test(vi-mock): inherit the real @object-ui/plugin-detail surface in 13 vi.mock factories - #8184
Merged
baozhoutao merged 2 commits intoSep 6, 2026
Conversation
…3 vi.mock factories
objectui#6892 slice 9. Every frozen `vi.mock('@object-ui/plugin-detail', ...)`
factory now obtains the real module through `importOriginal` and spreads it
FIRST, with the file's existing overrides after it, so a new export on the
barrel can no longer resolve to `undefined` and kill the file during
collection. `@object-ui/plugin-detail` joins the gate's `COVERED_SPECIFIERS`
in the same PR (13 -> 14), which is the documented widening precondition.
Census re-derived through the gate's own `scan()` on 310c0ab: 17 judged,
4 already inheriting, 13 frozen -> 0. All 13 sit in one syntactic shape
(zero-parameter object-literal arrows) across three packages: plugin-gantt 8,
plugin-calendar 4, plugin-tree 1. No test assertion changed.
STEP 0 measured rather than assumed, and the two measurements disagree.
The barrel's graph is the largest this worklist has walked -- 564 modules,
6181 module-scope statements, 127 ComponentRegistry registrations, 99 bare
side-effect imports -- and timed cold in an EMPTY test file `importOriginal`
of it costs a median 8.8s, which is objectui#6580's app-shell STOP range.
Measured in the real files it is roughly free: -0.1s to +1.0s per file,
because those files already import `@object-ui/react` plus their own view
component, whose combined 572-module graph is a strict SUPERSET of the
barrel's 564 (zero marginal modules). The cost of inheriting a barrel is a
property of the consuming file, not of the barrel.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FhBNJcLRZLe8M87VcUgpKr
… no plugin-detail edge
Fix-up on the slice-9 sweep. `check-phantom-dependencies` reads a
type-position `import('@object-ui/plugin-detail')` as a real specifier, and
`packages/plugin-tree` neither declares that package nor needs to: measured,
`ObjectTree`'s module graph reaches ZERO plugin-detail modules, against 50
each from `ObjectGantt` and `ObjectCalendar`. That factory therefore mocks a
module nothing under test ever loads, and the typed spelling would have made
the package declare a dependency its runtime does not have.
`<any>` inherits the same surface without asserting the edge, and it is the
spelling this tree's three already-inheriting plugin-detail factories use. The
twelve siblings keep the typed form, where the edge is real and declared. A
comment on the factory records the measurement so the difference is not read
as drift and "corrected" back.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FhBNJcLRZLe8M87VcUgpKr
Contributor
✅ Console Performance Budget
The eager closure is every chunk the entry reaches through static imports — what the browser fetches and parses before the app renders. The entry chunk on its own is a small fraction of it. 📦 Bundle Size Report
Size Limits
|
Contributor
|
Standing down on That check is informational (not in the Generated by Claude Code |
This was referenced Sep 6, 2026
baozhoutao
marked this pull request as ready for review
September 6, 2026 23:14
Contributor
|
Armed —
Landing stroke follows the merge: content probe on re-fetched Generated by Claude Code |
baozhoutao
deleted the
claude/issue-6892-vi-mock-inherit-slice9-plugin-detail
branch
September 6, 2026 23:30
Contributor
|
LANDED — Merged at 2026-09-06T23:30:16Z as Slice 10 ( Generated by Claude Code |
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 free
to 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 #6892 (slice 9:
@object-ui/plugin-detail)Base
origin/main310c0ab19(carries PR #8163, slice 8). Shipped head60a2c6a51. Every reading below is pinned to that head on a clean tree.The thirteen
vi.mock('@object-ui/plugin-detail', ...)factories that hand-listed the barrel's exports now obtain the real module and spread it FIRST, with each file's existing overrides after it, and@object-ui/plugin-detailjoins the gate'sCOVERED_SPECIFIERSin the same PR. No test assertion changed.STEP 0 — PROCEED, and the isolated probe would have answered it WRONG
Method. Static import graph walked from
packages/plugin-detail/src/index.tsxwith the TypeScript parser, every module-scope statement of every reached module classified (TS ESM'./x.js'resolved to./x.ts/.tsx,'@/x'as the per-package src alias). Corroborated empirically by importing the real barrel throughimportOriginalunder the lightdomproject — the single project all seventeen sites run in.The graph is the largest this worklist has walked: 564 modules, 6,181 module-scope statements, 2,610 import/export, 127
ComponentRegistry.registercalls, 99 bare side-effect imports and one CSS import. The barrel's OWN sixteen registrations are all namespaced (A2 as stated:plugin-detail3,record11,view1,field1) — but A2 is falsified for the transitive graph: the barrel reaches@object-ui/components, whoseimport './renderers'cascade contributes 111 further registrations, five of them BARE (no namespace) atpackages/components/src/renderers/layout/page.tsx:690-695, registeringpage,app,utility,home,record.Import cost, three runs each, median reported.
importOriginalof the barrel, isolated probe fileplugin-ganttrowCeiling, FROZENplugin-calendarrowCeiling, frozen then inheritingplugin-treerowCeiling, frozen then inheritingComparator: objectui#6580 measured
importOriginal('@object-ui/app-shell')at ~10.0 s per console test file, which the dispatch defined as a STOP. The isolated probe reproduces that range, and it is an artefact. Acting on it would have stopped a free conversion.Mechanism, measured rather than argued. All thirteen files already import
@object-ui/reactand their own view component (./ObjectGanttand friends) at module scope. Those two together reach 572 modules — a strict SUPERSET of the barrel's 564, withcomm -23over the two sorted module lists reporting exactly 0 modules that inheriting adds.importOriginalhere re-resolves an already-resident graph; it loads nothing new. The contrast is visible per package:ObjectGantt's graph contains 50 plugin-detail modules andObjectCalendar's 50, so the mock is live in those files.⇒ Carry-forward for slice 10 and beyond: a barrel's import cost is not a property of the barrel. Measure it in the files that will pay it, against what they already load. A probe that imports the barrel alone answers "what does this graph cost from cold", which is the wrong question whenever the consuming file already holds it.
⛔ No change to
packages/plugin-detail/src/**; STEP 0 called for none.STEP 1 — census, through the gate's own exported
scan()Covered set widened in-memory to the 22 workspace specifiers any
vi.mockcall site in the tree names;COVERED_SPECIFIERSwas never widened-and-reverted on disk.packages/plugin-ganttpackages/plugin-calendarpackages/plugin-treepackages/plugin-kanbanpackages/app-shellA1 CONFIRMED exactly, in both halves: the slice-8 forecast of 17 judged / 13 frozen is what
scan()re-derives on310c0ab19, and the PM's per-package file split (gantt 9, calendar 5, one each in app-shell, kanban, tree) is exact. No drift to report.All thirteen sit in one syntactic shape — zero-parameter object-literal arrows — the most uniform surface this worklist has swept. Population moves 74 → 61 frozen over a constant 659 judged. Direction check: a diff of the two per-specifier tables with the plugin-detail row removed is EMPTY, all 21 other rows byte-identical; no site moved the other way.
Free confirmation PRESENT: 4 of 17 sites already inherited the real barrel on
mainand passed, across four packages, so the module was known to load in this environment before anything was converted.Neighbour reading — per file, judged against the graph STEP 0 walked
Every other
vi.mockfactory in all thirteen files was read and classified before editing: 7 frozensonner(third-party, out of scope by construction) and 12 local whole-module replacements (./GanttView8,./CalendarView4, out of scope by the ruling). Slice 6's collection-death class did not fire, and the reason was established in advance rather than by a red run: this barrel's graph reaches neithersonnernor those local modules. Zero neighbouring repairs.A5 CONFIRMED on the base: no file among the thirteen mocks
@object-ui/app-shellor@object-ui/plugin-list(grep exit 1 on both). No hot-file overlap with the parked specifier. ⛔ No file mocking@object-ui/app-shellwas touched.One file diverges, and it is a finding rather than a style choice
packages/plugin-tree/src/ObjectTree.rowCeiling-7210.test.tsxinherits throughimportOriginaltyped as ANY rather than thetypeof import(...)form its twelve siblings use.check-phantom-dependenciesreads a type-positionimport()as a real specifier and reddened on the typed form:packages/plugin-treedoes not declare@object-ui/plugin-detail, and has no reason to. Measured,ObjectTree's module graph reaches 0 plugin-detail modules, against 50 each fromObjectGanttandObjectCalendar— that factory mocks a module nothing under test ever loads.The typed spelling would have forced the package to declare a dependency its runtime does not have. The ANY spelling inherits the same surface without asserting an edge that is not there, and it is what this tree's three already-inheriting plugin-detail factories use. A comment on the factory records the measurement so the difference is not read as drift and "corrected" back. The phantom mock itself is left in place — removing it is a decision for triage, not a rider here.
Positive control — three legs, from the committed tree, restore proven by STATE
Each leg under a
trapwith absolute paths; each mutation proven on disk by an anchored count AND a changed blob before the gate result is read; each restore proven bygit hash-objectequalling the HEAD blob ANDgit diff HEADnaming zero paths — never by a checkout exit code.plugin-gantt/.../ObjectGantt.rowCeiling-7210.test.tsx:64080e6cf56✓plugin-calendar/.../ObjectCalendar.unscheduled-7071.test.tsx:56ac18c89d0✓plugin-tree/.../ObjectTree.rowCeiling-7210.test.tsx:579c768873b✓Gate back to exit 0 after each restore, and
git statusempty at the end.finding):holdsObtainedModule()tests whether the obtained token is MENTIONED inside a spread, not whether the module is the thing spread. Measured on this tree, 0 of the 576 covered sites judgedinheritsuse that shape, so nothing evades the gate today. ⛔ Not repaired here — this PR widens the covered set, it does not touch the recogniser.No dist preflight is owed: the gate is a source-text scanner and both configs alias the specifier to source.
Covered set: 13 → 14
One line added to
COVERED_SPECIFIERS('@object-ui/plugin-detail', exact string match), the header count word thirteen → fourteen, the slice-9 per-specifier record added in the shape slices 5-8 used, and the remaining-population paragraph refreshed 74 → 61 with@object-ui/app-shellre-annotated as PARKED under objectui#8173.The pin
scripts/__tests__/check-vi-mock-inherit.test.tsneeded no edit — confirmed by RUNNING it, not by reading it: it derives names fromCOVERED_SPECIFIERSrather than copying them.Gates — exit codes captured by redirect-then-capture, pinned to
60a2c6a51node scripts/check-vi-mock-inherit.mjs@object-ui/plugin-detail(17 judged, 17 inheriting)scripts/__tests__/check-vi-mock-inherit.test.tsvitest run scripts/__tests__/(whole directory)vitest runthe 13 converted files310c0ab19vitest runplugin-gantt + plugin-calendar + plugin-treetype-checkof the three packagestsconfig.test.json --listFiles: 8/8, 4/4, 1/1 edited files present in programs of 1481, 1402, 1188pnpm type-check:scriptspnpm lint:rooteslint --no-inline-config --format jsonover the 14 non-changeset pathsno-explicit-anyon the plugin-tree ANY spelling, the same rule firing 89 times on untouched lines in these files)pnpm check:phantom-depspnpm check:control-bytesgrep -naPself-scan of all 15 changed paths, grep exit 1, no matchnode scripts/check-changeset-presence.mjsnode scripts/check-changeset-no-major.mjsnode scripts/check-governed-queue-guard.mjs --testpnpm check:entry-guardpnpm check:vi-mock-specifiersDependency closure built first (
pnpm --filter '...^...' build, exit 0) — without it the three type-checks fail withTS2307 Cannot find moduleon every workspace import, which reads exactly like a break this diff caused.Live E2E (informational)is base-red on every branch today for an upstream reason (#7990 / objectstack#16186), not this PR.Declared narrowing.
turbo ls --affectednames nine packages; the three OWNING packages were run whole andapps/console,apps/siteand the three examples were not. Backed by evidence rather than judgement: every changed file is a test file, and test files are leaves, so no dependent package has a path to reach this diff. CI runs the full farm regardless.Two commits, no amend and no force-push at any point: the branch history is the empty routing push, the sweep, and the plugin-tree fix-up.
🤖 Generated with Claude Code
https://claude.ai/code/session_01FhBNJcLRZLe8M87VcUgpKrGenerated by Claude Code