fix(indicator): a falsy child no longer deletes the state mark (#4893)

[cixzhang]

Fixes #4893 — a regression I introduced in #4890.

The bug

The busy idiom a host actually writes is:

<CheckIndicator state="checked">{isBusy && <Spinner/>}</CheckIndicator>

When isBusy is false that passes false. false is not null, and it is not caught by ?? — so every indicator took its children path, rendered nothing there, and deleted the state mark on every selected row.

It was wider than the issue says

#4893 was filed against CheckIndicator, because that is where #4890 moved the branch. I probed all three before writing anything, and all three were affected — the siblings' {children ?? mark} has the same hole, and it predates #4890:

children	CheckIndicator	CheckboxIndicator	RadioIndicator
false	mark lost, empty span	mark lost	mark lost
''	mark lost, empty span	mark lost	mark lost
null / undefined	ok	ok	ok
<b/>	replaces mark ✓	replaces mark ✓	replaces mark ✓

So CheckIndicator additionally leaked an empty <span>; the mark loss is the whole family. #4890 widened the CheckIndicator case from one state to every state — before it, the unchecked early return happened to mask it.

The fix

isRenderable(children) in all three — the util the repo already ships for exactly this (utils/isRenderable.ts), and what @astryx/no-nullish-jsx-guard steers to. That rule could not catch this: it sees JSX guards, not an early-return if or a ?? fallback.

0 deliberately still counts as content — it renders the character 0, so it is a real child. There is a test pinning that.

Two contained follow-ons in the same branch, both from the #4890 audit:

• CheckIndicator's children slot now reserves the glyph's box (1rem, Icon's sm, which is what both indicator sizes resolve to). The doc says children "keeps the indicator's place"; it did not — the row shifted when a Spinner swapped in.
• and carries the glyph's color (--color-icon-disabled when disabled, else accent), so an inherit-shade Spinner reads correctly. The sibling indicators already set color on their chrome for this reason. That closes the "size/isDisabled dropped on the children path" FIX.

Deliberately not fixed here: ref is dropped on CheckIndicator's glyph path. It looked like a one-liner and is not — Icon in string mode routes through IconFromRegistry, which accepts no ref, so fixing it means changing Icon's ref behavior for every caller. That is its own PR, not a rider on a regression fix. It stays open in the ledger.

New: apps/storybook/stories/Indicator.stories.tsx

This is scope I added on purpose. The indicators had no story file — rubric X9, open as #4894 — and the audit found that is why this bug shipped unseen: the children path had no rendered coverage anywhere, so nothing could have shown it. A regression fix with no way to see the regression would repeat that.

Five stories: AllStates, Sizes, Disabled, BusyChildren (the case above, both columns), BusyToggle (flip isBusy live). This also un-blocks rtl:audit, which currently reports 0 stories / 0 checks for Indicator.

Test plan

The new tests fail against main. I reverted the three source files in this same tree and re-ran:

× CheckIndicator keeps its mark when children is false — the `isBusy && …` idiom
× CheckIndicator keeps its mark when children is empty string
× CheckboxIndicator keeps its mark when children is false — the `isBusy && …` idiom
× CheckboxIndicator keeps its mark when children is empty string
× RadioIndicator keeps its mark when children is false — the `isBusy && …` idiom
× RadioIndicator keeps its mark when children is empty string
   Tests  6 failed | 23 passed (29)

All 29 pass with the fix. Each indicator also has a negative control — a real child still replaces the mark — so "always draw the mark" cannot pass this suite.

Visual, measured rather than asserted. Two Storybooks: main at 47f8bb4560 with this same story file copied in, and this branch. Real Chromium, deviceScaleFactor: 4, story core-indicator--busy-children, neutral light.

DOM count of visible marks in the not-busy column: before 0 of 4, after 4 of 4.

Pixel count of mark-ink inside each indicator's chrome:

	before	after
CheckIndicator	0 px	156 px
CheckboxIndicator (checked)	0 px	505 px
CheckboxIndicator (indeterminate)	0 px	372 px
RadioIndicator	0 px	1231 px

(For the filled chromes the mark is light-on-accent, so total non-white ink goes down slightly while the mark appears — counting light pixels enclosed by painted ones is the honest measure. Crops in /tmp/shots-4893/crops/.)

Suites: vitest 374 passed across Indicator + CheckboxInput + RadioList + Selector + DropdownMenu + CheckboxList; tsc core clean; CI=true eslint --no-cache 0 errors 0 warnings; check:changesets 55 valid.

Note on the audit trail

The ledger row for Indicator is 79.1 / C with three open BLOCKs (#4893, #4894, #4895). This PR closes #4893 and, via the story file, #4894. Per the rubric's closing protocol I will re-audit and update the ledger once this merges — the score has to move.

---

Two CI fixes on top, both pre-existing rather than caused here

build-storybook — storybook typecheck. My first push used satisfies Meta + StoryObj<typeof meta>, which makes args required on every story; state is required on an indicator, so the five render-only stories failed. The other 122 story files use const meta: Meta<typeof X> with StoryObj<typeof X>; mine now matches. My fault, caught by CI, fixed in a913c40406.

Vercel — the docsite build. This one is not from this PR. next build type-checks the emitted componentRegistry.ts against the docsite's own copy of the authoring doc types, and two fields had drifted out of that copy: deprecatedFor (added to ComponentThemingTarget by #4712) and replaces (on ComponentThemingDerivedVar, used by TextArea).

Verified with a control rather than asserted: a clean origin/main worktree fails the docsite build with the same 'deprecatedFor' does not exist in type 'ThemingTarget', and passes 315/315 static pages with only the one-file type fix applied. It surfaced here because this PR touches core .doc.mjs inputs, so Vercel's ignored-build-step ran a docsite build for the first time since #4712.

I diffed all six shapes the docsite duplicates; those two were the only gaps. Both authoring interfaces now carry a SYNC: breadcrumb, which is the mechanism the repo already uses to catch this.

One more thing seen while chasing it: /community prerenders a GitHub API call, and a 403 rate-limit from a build machine takes the whole build down (Error occurred prerendering page "/community"). I hit it locally, and the same build passed once the limit reset. The fetchContributors catch returns [], so something downstream of an empty list is the actual crash. Not this PR's to fix, but it is a real build-flake and worth its own issue.

---

Rebased after #4911 and #4915 landed. Both overlapped this PR while it sat in CI:

• docs(indicator): untag @example JSDoc fences so Storybook autodocs renders them #4911 untagged the @example fences — the same X14 fix ([audit] Indicator: every @example JSDoc fence in the Indicator sources is language-tagged ```tsx, which the repo documents as silently breaking Storybook autodocs rendering — 7 occurrences across 5 files, introduced by #4712 and missed by both prior audits #4895) I was holding back so it would not race this PR's armed merge. Dropped mine; its version is in the base now.
• fix(docsite): sync ThemingTarget/DerivedVar with the authoring types #4915 made the same generate-data.mjs type sync. Dropped my duplicate and took its file verbatim. What I kept is the pair of SYNC: breadcrumbs on the authoring interfaces — fix(docsite): sync ThemingTarget/DerivedVar with the authoring types #4915 does not have those, and they are what stops the drift coming back a third time.

What remains here is only the falsy-children fix, its tests, and the story file. Re-verified post-rebase: 280 tests green, core tsc clean, storybook tsc clean.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
astryx	Ready	Preview	Aug 11, 2026 11:10pm

[github-actions]

PR Analysis Report

📚 Storybook Preview

View Storybook for this PR
GitHub Pages may take up to a minute to hydrate after deploy.

🧪 Sandbox Preview

View Sandbox for this PR
GitHub Pages may take up to a minute to hydrate after deploy.

Modified Components

Indicator _{(@astryxdesign/core)} · View in Storybook

Metric	Before	After	Delta
Bundle Size (ESM)	N/A	N/A	N/A
Lines of Code	N/A	551	-
Complexity	N/A	High (26)	-

Bundle Size Summary

Package	Size (ESM)	Size (CJS)	Gzipped
@astryxdesign/core	N/A	4.7KB	1.2KB

Accessibility Audit

Status: No accessibility violations detected.

---

_{Generated by PR Enrichment workflow | Storybook | Sandbox | View full report}