Test debt: add missing a11y tests, prop-coverage utility, and behavioral tests

[IgorShevchik]

Background

Surfaced during the multi-angle review of the nuxt/ui → b24ui sync plan (.sync/PLAN.md). These are pre-existing test-suite gaps, independent of the sync pipeline, so they are split out here. They also matter for the pipeline: the planned sync relies on the test suite to catch incomplete ports, so closing these strengthens that safety net.

Scope

1. Missing accessibility (axe) tests

The component test suite uses vitest-axe (it('passes accessibility tests')) for ~92 of ~98 component specs. The following components have no axe case:

• Advice
• Countdown
• ProsePrompt
• ScrollArea
• TableWrapper
• Theme

(Advice/Countdown are b24ui-specific and won't be synced from upstream, but the a11y gap is still worth closing.)

Action: add an axe case to each of the six specs.

2. Prop-coverage utility

There is currently no check that every prop of a component is exercised by a renderEach snapshot case. This is the main blind spot the sync pipeline is exposed to: a prop can exist in the types (so typecheck passes) without being rendered/implemented, and no snapshot changes.

Action: add a small utility/test that, for a given component, compares the declared props against the keys exercised in its spec's renderEach cases and reports uncovered props. Wire it into CI for changed components.

3. Behavioral tests for interaction-heavy components

~25% of components have behavioral tests (events/keyboard/focus); the rest are snapshot-only. Snapshots don't catch behavior regressions (e.g. keyboard navigation, focus trap).

Action (initial set):

• Accordion — keyboard navigation
• Select / SelectMenu — selection events, filtering
• Modal — focus trap / dismiss behavior

Priority

• §1 a11y — P2 (small, mechanical)
• §2 prop-coverage — P2 (high leverage for the sync pipeline)
• §3 behavioral — P3 (incremental)

Reference

• Review context: .sync/PLAN.md "Quality requirements for every ported PR"

[IgorShevchik]

Done and merged in #529 — 1da904d6. All three sections are settled, though only two of them the way this issue expected.

§1 — axe cases ✅

Advice, Countdown, ProsePrompt and TableWrapper have one now.

ScrollArea and Theme deliberately do not, and this issue's list was two entries too long. nuxt/ui@v4 skips exactly those two as well, and the reason is right there in the source — Theme.vue is <template><slot /></template>. Running axe on a renderless provider measures whatever the test put in the slot, not the component.

The more useful finding: an axe test that runs no rules passes for free. Two of the four cases in the first draft ran zero rules and were perfectly green. TableWrapper was handing its table to the slot as a string, which Vue escapes to text, so no table ever reached the DOM. Countdown's ring and digits match no axe rule at all. Both now mount something axe can actually look at, and each of the four was checked against a real regression:

Mutation	What fails
:alt stripped from the avatar <img> in Avatar.vue	Advice, Countdown → image-alt
:label stripped from the prompt's action buttons	ProsePrompt → button-name
role="row" on the TableWrapper root	aria-required-children, aria-required-parent

And a footnote worth keeping: the first TableWrapper mutation did go red — on aria-required-children, because there were no rows to find, because there was no table. A mutation going red is not by itself proof the test works.

§2 — prop coverage ✅, but not as a utility

The scan happened; the standing checker did not. 1140 declared props, 242 never mentioned in their own spec — 21 %, and that is a floor.

The concentration is what made it worth doing. The densest offenders are props nuxt/ui does not have at all:

                Input.vue  Select.vue   (upstream)
noPadding           0          0
noBorder            0          0
underline           0          0
tag                 0          0
tagColor            0          0

An upstream prop arrives with an upstream test attached. A Bitrix24-only prop arrives with nothing, and no future sync will bring one — so the safety net is thinnest exactly where the code is most ours. Cases added for the Bitrix24-only props of Input, Textarea, Button, Select and InputNumber.

No utility was added and no CI job was wired up, deliberately: a permanent checker is machinery to maintain forever and one more divergence from upstream. The script, the numbers and the caveats are recorded as item 6 on #75 instead, next to the other things that gate trusting the automated sync — because "a prop exists in the types with no behaviour behind it" is precisely what an incomplete port looks like.

It caught three dead props on the first run

Button.depth, Button.activeDepth and Textarea.fixed are typed, documented, forwarded to the theme, and render nothing. All three are now @deprecated with @removed 3.0.0.

Button.depth turned out to have been noticed before and tagged @depricate — misspelled, so no IDE, no vue-tsc and no docs generator ever saw it. It had been quietly hidden from the docs by a hardcoded name list instead.

Which surfaced the follow-on: nothing in the docs read @deprecated at all. Both the props table and the /raw/ markdown that llms.txt serves now do, and both hardcoded lists are gone. A deprecated prop reads like this in the raw page:

  /**
   * Keep the mobile text size on all breakpoints. (Left for backward compatibility.)
   * @deprecated Does nothing. It works through `fixed` x `size` compound
   *   variants, and `Textarea` deliberately has no `size`, so no variant ever
   *   matches.
   * @removed 3.0.0
   */
  fixed?: boolean | undefined;

So if an assistant tells you to reach for <B24Textarea fixed /> after this ships, it now has no excuse. Try it:

Using @bitrix24/b24ui, build a comment box with a B24Textarea that grows as you type, capped at six rows, with a character counter underneath. Check https://bitrix24.github.io/b24ui/raw/docs/components/textarea.md first and tell me if any prop you were about to use is deprecated.

§3 — behavioural tests ✅ already done

Nothing to do here. 36 of 110 of our component specs carry behavioural assertions against 28 of 116 in nuxt/ui@v4 — we were already ahead when this was written. Of the initial set, Select, SelectMenu and Modal were covered; Accordion is snapshot-only upstream too, so there is nothing to catch up to.

---

Thanks from the maintainer for filing this one — it was written as three chores and turned into three dead public props and a docs pipeline that had never read its own deprecation tags. Closing.

Docs: Button · Textarea

---

Generated by Claude Code