[poc] feat(webkit): classic W3C WebDriver backend#41150
Draft
yury-s wants to merge 8 commits into
Draft
Conversation
Adds a wd* backend (next to webview/) that drives Safari over the classic W3C WebDriver protocol exposed by safaridriver — both by launching safaridriver locally (webdriver+launch://) and by connecting to an existing endpoint (webdriver://), reached via webkit.connectOverCDP(). WebDriver pushes no events, so the backend synthesizes the frame, execution context and lifecycle events the core Page/Frame machinery expects. Since WebDriver has no persistent object handles, live handles are emulated with a page-side registry (window.__pwHandles), which lets the InjectedScript hit-target interceptor, element handles and waitForFunction-style polling work across the handle-less execute boundary. Verified end-to-end against real Safari: goto, evaluate, screenshot, fill/type/click, and locator queries. Also adds tests/webdriver/ which reuses the existing tests/page/ suite via PWPAGE_IMPL=webkit-webdriver (mirroring tests/webview/). Run with `npm run wdtest`.
Makes tests/page/page-check.spec.ts pass (and unblocks setContent/console-based tests generally). - Capture console messages + document readyState as a side-channel of every execute, then synthesize console and lifecycle events. This drives Frame.setContent (which logs a console.debug tag to clear lifecycle) and resolves waitForLoadState over a protocol that pushes no events. - Dispatch input through the same injected page-side dispatcher the webview backend uses (window.__pwWebViewInput) instead of the W3C Actions API. The synthetic events are handled deterministically by Playwright's hit-target interceptor at the exact target, eliminating the intermittent click flake (only mousedown delivered) that trusted OS-level Actions events caused.
Switches the WebDriver backend back to the native W3C Actions API for input (instead of the injected synthetic dispatcher) and fixes the real root cause of the intermittent dropped clicks. - Serialize all WebDriver commands onto a single chain. safaridriver processes one command per session at a time; Playwright issued some concurrently (races, lifecycle-triggered follow-ups), and overlapping /actions + /execute requests were reordered by the driver — landing pointerUp a cycle late and swallowing parts of a click. Serializing eliminates the desync. - Cache UtilityScript/InjectedScript as per-document singletons in the page registry instead of rebuilding them (and re-installing their global event listeners) on every evaluate. - Test fixture: reuse one session per worker and reset to about:blank between tests, and stop safaridriver gracefully (SIGTERM + await exit, never SIGKILL) so it deletes its session and unpairs Safari — avoiding the "Continue Session" dialog / "already paired" failures on the next run.
…reation - Trim verbose comments and drop dead WDSession methods. - Reuse one WebDriver session per worker (resetting to about:blank between tests): creating a session re-prompts Safari's "remotely controlled" dialog, so opening it once per worker minimises the prompt versus a fresh session per test. - Retry session creation while Safari still reports the previous session as paired, so a just-released session doesn't fail the next connect.
Runs the basic page-check/page-goto/page-click suites against Safari over the WebDriver backend on a macos-15-xlarge runner (workflow_dispatch + PR on webdriver paths). Temporarily guards the other PR-triggered workflows with `github.head_ref != 'webdriver-safari-poc'` so they don't fire on the draft PR used to exercise the bot. Revert these TEMP guards before merging.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
wd*) backend that drives WebKit over classic W3C WebDriver, alongside the existing webview backend. Reached viawebkit.connectOverCDP('webdriver://host:port')(or a local-launch sentinel).tests/webdriver/reusing thetests/page/suite (PWPAGE_IMPL=webkit-webdriver,npm run wdtest).Draft / CI note
macos-15-xlargebot (runs page-check / page-goto / page-click only).