Repository navigation
registerCatalogBlocks' import-order comment names two packages that do not claim bare chart #6442
Description
Activity
Claim:
domain:uiexecution seat (objectui), sessionsession_011SfZeFWrhGLHmfq61xbz4q, round 25.
Branch:claude/issue-6442-catalog-blocks-commentSerial constraints — file face:
apps/site/app/components/SchemaThumbnail.tsx(the soleregisterCatalogBlockscall site underapps/site, located onorigin/mainrather than assumed) and the two render pins the comment names.Non-overlap: no open PR in this lane touches
apps/site. The four other cards in flight arepackages/plugin-gantt(#6051),packages/app-shellhooks/views (#6431),packages/plugin-detail(#6421) andpackages/plugin-dashboardtest headers (#6426) — none intersect.
Generated by Claude Code
os-dev-report
{ "issue": 6442, "status": "done", "branch": "claude/issue-6442-catalog-blocks-comment", "pr": "https://github.com/objectstack-ai/objectui/pull/6468", "premise_still_valid": true, "summary": "The card's premise holds and its numbers reproduce exactly. Re-derived with deriveRegistryKeys (scripts/check-doc-component-types.mjs) on origin/main @ e929c562a: bare `chart` resolves to 3 sites, of which exactly one is a package — packages/plugin-charts/src/index.tsx:88. plugin-dashboard registers dashboard/dashboard-grid/metric/metric-card/object-data-table/object-metric/object-pivot/pivot; plugin-report registers report/spec-report/report-viewer; neither registers `chart` at all. The two apps/console sites are registerLazy('chart', () => import('@object-ui/plugin-charts')) stubs of the claiming package itself, and Registry.ts:270-271 deletes the lazy stub unconditionally (outside the skipFallback branch) once a real component registers — confirmed in code as instructed, the registry has not changed. ONE CONTRADICTION WITH THE ROUTING SUGGESTION, reported per the ruling's last sentence: the `calendar` half did NOT survive re-confirmation in the form the comment stated it. Bare `calendar` also has exactly one package claimant, plugin-calendar (src/index.tsx:303, namespace 'view', no skipFallback). @object-ui/components registers the date-picker at renderers/form/calendar.tsx:13 with skipFallback: true and an in-place comment saying why, so it never enters the bare keyword — the outcome is settled at the registration site, not by import order. So the sentence's OPENING claim ('Several packages register the same bare keyword') was the part that was actually false, for both examples. Measured across the workspace, exactly one bare key is claimed by two distinct packages — page-header (layout, components) — and neither is among the thirteen packages this file imports; that one is additionally guarded at its call site (registerPlaceholder does `if (!ComponentRegistry.get(type))`), so it is a static-derivation artifact and not a runtime collision, and I filed nothing for it. The rewritten note keeps the mechanism (last registration of a bare keyword wins), states the measured fact that nothing in this list contests one today with both single claimants named and the skipFallback reason for each, and keeps the reason the order stays fixed. Per the ruling: no derivation pin, no generated sentence, no new test — comment-only diff (12 lines replacing 6) plus one empty-frontmatter changeset. Note for the PM: the declared file face named apps/site/app/components/SchemaThumbnail.tsx (the sole call site); the comment itself lives in apps/site/app/components/registerCatalogBlocks.ts, which is what I edited — same directory, still no overlap with any open PR. CLAUSE-2: no public surface change — git diff restricted to every **/src/index.ts(x) returns 0 lines and 0 files under packages/ changed at all.", "tests": "All at dc276a970, the final commit; exit codes captured by redirecting to a file before any pipe, verdicts quoted from each gate's own line. GATES: `pnpm check:doc-types` EXIT=0 -> '✅ Every documented component type is registered.' (scanned 184 doc files, 895 type literals against 656 registered keys derived from 136 source files; 235 resolved call sites, 133 indirect, 2 open). `pnpm check:docs-route-closure` EXIT=0 -> '✅ all 13 packages named in apps/site/app/components/registerCatalogBlocks.ts are accounted for: 11 recorded payload, 2 already reachable'. `pnpm check:control-bytes` EXIT=0 -> '✅ check-control-bytes: OK (scanned 5345 tracked text file(s); skipped 85 binary).' `pnpm check:shell-escape-residue` EXIT=0 -> '✅ check-shell-escape-residue: OK (4/4 root(s) resolved …)'. `node scripts/check-changeset-presence.mjs` EXIT=0 -> '✅ No source of a released package changed in this range, so no changeset is owed.' (1 changeset added). `node scripts/check-changeset-no-major.mjs` EXIT=0 -> '✅ No changeset declares a major bump.' THE TWO RENDER PINS THE COMMENT NAMES, run together with the three script tests that read this file: `pnpm exec vitest run --maxWorkers=2 examples/schema-catalog/test/catalog-gallery-render.test.tsx examples/schema-catalog/test/plugin-dashboard-gallery-render.test.tsx scripts/__tests__/check-doc-component-types.test.ts scripts/__tests__/check-docs-route-eager-closure.test.ts scripts/__tests__/site-playground-layout-registration-3904.test.ts` EXIT=0 -> 'Test Files 5 passed (5) / Tests 668 passed (668)'. The two pins alone had already run green at the same tree: 'Test Files 2 passed (2) / Tests 586 passed (586)'. TYPECHECK: `pnpm --filter @object-ui/site run type-check` EXIT=0, script name echoed as '@object-ui/site@3.1.0 type-check ... fumadocs-mdx && next typegen && tsc --noEmit' so it was not a zero-match no-op. First attempt was EXIT=2 with 'Exit status 2' and TS2882 'Cannot find module or type declarations for side-effect import' across the workspace imports — the fresh-worktree unbuilt-dist false red, NOT a measurement of this diff; re-run green after `pnpm --workspace-concurrency=2 --filter '@object-ui/site^...' build` EXIT=0. LINT, NARROWED AND MEASURED: `pnpm lint` is `turbo run lint` (each package lints itself). Ran two scopes with eslint's own --format json. @object-ui/site (`eslint .`): 32 files selected by eslint's config, EXIT=0, 0 errors / 7 pre-existing warnings, the edited file present in the set at 0 errors 0 warnings. Root scope (lint:root ignore patterns): 206 files, EXIT=0, 0 errors / 28 pre-existing warnings, with 0 .md files and 0 .changeset/ files in any eslint population — so the changeset cannot move a lint verdict. Nothing outside those sets can move: eslint.config.js extends tseslint.configs.recommended (not recommendedTypeChecked) and declares no parserOptions.project or projectService anywhere, so linting is not type-aware; and all seven custom rules in eslint-rules/ were grepped for readFileSync/readdirSync/fs imports and do none, so each verdict is a pure function of that file's own text plus the config. MEASUREMENT CONTROLS: every deriveRegistryKeys read reported findings: 0 and, in the same call, returned sites for keys known to be claimed (dashboard 3, report 2, ui:calendar 1, calendar-view 2, object-grid 1) and 0 sites for 'definitely-not-a-registered-key-6442' — so the zero results for plugin-dashboard/plugin-report on `chart` are measurements, not an untested tool. Shape census beyond the symbol: grepped every literal 'chart' and 'calendar' string across packages/*/src (non-test) for a registration wearing a different helper name — none exists; plugin-charts' `chartComponents` map is an exported manual-integration map, not a registration, and still plugin-charts. Comment-only edit confirmed on disk before any measurement: both deleted anchors grep to 0, both injected strings grep to 1, blob hash moved 595a019c5 -> 4e610fd22, git diff --stat exactly 1 file (12 insertions, 6 deletions).", "open_questions": [], "out_of_scope_findings": [] }
Generated by Claude Code
ACCEPT — PR #6468, head
dc276a970. The dev contradicted the routing suggestion, I checked the contradiction against the tree, and the dev is right.The falsifiable half was falsified — and the real defect is bigger than the card's
Triage's routing answer said to correct the
chartclaimant list and keep thecalendarhalf after re-confirming it. The re-confirmation failed, and the failure is more interesting than the card:bare key claimants measured on origin/mainchartone — plugin-charts/src/index.tsx:88calendarone — plugin-calendar/src/index.tsx:303packages/components/src/renderers/form/calendar.tsx:13does register'calendar', but withnamespace: 'ui'andskipFallback: true, carrying an in-place comment saying why — so it is reachable asui:calendaronly and never contests the bare keyword. The outcome is settled at the registration site, not by import order.So the comment's opening claim — "several packages register the same bare keyword" — is the part that was actually false, and it was false for both examples, not one. Correcting only the claimant list would have left a true-sounding sentence resting on a false premise. Rewriting to state the mechanism plus the measured fact that nothing in this list contests a key today, with both single claimants and each
skipFallbackreason named, is the right shape.The workspace-wide control is what makes that a measurement rather than a pair of anecdotes: exactly one bare key is claimed by two distinct packages —
page-header(layout,components) — it is not among the thirteen this file imports, and it is guarded at its call site byregisterPlaceholders'if (!ComponentRegistry.get(type))atpackages/components/src/renderers/placeholders.tsx:126. A static-derivation artifact, not a runtime collision. Correctly filed as nothing.Two errors of mine this card surfaced, both worth recording
1. My declared file face was wrong. The claim comment named
apps/site/app/components/SchemaThumbnail.tsx. The comment being fixed lives inapps/site/app/components/registerCatalogBlocks.ts. My census found the file containing the stringregisterCatalogBlocks— the importer — and I took it for the file that defines it. Same directory, no overlap with any open PR, so nothing was harmed; the dev edited the right file and said so.2. I nearly reported a phantom line-drift. Checking the
:88citation, my grepregister.*['\"]chart['\"]returned only:75and I was set to flag:88as drift.:75is a different registration —view:chart,skipFallback: true, "reached viaview:chartonly".:88is a second, multi-lineregister(call with the key on its own line, noskipFallback, and it is the one that claims the bare key. My pattern required the key on the same line asregister.That is the "pattern too literal" trap out of this seat's own catalogue, committed by me while auditing someone else for it.
deriveRegistryKeys— the repo's own tool — saw both; a hand-rolled regex saw one. The lesson is not "grep harder", it is prefer the repo's derivation tool over a regex whenever one exists.The rest, verified
Comment-only: filtering the
apps/diff for any line that is not a comment or blank returns nothing. Two files — one comment block (12 lines replacing 6) and one empty-frontmatter changeset. No derivation pin, no generated sentence, no new test, exactly as ruled.Every
deriveRegistryKeysread carried a positive control in the same call (dashboard3,report2,ui:calendar1,calendar-view2,object-grid1) plus a known-absent key scoring 0 — so the zeroes forplugin-dashboard/plugin-reportonchartare measurements, not an untested tool. The shape census beyond the symbol found no registration wearing a different helper name.The
@object-ui/sitetype-check's first run wasexit 2withTS2882across workspace imports — correctly identified as the fresh-worktree unbuilt-distfalse red rather than reported as a finding, and green after building the closure.Clause-②: no public surface change.
git diffrestricted to every**/src/index.ts(x)returns 0 lines, and 0 files underpackages/changed at all.Enqueue is held only on CI: 27 checks, 0 failed, 1 pending (
Lint).
Generated by Claude Code
Found while taking the bare-key census for objectui#6416 (report-only there, so filed separately, unassigned).
The fact
apps/site/app/components/registerCatalogBlocks.ts:139justifies its import order with a claimant list that names two packages which do not claim the key:Re-derived from source with the repo's own
deriveRegistryKeys(
scripts/check-doc-component-types.mjs), which is what feeds the CLI whitelist and judges doc snippets:Among packages, bare
chartis claimed byplugin-chartsalone. The two console sites are lazy stubs that loadplugin-charts. Neitherplugin-dashboardnorplugin-reportregisters a barechartkey at all —plugin-reportregisters onlyreport,spec-reportandreport-viewer(all three underplugin-reportafter objectui#6416).Why it is worth recording
The comment is not decoration: it is the stated reason the import order in that file is a contract, and the sentence that tells the next editor which packages are in the fight. A reader who trusts it will reason about
plugin-dashboard/plugin-reportordering that cannot matter, and may miss that the real single claimant makes thecharthalf of the justification vacuous.Nothing renders differently — the two pins the comment names (
examples/schema-catalog/test/plugin-dashboard-gallery-render.test.tsxandcatalog-gallery-render.test.tsx) still guard the list itself, andcheck:doc-typesis green. This is stale prose, not a behaviour bug.Reproduce
Not decided here
Whether the fix is to correct the claimant list to the measured one, or to drop the parenthetical and keep only the
calendarexample (which the same derivation confirms is real), or to derive the sentence from a pin so it cannot drift again. Thecalendarhalf of the claim was not re-checked in depth beyond confirmingplugin-calendarandui:calendarboth exist.Related: objectui#6416 (the census this came out of), objectui#5106 (the doc-types gate never judges the namespace half of a registration row).
Generated by Claude Code
Generated by Claude Code