statement rather than as debt.
  • Something upstream-shaped for react-grid-layout / maplibre-gl specifically.

The second option is a legitimate answer and would make the ledger honest rather than aspirational. ⚠️ Note it also settles plugin-map, so whichever way this goes it should be taken over the style-carrying packages as a group, not one at a time.

Why LOAD_DEBT is the right place for these three today, and what it costs

LOAD_DEBT is deliberately kept separate from SPECIFIER_DEBT — its own header says merging them "would let this gate become a grab-bag in which the defect it exists for stops being visible." These three genuinely answer a different question (is an unbundled consumer supported at all?), so the entries are truthful.

The cost, recorded honestly by PR5383's dev rather than discovered later: LOAD_DEBT matches by package name, not by error code. So the load leg alone would now also stay quiet about a genuine specifier regression in any of these three. What carries that guard instead is the specifier leg, which became a hard requirement the moment SPECIFIER_DEBT emptied — PR5383's ablation confirmed it reddens on a reverted app-shell while the load leg prints ledgered.

That is a working arrangement, not a hole — but it is one more reason this card is worth closing rather than living with: every package in LOAD_DEBT is a package the load leg cannot speak for.

Related

Activity

  1. os-zhuang
    added
    needs-user-decision
    on Aug 20, 2026
  2. os-zhuang
    added the
    Task
    issue type on Aug 20, 2026

  3. os-zhuang commented on Aug 20, 2026

    @os-zhuang
    os-zhuang
    on Aug 20, 2026 – with Claude
    Contributor

    Triage: entered the decision inbox (needs-user-decision, type Task). The ruling should be taken over the three style-carrying packages (plugin-dashboard / plugin-map / the family) at once — it also decides whether their LOAD_DEBT entries are debt or permanent.

    <!-- os-decision-facets -->

    • 项目长远合理性:A(把 CSS 移出模块作用域,给样式一个显式入口)延续 ESM 可加载性程序的方向(34→36/39),收敛特例;B(声明带样式插件不支持裸 Node)把两条 ledger 债转成永久豁免,特例扩大且永远要人记得为什么。
    • 实际业务拉动:今天没有终端用户撞上 —— 拉动来自门禁台账(LOAD_DEBT 清零)与未来 SSR/Node 直载场景;拉动是真实的但不紧迫。
    • 防 AI 犯错:A 之后「样式怎么进来」是声明式的(显式 import / 显式样式入口),bundler 魔法消失;B 让「在 Node 下坏」成为被允许的隐性状态,AI 复制该形态到新插件不会被任何门拦住。
    • 创业阶段不扩散:A 的边界是三个包各一处改动 + 消费端一行文档;B 零代码但买入永久台账义务。两者都不扩散需求,A 略贵但一次性。
    • 推荐:A(三包一次做完,样式入口显式化;changeset 注明消费者需显式引样式)。
    • 置信缺口:本分析看不见现有消费者对隐式样式注入的依赖面 —— 若有已发布应用依赖 bundler 隐式注入,A 是一次 breaking,需要 changeset major 语义与迁移一行;这一点只有维护者/实测能答。

    Generated by Claude Code

  4. os-zhuang commented on Aug 20, 2026

    @os-zhuang
    os-zhuang
    on Aug 20, 2026 – with Claude
    Contributor

    Maintainer ruling recorded (2026-08-20, live decision-inbox session with the triage seat, session session_01PjAP6vbcsg2yMtvySPv1Qo)

    Ruled: the second direction — unbundled Node consumption is NOT supported for style-carrying plugin packages, as a permanent statement rather than debt. Provenance: maintainer accepted the batch, verbatim: 「其他接受你的建议。」Taken over the group as the card asks: @object-ui/plugin-dashboard, @object-ui/app-shell (via its static imports), and @object-ui/plugin-map together. No known unbundled-Node consumer exists; converting a zero-pull capability gap into permanent maintained machinery is the shape the startup rule rejects.

    Implementation: reword LOAD_DEBT for these three from debt to a stated permanent boundary (the ledger becomes honest rather than aspirational), keep the specifier leg as the hard guard it already is, and note the boundary in the packages' READMEs so a future Node consumer learns it from the package, not from a red import. Revisit only on a real consumer request — which would reopen this as a design card, not a debt payment.

    State transition in the same stroke: needs-user-decision → pm:queue.

    Generated by Claude Code

  5. os-zhuang
    added
    pm:queue
    and removed
    needs-user-decision
    on Aug 20, 2026
  6. os-support-ai
    added
    pm:dispatched
    and removed
    pm:queue
    on Aug 20, 2026
  7. os-support-ai
    self-assigned this
    on Aug 20, 2026

  8. claude commented on Aug 20, 2026

    @claude
    claudeBot
    on Aug 20, 2026 – with Claude
    Contributor

    Claim: PM loop round 21
    Session: session_01RV6yuVCxymHYE16PL9vQkE
    Branch: claude/issue-5384-loaddebt-permanent-exemption
    Worktree: objectui-issue-5384
    Domain: repo:objectui
    File surface: the LOAD_DEBT ledger + its gate/docs; ⛔ NOT the CSS imports (stop on breach; explain in the report)
    Container & model: S, mode:subagent
    Clause-②: no
    Serial constraints cleared: none

    Generated by Claude Code

  9. os-support-ai
    added a commit that references this issue on Aug 20, 2026

    docs(esm-load-gate): state the unbundled-Node boundary instead of car…

    77d894b

  10. os-support-ai commented on Aug 20, 2026

    @os-support-ai
    os-support-ai
    on Aug 20, 2026 – with Claude
    CollaboratorAuthor
    {
  "issue": 5384,
  "status": "done",
  "branch": "claude/issue-5384-loaddebt-permanent-exemption",
  "pr": "https://github.com/objectstack-ai/objectui/pull/5455",
  "premise_still_valid": true,
  "summary": "Reworded the LOAD_DEBT ledger from debt into a stated permanent boundary over the three style-carrying packages as a group, per the ruling — the CSS imports are untouched (that was option A, ruled against). The map is now UNBUNDLED_NODE_UNSUPPORTED; its docblock carries the ruling, its date, the verbatim provenance, the measured failure (ERR_UNKNOWN_FILE_EXTENSION for all three), and what would reopen it; each entry states its own boundary and cites objectui#5384; the run prints them as `by design` rather than `ledgered` plus a named summary line. The ruling's other two clauses are done too: the specifier leg is untouched and is documented as the guard that carries these three, and plugin-dashboard/plugin-map/app-shell each gained a 'Requires a bundler' README section with the real error text. A new unit pin makes a bare, un-cited exemption fail the suite. The ruling's premise was re-measured, not inherited: every consumer goes through a bundler (vite; Next transpilePackages, not serverExternalPackages), examples/schema-catalog only imports JSON files from plugin-named directories, and the sibling objectstack repo's four hits are all string literals — every zero counter-probed with a term known present.",
  "tests": "Local union re-run at HEAD=77d894b76 (tree clean), all green: check:esm-specifiers ('no un-ledgered package emits an extensionless relative specifier'), check:control-bytes (4510 files), changeset:check, docs:check-links ('valid across 13 scan roots'), type-check:scripts, check:doc-snippets ('87 of 87 block(s) judged, 0 failed'), vitest check-node-esm-load.test.ts + ci-cd-pipeline-doc.test.ts ('Tests 62 passed (62)'), and the full load leg. BEFORE/AFTER counts, measured over ONE build of every published package (turbo: '43 successful, 43 total') by running the origin/main gate and this branch's gate against the SAME artifacts with --no-build: before = 'Load leg: 34 of 39 published ESM entries imported and evaluated', exit 0; after = 'Load leg: 34 of 39 published ESM entries imported and evaluated', exit 0. THE NUMBER DOES NOT MOVE, as this card intends — only the three lines change from 'ledgered …' to 'by design …' plus a new summary line naming them. All three fail with ERR_UNKNOWN_FILE_EXTENSION, confirming the ledger's stated reason. Reverse verification of the new pin, predicted direction 'reddens' and observed as predicted: stripping 'by ruling objectui#5384' from the plugin-map entry gave 'Tests 1 failed | 29 passed (30)' on expect(reason).toMatch(/objectui#\\d+/); restoring gave '30 passed (30)'. No rebuild applies to this ablation and none was needed — the suite imports the gate as a source .mjs by relative path, not through a package exports map, so no dist/ sits between the edit and the assertion; the fix was committed before the mutation and restored by git checkout, verified byte-identical by a clean git status.",
  "open_questions": [],
  "out_of_scope_findings": []
}

    Generated by Claude Code

  11. os-support-ai commented on Aug 20, 2026

    @os-support-ai
    os-support-ai
    on Aug 20, 2026 – with Claude
    CollaboratorAuthor

    os-dev-report

    (Supersedes the previous comment, whose HTML-comment marker the GitHub sanitizer stripped on write — the marker is spelled as literal text here so a PM scan can see it. Same report, unchanged.)

    {
  "issue": 5384,
  "status": "done",
  "branch": "claude/issue-5384-loaddebt-permanent-exemption",
  "pr": "https://github.com/objectstack-ai/objectui/pull/5455",
  "premise_still_valid": true,
  "summary": "Reworded the LOAD_DEBT ledger from debt into a stated permanent boundary over the three style-carrying packages as a group, per the ruling — the CSS imports are untouched (that was option A, ruled against). The map is now UNBUNDLED_NODE_UNSUPPORTED; its docblock carries the ruling, its date, the verbatim provenance, the measured failure (ERR_UNKNOWN_FILE_EXTENSION for all three), and what would reopen it; each entry states its own boundary and cites objectui#5384; the run prints them as `by design` rather than `ledgered` plus a named summary line. The ruling's other two clauses are done too: the specifier leg is untouched and is documented as the guard that carries these three, and plugin-dashboard/plugin-map/app-shell each gained a 'Requires a bundler' README section with the real error text. A new unit pin makes a bare, un-cited exemption fail the suite. The ruling's premise was re-measured, not inherited: every consumer goes through a bundler (vite; Next transpilePackages, not serverExternalPackages), examples/schema-catalog only imports JSON files from plugin-named directories, and the sibling objectstack repo's four hits are all string literals — every zero counter-probed with a term known present.",
  "tests": "Local union re-run at HEAD=77d894b76 (tree clean), all green: check:esm-specifiers ('no un-ledgered package emits an extensionless relative specifier'), check:control-bytes (4510 files), changeset:check, docs:check-links ('valid across 13 scan roots'), type-check:scripts, check:doc-snippets ('87 of 87 block(s) judged, 0 failed'), vitest check-node-esm-load.test.ts + ci-cd-pipeline-doc.test.ts ('Tests 62 passed (62)'), and the full load leg. BEFORE/AFTER counts, measured over ONE build of every published package (turbo: '43 successful, 43 total') by running the origin/main gate and this branch's gate against the SAME artifacts with --no-build: before = 'Load leg: 34 of 39 published ESM entries imported and evaluated', exit 0; after = 'Load leg: 34 of 39 published ESM entries imported and evaluated', exit 0. THE NUMBER DOES NOT MOVE, as this card intends — only the three lines change from 'ledgered …' to 'by design …' plus a new summary line naming them. All three fail with ERR_UNKNOWN_FILE_EXTENSION, confirming the ledger's stated reason. Reverse verification of the new pin, predicted direction 'reddens' and observed as predicted: stripping 'by ruling objectui#5384' from the plugin-map entry gave 'Tests 1 failed | 29 passed (30)' on expect(reason).toMatch(/objectui#\\d+/); restoring gave '30 passed (30)'. No rebuild applies to this ablation and none was needed — the suite imports the gate as a source .mjs by relative path, not through a package exports map, so no dist/ sits between the edit and the assertion; the fix was committed before the mutation and restored by git checkout, verified byte-identical by a clean git status.",
  "open_questions": [],
  "out_of_scope_findings": []
}

    Generated by Claude Code

  12. os-support-ai
    closed this as completedin #5455on Aug 21, 2026
  13. akarma-synetal
    added a commit that references this issue on Aug 23, 2026

    docs(esm-load-gate): state the unbundled-Node boundary instead of car…

    cfcff30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

pm:dispatched

Type

Task

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions