1.0.3

Fixed

• userEvent.type, clear, keyboard and selectOptions do nothing on disabled controls, read-only fields and disabled options (no value change, no input events).
• userEvent.click, pointer, dblClick and tab use the matchers' disabled check (disabled fieldset ancestry); a stray disabled attribute on a div no longer blocks a click.
• toHaveStyle("...") splits declarations outside quotes and parentheses and at the first :, and throws a TypeError for a declaration or text that does not parse.
• toBeVisible treats opacity: 0 and the content of a closed <details> (other than its <summary>) as hidden.
• toHaveTextContent("") passes only for an element with no text.
• toHaveTextContent, toHaveAccessibleName and toHaveAccessibleDescription give the same answer on every call for a g or y regular expression.

1.0.2

Fixed

• render() and renderDirective() remove the container or element they appended when the UI, the directive or its first effect throws (they leaked into the document), and render({ hydrate: true }) without a container throws before it appends anything.
• userEvent.keyboard("{Tab}") dispatches one keydown instead of two and sends the keyup to the element that received focus.
• toBeDisabled() and toBeEnabled() apply to form controls and custom elements only (a div inside a disabled fieldset is not disabled) and spare the controls in the first legend of a disabled fieldset.
• waitFor works with Vitest's fake timers (vi.useFakeTimers()): each poll advances the fake clock by one interval instead of waiting on a clock that never moves, and timeout counts fake milliseconds. It finds the fake clock on setTimeout itself, so nothing is imported from vitest.
• findBy* and findAllBy* (on screen, within and a render result) and waitForElementToBeRemoved work with Vitest's fake timers like waitFor, and a container passed in their wait options now picks the DOM snapshot in the timeout error, as in Testing Library.
• startApp takes readyMatch (text or pattern): the body of the readyPath answer must match it, so another process that took the port is not mistaken for the app. An app that ended while the probe was in flight is reported as ended, not ready.
• freePort (and so startApp) picks a random port from 20000 up to 45000 that it can listen on on the given host, and never returns the same port twice in a process, instead of asking the system for port 0: the next ephemeral port is the pool every outbound socket draws from, so another connection could take it before the app started. It rejects at once with the listen error when the host cannot be bound, and after 64 busy ports with an error that names the range.

1.0.1

Changed

• Released at 1.0.1 with the other packages, which share one version. No change to this package.

1.0.0

Added

• Render, hydrate, DOM queries and automatic root cleanup for component tests.
• User-event helpers and utilities for testing effects, hooks and directives.
• @shadow-js/testing/playwright: waitForHydration (the dusk:hydrated mark), collectProblems
  (console errors and warnings, page errors, failed requests, 4xx/5xx answers), clientNavigate
  (follows a link and fails on a full document load), startApp and freePort (the app under test on a
  free port, stopped with SIGINT then SIGKILL), and createTest, which adds the app, problems
  and hydrated fixtures to a @playwright/test test. playwright-core is an optional peer
  dependency; the entry imports no Playwright package.