Uh oh!
There was an error while loading. Please reload this page.
test(parity): adapt the spec parity gates to 17.0.0 GA (ruled items 1-5); fork the four-block exemption - #4671
Merged
Conversation
… new keys Pre-re-point adaptation for the GA pin bump (#4636 / PR #4639), implementing objectui#4648's delegated ruling for the items whose measurement held. - registry-inputs parity: `EXPECTED_COVERED` becomes pin-aware, so the four `object-*` blocks GA adds to `ComponentPropsMap` are expected exactly when the installed spec carries them, with the pin-dependence asserted rather than assumed. - five GA-added keys the renderers already honour (`page:header.maxVisible` / `.mobileMaxVisible`, `page:tabs.alwaysShowStrip`, `record:details.inlineEdit` / `.showHeader`) get GA-pending unpublished-key exemptions: declaring the inputs on the rc.6 pin would fail this file's own forward direction, so #4668 owns declaring them once the pin moves. - `sectionFields.spec-parity`: `FormFieldSchema.publicPicker` gets a reasoned exemption (ruling item 5) — it gates objectstack's public-lookup REST route and has zero read points here. Both pending mechanisms are pinned sets whose dormancy is asserted to track the installed spec's element set in both directions, so an entry cannot rot into a silent allowlist. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RnQd8iMMUwXQEV1crFmQiQ
…tes read GA's refusal, not rc.6's silent strip objectui#4648 ruling items 3 and 4, measured. Neither key set moved between 17.0.0-rc.6 and 17.0.0 GA and objectui publishes neither key, so there was nothing to stop publishing; what moved is HOW the spec says no. GA closed `RecordHighlightsProps` and the `record:details` section object under objectstack#4001 batch A, so an undeclared key now raises `unrecognized_keys` with a named message where rc.6 dropped it in silence. Both gates now assert the same verdict through whichever evidence the installed contract gives, with the mode probed behaviourally rather than read off a version string. A broken probe cannot go vacuously green: taking the wrong branch fails on either pin. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RnQd8iMMUwXQEV1crFmQiQ
The latest updates on your projects. Learn more about Vercel for GitHub. |
Contributor
✅ Console Performance Budget
📦 Bundle Size Report
Size Limits
|
The empty-frontmatter changeset is this repo's explicit exemption for a diff that touches released packages' src/ without changing published behaviour — the same answer PRs #4641 and #4666 gave the same gate. Two of the four touched test files live under plugin-form / plugin-detail src/, which the gate counts. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RnQd8iMMUwXQEV1crFmQiQ
Contributor
✅ Console Performance Budget
📦 Bundle Size Report
Size Limits
|
yinlianghui
marked this pull request as ready for review
August 15, 2026 07:04
This was referenced Aug 15, 2026
Uh oh!
There was an error while loading. Please reload this page.
This was referenced Aug 15, 2026
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Part of #4648 — the ruled items whose measurement held. The card's headline item (reasoned exemptions for the four new GA
object-*blocks) is not implemented here: the implementation measurement contradicted the ruling's premise, so it goes back as a fork report rather than being re-decided in a diff. The card stays open for that.Pre-re-point adaptation for the GA pin bump (#4636 / PR #4639), alongside PR #4660 (symbol collisions) and #4649 (retirement unwind).
Per-item disposition
object-form,object-grid,object-master-detail-form,object-metric)element:record_picker.filterrecord-picker.tsxcomposescomposed?.filter ?? props.filterand the fetch effect setsquery.$filter = filterbeforeadapter.findelement:text_input.defaultValuetext-input.tsx:73-76seeds the bound page variable once on mount,:119passes it as the uncontrolled initial valuerecord:highlightstop-levelreadonlyfields/layout/ariatop-level andreadonlyper entry), and this repo never published it. What moved is the refusal moderecord:detailsstripped section member keyname/label/columns/fields); the renderer-onlytitle/showBorder/hideEmptyare the undeclared three, and they are now refused instead of strippedFormFieldSchemakeypublicPicker; zero read points repo-wide — it gates objectstack's public-lookup REST route (403 LOOKUP_NOT_PUBLICwithout it, objectstack#7467), a server capability with no destination innormalizeSectionFieldsectionFields.spec-parity$like/$ilikepage:header.maxVisible/.mobileMaxVisible,page:tabs.alwaysShowStrip,record:details.inlineEdit/.showHeaderare all honoured todayThe forked item, in one paragraph
The ruling chose reasoned exemptions on the measured premise that nothing in this repo reads the four blocks' authoring surfaces. Measured in implementation, the opposite holds: all four blocks are long-standing objectui renderers, and every one of the 78 keys their
inputsomit is honoured today — either read offschema.*in the renderer, or delivered bySchemaRenderer's rest-prop spread onto a component whose props declare it one for one.ObjectGridSchemain@object-ui/typesdeclareslabel/fields/sort/pagination/pageSize/showSearch/selectable/staticDataand the rest of GA's list (several as its own@deprecatedlegacy spellings, which GA has now declared);ObjectMetricWidgetdeclaresformat/currency/prefix/suffix/trend/compareTo/colorVariant/invert/fallbackValue; the form renderers declaresubmitText/cancelText/showSubmit/readOnly/initialValuesand their siblings.That inverts the defect class. These are not keys the renderer ignores (the silent-success class the ruling wanted to close) — they are keys that work while
sdui-parserreportsunknown-propon the author who writes them, the designer panel hides them andsdui-intrinsics.d.tsdenies them. Under this gate's own stated bar ("a spec key the renderer HONOURS andinputsomits is a plain defect and gets declared") they are A-class defects to declare. Writing the exemptions anyway would have meant acting on a falsified premise and rewriting the bar that governs the list. Both are the maintainer's call, so the measurement goes back to #4648 and nothing was added toUNPUBLISHED_EXEMPTIONSfor those four blocks — a comment block in the file says so, so the next reader does not "fix" the red by quietly adding them.Consequence, stated plainly: on a GA tree this PR takes the parity suites from 8 failures to 4, and the 4 remaining are exactly that forked item. It does not make the GA tree fully green, and it cannot until #4648 rules again.
What the code does
apps/console/src/__tests__/registry-inputs-spec-parity.test.tsEXPECTED_COVEREDbecomes pin-aware: the four GA blocks are expected incoveredexactly when the installed spec carries them. The expectation stays EXACT on either pin — the shrinking-contract reading is not weakened — and the pin-dependence itself is asserted, together with the pin-INDEPENDENT half (this repo registers all four withinputseither way, so a block leavingcoveredcan only ever mean the spec stopped carrying it).UNPUBLISHED_EXEMPTIONS, one per honoured-but-unpublished key, each naming its read site and page:header / page:tabs / record:details: five keys spec 17.0.0 GA declares and the renderers already honour are never published as inputs #4668. They are dormant on the rc.6 pin and fully judged on a GA tree; the set allowed to be dormant is pinned by name, and its dormancy is asserted to track the installed element set in both directions, so a GA release that dropped one of these keys fails here instead of leaving an entry that covers nothing.Why exemptions and not declarations: the pinned
17.0.0-rc.6does not declare any of the five, so publishing the inputs today fails this same file's FORWARD direction, and a forward exemption to cover that would itself go stale — and red — the moment the pin moves. "Must stay green on current main" wins, the same collision PR #4660 recorded forSECRET_MASK. #4668 declares them once the pin lands, and each entry then dies by the file's own stale check.packages/plugin-form/src/sectionFields.spec-parity.test.ts—publicPickerexemption with the same GA-pending mechanism, plus an "every exemption cites an issue" assertion the console gate already had and this file did not.packages/plugin-detail/src/__tests__/record{Details,Highlights}Inputs.spec-parity.test.ts— GA closedRecordHighlightsPropsand therecord:detailssection object under objectstack#4001 batch A, so an undeclared key raisesunrecognized_keyswith a named message where rc.6 dropped it silently. Both gates now assert the same verdict through whichever evidence the installed contract gives, with the mode probed behaviourally (a probe key no spec would declare) rather than read off a version string. The rejection branches assert the envelope — code plus the keys named — not a bare failure, so a refusal offieldsorlabelcannot satisfy them.A wrong probe cannot go vacuously green: taking the wrong branch fails on either pin, which is how these two were caught in the first place.
Evidence — both pins, at
6f96911dc(working tree clean)GA readings use the npm-pack overlay from PR #4660's body:
npm pack @objectstack/spec@17.0.0unpacked inside this worktree's virtual store with the peer links mirrored, every@objectstack/specsymlink repointed at it, the real suites run unmodified, rc.6 restored afterwards. Version confirmed per package before each run (17.0.0-rc.6vs17.0.0).Current pin,
17.0.0-rc.6— green:GA tree,
@objectstack/spec@17.0.0— 4 failures, all of them the forked item:Before this branch the same GA tree failed 8: the four above plus the coverage-set pin,
page:header,page:tabsandrecord:details— and, on the per-block gates next door, the two strictness assertions.plugin-detailpackage suiteapps/console+plugin-form+plugin-detailscope sweeptypes,components,plugin-list,plugin-grid,plugin-dashboard)check-changeset-presencecheck-changeset-fixed/check-changeset-no-majortype-check(plugin-detail,plugin-form,console)eslint --quietover the 4 changed filescheck:control-bytesThe suite, type-check, lint and control-byte readings above were first taken at
7eeda7e93and re-confirmed on both pins at6f96911dc; the only delta between those two commits is the changeset file (git diff --stat 7eeda7e93 6f96911dc= 1 file, 4 insertions).Reverse verification, direction predicted before running, both pins:
publicPickerentry ⇒ GA: 9 failures (the block assertion, both pending-set guards,has a behavioral row,has no stale rows); rc.6: 4 failures (the pending-set guards and the two plugin-form checks only — the block assertion stays green because rc.6 has no such key, which is the dormancy the mechanism claims). Restored from the commit,git statusclean, re-run green.Scope and the changeset
Four test files, no source, no registry
inputsand no published authoring surface change. Two of the four nevertheless live under released packages'src/(plugin-form,plugin-detail), which the Changeset Declaration gate counts, so this carries the repo's explicit exemption for that case: an empty-frontmatter.changeset/brave-pandas-invent.md, the same answer PRs #4641 and #4666 gave the same gate. Its output at this head:(An earlier revision of this body claimed a
skip-changesetlabel instead. That label does not exist in this repo — the empty-frontmatter changeset is the mechanism, and the label has been removed from this PR.)No file outside this card's declared surface; the in-flight siblings' files (
app-shell/views/RecordDetailView.tsx, plugin-grid +useConsoleActionRuntime, the catalog calendar entries) are untouched.Generated by Claude Code