Changelog

Copy Markdown View Source

Unreleased

0.4.0 — 2026-09-15

Added

  • Filter download capture by filename: (string or regex) and url: (absolute string, regex, or URI predicate) on both backends.
  • Accept fn %URI{} -> boolean end in page and captured-result URL assertions, including negation.

Changed

  • Centralize deadline calculations across event capture, browser navigation, and LiveView retries while preserving their timeout behavior.
  • Retry failed CI tests once with the same seed, reporting the retry and retaining browser failure artifacts from both attempts. Formatting, compilation, and lint checks must pass before tests run.
  • Require playwright_ex 0.11 or newer for Frame.snapshot/2, Request.response/2, and managed event waiters.
  • Delegate browser URL waiting to Frame.wait_for_url/2 and event capture to EventWaiter. Captures retain the first matching event, even when several arrive before the action returns; the capture timeout starts when arming.
  • Save browser downloads through PlaywrightEx.Download. Temporary copies are removed after reading; source artifacts remain available until the browser context closes.
  • Use waiter transforms for captured metadata and dialog decisions, removing Fluffy’s event listener, dialog handler process, subscription registry, and snapshot messages. The connection manages event subscriptions.
  • Unexpected predicate and handler failures propagate instead of being converted into event errors.

Fixed

  • Read browser URLs and document identity from atomic frame snapshots. Same-URL reloads adopt their new response; requestless documents clear the previous HTTP status.

  • Preserve zero-timeout URL assertions as a single current-state check instead of a one-millisecond wait.

  • Keep captured navigation and HTTP metadata readable after the action closes its page.

  • Retrieve navigation responses directly from their requests, avoiding the response-cache handoff race. Read committed document requests from playwright_ex for automatic navigation and popup initialization, removing Fluffy’s navigation recorder and its subscriptions.

  • Give captured popups a separate session timeout for initialization, even when the action outlives the capture deadline.

  • Use the session timeout to save a captured browser download, including when the callback returns after the event deadline.

  • Ignore Phoenix navigation and download events arriving after the armed capture deadline, before applying download filters or byte limits.

  • Load browser download metadata before decoding the first download event.

  • Retain the first navigation on Phoenix's Static and LiveView drivers, including patches and fragment changes, when the callback continues navigating.

  • Ignore history metadata updates that leave the URL unchanged when capturing navigation; same-URL document reloads still count.

  • Associate navigation results with the committed request and popup responses with the captured page, preserving final redirect status. Capturing an earlier navigation no longer overwrites a later page's status.

0.3.1 — 2026-09-14

Fixed

  • Correct the session type declaration to reflect its shared implementation across backend modules. Consumer tracing, screenshot, and evaluation pipelines now pass Dialyzer without suppressions; session fields remain internal.
  • Improve Playwright assertion reliability with short timeouts.

0.3.0 — 2026-09-09

Changed

  • Breaking: consolidate assertion execution, negation, and constructors in Fluffy.Expect. Import it for expect and not_. Page and captured-result constructors now use target prefixes, such as page_to_have_url and response_to_have_status; the old constructors on subject modules are removed.
  • Add use Fluffy.Assert for pipeable assert and refute, with short constructors such as visible, page_url, and response_status. Ordinary ExUnit assertions remain available.

Fixed

  • Correct negated count expectations in the Static driver.

0.2.0 — 2026-09-08

Changed

  • Breaking: rename all Fluffy.Expect assertion constructors to fluent names. State assertions now use to_be_* (visible → to_be_visible, checked → to_be_checked, and likewise for disabled, editable, enabled, and focused). Property assertions use to_have_* (count → to_have_count, value → to_have_value, values → to_have_values). Captured-result assertions move to subject modules: Download.to_have_url, Dialog.to_have_message, Navigation.to_have_url, Request.to_have_method, and Response.to_have_status. Response request metadata is explicit in Response.to_have_request_method and Response.to_have_request_post_data.
  • Separate internal navigation transitions from public captured-navigation assertions. Remove duplicated subjects from captured-result assertion failures.
  • Captured URL assertions accept exact strings, regexes, and structured path, query, and fragment matchers. Structured matching shares page URL semantics, including exact and subset query matching; relative strings are not resolved.
  • Keep Page.to_have_*, assertion options, negation, and retry behavior unchanged. Update existing calls to the new names; the old names are no longer exported.

0.1.1 — 2026-09-08

Fixed

  • Preserve entered Phoenix form values when nested LiveView session tokens change and when phx-trigger-action submits the form.

0.1.0 — 2026-09-08

Added

  • Initial release