perf(test): avoid redundant stack capture

[Skn0tt]

Reuses already captured locations instead of taking a second stack trace.

Node 26, headless Chromium:

• page.evaluate(): 399 → 274 µs (-31%)
• polling expect: 568 → 493 µs (-13%)
• synchronous expect: 164 → 100 µs (-39%)

[Skn0tt]

Inspired by microsoft/playwright-python#3147. I also checked Java and DotNet, couldn't find a significant improvement there.

[github-actions]

Test results for "MCP"

1 failed
❌ [firefox] › mcp/cli-core.spec.ts:179 › dialog-dismiss @mcp-windows-latest-firefox

7785 passed, 1262 skipped

---

Merge workflow run.

[github-actions]

Hi, I'm the Playwright bot and I took a first look at the CI failures.

🟡 The one failure is very unlikely to be caused by this PR, but I can't outright prove it's a flake

[firefox] › mcp/cli-core.spec.ts:179 › dialog-dismiss (bot mcp-windows-latest-firefox) is the only failure in the merged report. This PR only touches test-runner stack/location capture (expect.ts, testInfo.ts), which can't reach the MCP CLI snapshot output — so I lean strongly toward noise, but the only failure of this test on record is on this PR, so I'm calling it yellow rather than green.

Details

Uncertain (leaning not-caused-by-PR)

• [firefox] › mcp/cli-core.spec.ts:179 › dialog-dismiss — failed with expect(inlineSnapshot).not.toContain('MyAlert') where inlineSnapshot was undefined, after the step ran ~15s (looks like the snapshot CLI call hung/returned nothing). The diff here reuses already-captured stack frames for step location/boxedStack in the test runner's reporting path — it never influences a matcher's received value or the MCP CLI's snapshot output. So the diff has no path to this assertion. It's Firefox-on-Windows only; the same test passed on every other browser/OS in this run.

  What keeps it out of the green bucket: per our flake rule I need the same test failing where this PR can't be responsible, and I couldn't find that. In the aggregated CI results this test is essentially never red — mcp-windows-latest-firefox is 179/180 passed, and it's 0 failures across 400+ runs on every other bot. The single red on record is this PR's own run (sha a3587afb, branch skn0tt-optimize-stack-capture), and it wasn't retried, so there's no green-on-rerun to lean on either.

  To settle it: re-run the failed MCP shard. Given the change can't reach this code path, I'd expect it to go green.

_{Triaged by the Playwright bot - agent run}

[github-actions]

Test results for "tests 1"

4 flaky

⚠️ [chromium-library] › library/video.spec.ts:664 › screencast › should capture full viewport `@chromium-ubuntu-22.04-arm-node20`
⚠️ [chromium-library] › library/popup.spec.ts:260 › should not throw when click closes popup `@chromium-ubuntu-22.04-node20`
⚠️ [chromium-library] › library/video.spec.ts:736 › screencast › should work with video+trace `@chromium-ubuntu-22.04-node22`
⚠️ [playwright-test] › ui-mode-test-watch.spec.ts:185 › should watch new file `@windows-latest-node22`

49846 passed, 1146 skipped

---

Merge workflow run.