fix(highlight): resolve highlights relative to the frame of the locator

[Yury Semikhatsky (yury-s)]

Summary

• HighlightController resolved every selector from the main frame, so frame.locator(...).highlight() highlighted elements in the page instead of the frame
• Thread the frame through the highlight entry, key entries by frame so the same selector can be highlighted from different frames, and drop entries when their frame detaches

Fixes #42850

[github-actions]

Test results for "tests 1"

4 flaky

⚠️ [chromium-library] › library/browsercontext-page-event.spec.ts:173 › should work with Ctrl-clicking `@chromium-ubuntu-22.04-arm-node20`
⚠️ [chromium-library] › library/video.spec.ts:762 › screencast › should work with video+trace `@realtime-time-library-chromium-linux`
⚠️ [firefox-library] › library/browsercontext-cookies-third-party.spec.ts:257 › third party 'Partitioned;' cookies `@firefox-ubuntu-22.04-node20`
⚠️ [firefox-library] › library/browsercontext-cookies-third-party.spec.ts:470 › top level 'Partitioned;' cookie and same origin iframe `@firefox-ubuntu-22.04-node20`

52160 passed, 1253 skipped

---

Merge workflow run.

[github-actions]

Test results for "MCP"

2 failed
❌ [chrome] › mcp/cli-webmcp.spec.ts:135 › webmcp-list stitches tools across frames @mcp-macos-latest-chrome
❌ [firefox] › mcp/cli-session.spec.ts:54 › idle timeout shuts the session down @mcp-windows-latest-firefox

8690 passed, 1474 skipped

---

Merge workflow run.

[github-actions]

🟢 Both failures are pre-existing flakes — this PR is clear

Hi, I'm the Playwright bot and I triaged the red CI here. The two failures are both MCP tests that
fail the same way on other people's PRs, and neither touches the highlight code this PR changes.

Details

The diff is confined to highlightController.ts / frameDispatcher.ts (Frame.highlight /
Frame.hideHighlight) plus a new tests/library/locator-highlight.spec.ts case. HighlightController
only does work once highlight() is called, and no MCP test in this report calls it. The library
suite that does cover the changed code is fully green (52160 passed, 4 unrelated flaky).

Pre-existing flake / infra

• [chrome] › mcp/cli-webmcp.spec.ts:135 › webmcp-list stitches tools across frames — the iframe's
  tool doesn't register in time, so the assertion sees Found 1 WebMCP tool(s) instead of 2. The
  byte-identical failure has hit unrelated PRs: chore(coverage): istanbul instrumentation for the trace viewer and html reporter #42785 (twice, runs
  35406976347 and
  35411614391) and feat(test-runner): support locks in test.describe.configure #42792
  (35419927784), all on
  mcp-macos-latest-chromium. Across the results DB the test is 4 failed of 574 runs — a rare,
  macOS-only registration race.

• [firefox] › mcp/cli-session.spec.ts:54 › idle timeout shuts the session down — a well-established
  flake on mcp-windows-latest-firefox: 57 of 182 runs failed (31%), while the same test is
  0-for-1884 failures on chromium, chrome, msedge and webkit. Recent unrelated victims include fix(fetch): apply proxy.bypass to redirect targets #42848,
  fix(mcp): rebind page-registered WebMCP tools when the tab or frame changes #42821, fix(highlight): highlight in the world the selector resolves in #42864, feat(reporter): and printOnlyFailures and --reporter-only-failures #42766, fix(selectors): reset regex state before matching each element #42818 and two plain push runs
  (35748035194,
  35769728467). Timing-sensitive
  shutdown assertion on a slow Windows bot.

I'm a first pass, not the final word — but nothing in the evidence points at this change.

_{Triaged by the Playwright bot - agent run}