WIP - Add experimental Firefox integration test support

[kzar]

Warning: These changes were initially generated with Cursor, and I am now in the
process of tidying them up.

---

Note

Adds experimental Firefox support to integration tests and CI.

• New Firefox harness: helpers/firefoxHarness.js implements RDP client, background evaluation (FirefoxBackgroundPage), context setup/cleanup
• Extends helpers/playwrightHarness.js with Firefox detection, context wiring, unified routing, addScriptTag shim to bypass CSP on Firefox
• New test background-eval.spec.js validating background evaluation API across browsers; updates fingerprint tests to use addScriptTag
• Adds playwright.firefox.config.js limiting supported tests, single-worker stability; new npm script playwright-firefox
• CI: uploads shard-suffixed artifacts and adds playwright-tests-firefox job (sharded, continue-on-error) installing Firefox and running the new script
• Minor robustness tweaks in backgroundWait.js; README documents experimental Firefox testing

^{Written by Cursor Bugbot for commit 95959b0. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 4 potential issues.

[cursor]

Default manifest version changed breaks direct test execution

Medium Severity

The getManifestVersion() logic change inverts the default return value. Previously it returned 3 unless playwright-mv2 was specified; now it returns 2 unless playwright is specified. When running npx playwright test directly (documented in the README for re-running tests without rebuilding), npm_lifecycle_event is undefined, causing manifestVersion to be 2 instead of 3. This breaks the documented workflow because the test harness would try to load the extension from build/chrome-mv2/dev instead of build/chrome/dev.

---

_{Please tell me if this was useful or not with a 👍 or 👎.}

 

[cursor]

Debug console.log not commented out

Low Severity

An active console.log statement prints to output for every Firefox test. The established convention in this file is to comment out debug logging statements (as seen on lines 66, 75, and 169). This appears to be debug code that was accidentally left enabled given the PR mentions tidying up Cursor-generated code.

---

_{Please tell me if this was useful or not with a 👍 or 👎.}

 

[cursor]

Firefox waitForFunction signature incompatible with caller

Medium Severity

The FirefoxBackgroundPage.waitForFunction has signature (pageFunction, options = {}, ...args) but the existing forFunction in backgroundWait.js calls it expecting Playwright's signature (pageFunction, arg?, options?). When forSetting(bgPage, key) is called, the key argument gets misinterpreted as options instead of being passed to the evaluated function. This would cause silent failures when Firefox tests use functions like forSetting that pass arguments.

---

_{Please tell me if this was useful or not with a 👍 or 👎.}

Additional Locations (1)

• integration-test/helpers/backgroundWait.js#L33-L41

 

[cursor]

Firefox browser and temp directory leak on setup failure

Medium Severity

In createFirefoxContext, if installExtensionViaRDP fails after Firefox has been launched (line 506), the function throws without closing the browser context or cleaning up the temp directory created at line 490. Since context._firefoxUserDataDir is only assigned after the RDP call succeeds, even calling cleanupFirefoxContext wouldn't remove the orphaned temp directory. This causes leaked Firefox processes and accumulating temp directories when extension installation fails.

---

_{Please tell me if this was useful or not with a 👍 or 👎.}