Dmitry Gozman

This commit introduces several **Edge-specific command-line flags** to the **Playwright core server's Chromium configuration**. This **maintenance chore** enhances the browser launch process by disabling unwanted behaviors such as forced browser sign-in, update services, and the Edge updater. By modifying `chromiumSwitches.ts`, the change ensures a more stable and predictable environment for **automated testing with Microsoft Edge**, preventing interruptions and improving test consistency.

This commit provides significant **documentation updates** in preparation for the **version 1.59 release**. It **adds comprehensive release notes** for the **C#, Java, and Python language bindings**, detailing new features such as the Screencast API, Snapshots, and Locators, alongside breaking changes and updated browser versions. The **JavaScript release notes** are also updated with the latest browser versions, and the `Page.screencast` property documentation is clarified regarding its `onFrame` callback type. Finally, a minor **tooling update** ensures 'Promise' is correctly mapped to 'Task' during .NET API documentation generation.

This commit primarily focuses on **fixing and unflaking various tests** to improve the overall stability of the test suite. It addresses test flakiness by marking `screencast.spec.ts` as slow, adding a Firefox-specific `fixme` for worker behavior in `workers.spec.ts`, and enhancing error messaging in the `TestServer`. Additionally, it includes **bug fixes** within **`playwright-core`** to correctly format `chrome:` and `edge:` protocols in `protocolFormatter.ts` and to use `startedDateTime` for unique resource IDs in `traceModel.ts`. These changes collectively enhance the **reliability of the test suite** and the **accuracy of core library functionalities**. The CI workflow for secondary tests was also adjusted to use more workers on macOS.

This commit **reverts** a previous change, making the **Playwright API's** `Page.video()` method **nullable** once more. It **fixes** an issue where `page.video()` would not correctly return `null` when video recording was not configured, preventing potential errors with invalid video objects. The `playwright-core` module's `video()` method is updated to return `null` in this scenario, and `playwright` now filters to ensure only valid video objects are processed. This **API correction** is accompanied by **documentation updates** for `Page.video` and a new test case to verify the expected `null` return behavior.

This commit **enhances the clarity and type precision of the API documentation** by introducing explicit struct names for various return types and parameters. It **refactors the documentation generation tooling** (`doclint`) to parse and utilize these new explicit struct definitions, replacing generic `Object` types with specific, named types like `Header` or `FilePayload`. This **documentation improvement** spans across numerous API classes, including `APIResponse`, `BrowserContext`, `Request`, and `Response`, making the API signatures more explicit. Furthermore, it **streamlines the .NET API generation** process by leveraging these explicit names, thereby eliminating a significant amount of hardcoded type name overrides and improving consistency.

This commit primarily focuses on **test maintenance and fixes** across several Playwright test suites. It **updates and skips various tests** within the `trace-viewer`, `page-drag`, and `playwright-test` modules to improve their reliability and align with current expected behaviors, including a new `fixme` condition for drag-and-drop in headless environments. Concurrently, a **minor feature enhancement** is introduced in `playwright-core`'s **trace snapshot rendering** utility, allowing the `applyPlaywrightAttributes` function to accept a `blankSnapshotUrl` parameter. This work collectively ensures test accuracy and provides more granular control over trace rendering.

This commit performs a **maintenance chore** by updating the **`ThirdPartyNotices.txt`** file within the **`playwright-core`** package. Specifically, it bumps the listed versions for the `diff` and `yauzl` third-party dependencies to reflect recent dependency updates. This ensures that the project's legal and compliance documentation accurately reflects the versions of external libraries being used. This is a **documentation update** with no functional impact on Playwright's core capabilities, primarily affecting the accuracy of **third-party notices**.

This commit provides a **bug fix** for the Playwright **test runner**, ensuring that only test files explicitly matching the `--test-list` argument are loaded, rather than all available files. It introduces a `loadFileFilters` property to the internal configuration in `config.ts` and **refactors the test file collection logic** within `loadUtils.ts` and `tasks.ts` to apply these filters, which are derived from CLI arguments. This change significantly **improves performance** by reducing memory consumption and startup times when running a targeted subset of tests. A new test case in `test-list.spec.ts` has been added to verify the correct implementation of this targeted file loading behavior.

This commit performs a **maintenance chore** by **bumping dependencies** listed in the **third-party notices documentation**. It updates the version numbers for `express-rate-limit`, `ip-address`, `qs`, `yaml`, and `picomatch` within the `ThirdPartyNotices.txt` files for both the **`playwright-core`** and **`playwright`** packages. This ensures the project's legal and compliance documentation accurately reflects the current versions of external libraries in use, with no functional changes to the Playwright product.

This commit primarily performs a **Chromium browser update** to revision `r1217`, which involves updating the **Playwright core browser configuration** and the Chrome version within **device emulation user agent strings**. Concurrently, the internal `utils/roll_browser.js` utility is **enhanced** to enforce version requirements and improve validation for Chromium-based browsers. Separately, the **API documentation generation script** (`utils/doclint/generateApiJson.js`) receives a **refactoring** to extract and integrate language-specific serialization logic. This ensures Playwright's compatibility with the latest Chromium features while improving internal tooling and documentation processes.

This commit performs a significant **refactoring** and **API update** to the **Debugger API** following a review. The `Debugger.pause` method has been **renamed to `Debugger.requestPause`** to clarify its asynchronous behavior, and the `Debugger.pausedDetails` property now consistently returns a single object or `null` instead of an array, simplifying paused state management. These changes are reflected across the **Playwright core debugger implementation**, its **protocol definition**, and **documentation**. Consequently, various internal components like the **recorder** and **DevTools backend**, along with Playwright fixtures and tests, have been updated to align with the new API. This ensures a more consistent and intuitive debugging experience.

This commit provides a **follow-up enhancement to the HTML reporter's step filtering capabilities**. It **refactors the core filtering logic** within the `StepTreeItem` component, introducing a new `stepChildrenMatchFilter` function to accurately determine child step matches. This improvement allows the reporter to **highlight filtered text** in step titles using the new `step-title-highlight` CSS class and intelligently control step expansion. The changes result in a more intuitive user experience when navigating filtered test results, with new tests added to verify the improved filtering and expansion behavior.

This commit **fixes a regression** in the `ariaSnapshot` API, ensuring that `locator.ariaSnapshot()` and `page.ariaSnapshot()` correctly apply auto-waiting when a selector is provided but the `mode` is not explicitly set to 'ai'. The **Playwright core server logic** within `frames.ts` is updated to restore this expected default behavior for `ariaSnapshot` calls. Additionally, **documentation** for `Locator.ariaSnapshot` and `Page.ariaSnapshot.mode` is clarified to distinguish between 'ai' and default behaviors. **New test cases** are introduced to validate auto-waiting, depth limiting, and iframe snapshotting for `locator.ariaSnapshot()`, preventing future regressions.

This commit performs **test maintenance** to **fix and unflake several tests** across the Playwright suite, enhancing overall test reliability. It addresses a brittle assertion in `tests/library/browser-server.spec.ts` by making the `pipeName` regex less specific and resolves a timing issue in `tests/library/browsercontext-reuse.spec.ts` by awaiting service worker activation. Furthermore, the commit updates expected hashes for **HTML reporter screenshot attachments** in `tests/playwright-test/reporter-html.spec.ts` and refines a CSS assertion for **UI mode test output** in `tests/playwright-test/ui-mode-test-output.spec.ts` to specifically check `outline-color`. This targeted **test stabilization** reduces flakiness and ensures more consistent results in the **Playwright test infrastructure**.

This commit introduces a **new capability** to the **Playwright CLI `install` command**, enabling users to specify a custom target directory for skill installation. It specifically adds support for the `--skills=agents` option, which directs the installation of skills into a dedicated `.agents` directory. This **enhancement** modifies the `install` command within both the `cli-client` (`program.ts`) and `cli-daemon` (`commands.ts`) components to accept a string value for the `skills` option. This allows for more organized skill management within projects and is validated by a new test case in `cli-misc.spec.ts` ensuring correct behavior.

This commit **fixes a bug** in the **Chromium network management** within `playwright-core`, ensuring that **internal redirects** are correctly processed even after a regular HTTP redirect has occurred. It resolves an issue where the `crNetworkManager` incorrectly used `_route` instead of `_originalRequestRoute` for handling internal redirects and fulfilled requests, leading to improper network request lifecycle management. This **enhances the reliability of network interception and mocking** for Chromium, particularly in scenarios involving complex redirect chains, by correctly managing `InterceptableRequest` objects.

This commit primarily **refactors the Playwright test suite** by introducing and widely adopting the `inheritAndCleanEnv` utility, which standardizes environment variable management across numerous test fixtures and browser launchers, including those for Electron, Firefox, and Playwright Test. It also includes a **bug fix** for **Firefox screencasting**, ensuring `width`, `height`, and `quality` options are explicitly passed to `Page.startScreencast`. Additionally, several **test reliability improvements** are made, such as enhancing locale propagation assertions, replacing `about:blank` with `server.EMPTY_PAGE` for consistency in navigation tests, and improving form navigation test stability. This work significantly enhances the robustness and maintainability of the **Playwright test infrastructure** and resolves a specific Firefox-related issue.

This commit implements a **logging improvement** to **suppress `pw:api` logs** when they originate from internal Playwright instances. It introduces a new `apiLog` method on `SdkObject` in `packages/playwright-core/src/server/instrumentation.ts` and modifies `ProgressController` in `packages/playwright-core/src/server/progress.ts` to conditionally prevent these logs. Existing direct `debugLogger.log` calls within `packages/playwright-core/src/server/frames.ts` are **refactored** to utilize this new `apiLog` mechanism. This change significantly **reduces logging noise** for users, providing clearer debug output by filtering out internal API calls.

This commit **fixes** a bug in the **`clock` module** by ensuring that `performance.mark` and `performance.measure` return proper `PerformanceEntry`-like objects when the clock is faked. It introduces a `FakePerformanceEntry` class within `packages/injected/src/clock.ts` to provide realistic fake entries for these performance API calls. This **bug fix** enhances the fidelity of the faked performance APIs, making performance-related tests more reliable and accurate. A new unit test in `tests/library/unit/clock.spec.ts` has been added to validate this corrected behavior.

This commit **updates the documentation** for **emulation** by adding a crucial recommendation regarding user agent handling. It advises users to **unset the `userAgent`** when aiming for platform-specific user agents to ensure correct behavior. The update, located in `docs/src/emulation.md`, includes practical **code examples** in JavaScript, Python, and C# to guide developers in implementing this best practice. This **documentation enhancement** improves user understanding and helps prevent common configuration issues related to browser emulation.