Revert "feat(getByRole): add support for aria-busy (#41269)"

[dgozman]

This reverts commit f215af1.
Reopens #36233.

[github-actions]

🟢 CI is clear — the one failure is a pre-existing flake

Hi, I'm the Playwright bot and I took a look at the failing CI. This PR reverts the aria-busy getByRole support (roleSelectorEngine.ts, roleUtils.ts, locatorUtils.ts, docs, types). The single failure is an MCP screencast test that never touches that code path — and it flakes on plenty of SHAs the PR can't be responsible for. Nothing here looks caused by this PR.

Details

Overall: 1 failure across the merged reports, and it's a known flake. The diff is confined to the role selector engine and getByRole types/docs; the failing test exercises the MCP show --annotate screencast flow, which has no connection to aria-busy or role matching.

Pre-existing flake / infra

• [webkit] › mcp/annotate.spec.ts:447 › should switch screencast to -s session on show --annotate (@mcp-macos-latest-webkit) — flake. In the test-results DB this test flips verdict across unrelated runs on every browser: webkit failed on 8 distinct SHAs and passed 378 times, and the webkit failures include other PRs (fix(core): surface abort reasons in errors #41752, fix(page): collect DOM elements retained by locator hit-target interceptor #41575, feat(dashboard): add a debugger actions panel #41711, chore(protocol): make timeout a protocol-level field #41712, devops(test-results-db): CLI to compact CI test results into DuckDB #41673) and push builds with no PR — all places this revert can't have caused it. Same intermittent pattern on chromium (7 fails), firefox (12), and chrome (4). The revert doesn't go anywhere near the MCP annotate/screencast flow.

_{Triaged by the Playwright bot - agent run}

[github-actions]

🟡 One failure I can't clear — likely a race, but I couldn't prove it

Hi, I'm the Playwright bot and I took a look at the failing CI. This PR reverts the aria-busy getByRole support (role selector engine, roleUtils.ts, locatorUtils.ts, docs, types). Two tests failed across the merged reports. The MCP one is a known flake. The trace one is an action-ordering race that the revert can't plausibly reach — but it's the only time that test has ever failed, so I can't call it a proven flake.

Details

Overall: 2 failures. One is a well-established flake on unrelated SHAs; the other is a single ordering race with no cross-run history, so it stays uncertain. The diff is confined to role matching (getByRole/aria-busy) and its types/docs — neither failing test exercises that path.

Uncertain

• [playwright-test] › playwright.trace.spec.ts:1267 › should not nest top level expect into unfinished api calls (@ubuntu-latest-node20) — the action tree came back with Expect "toBeVisible" and GET "/hang" swapped:

  -   "GET \"/hang\"",
      "Expect \"toBeVisible\"",
  +   "GET \"/hang\"",
  

  That's a timing race in which async action lands in the trace first, against a /hang route that never responds. It failed only on ubuntu-latest-node20 and passed on macos, windows, and node22/24/26 in the same run. In the results DB this test is 266/266 green except this one run — and that run is this PR, so I have no independent SHA where the same test flips. By our flake rule that's not enough to call it a flake. That said, the revert only touches getByRole/aria-busy, and this test uses getByText + route.fetch with no role matching, so there's no mechanism I can see for the diff to reorder trace actions. My read: most likely a pre-existing race, but I can't prove it. A rerun of tests 1 would settle it.

Pre-existing flake / infra

• [webkit] › mcp/annotate.spec.ts:447 › should switch screencast to -s session on show --annotate (@mcp-macos-latest-webkit) — flake. In the results DB this test flips verdict on every browser: webkit failed 6 runs / passed 190, firefox 11/180, chromium 5/189, chrome 2/193. The webkit failures land on unrelated PRs (devops(test-results-db): CLI to compact CI test results into DuckDB #41673, feat(dashboard): add a debugger actions panel #41711, fix(core): surface abort reasons in errors #41752) and push builds with no PR — all places this revert can't be responsible for. The MCP show --annotate screencast flow has nothing to do with role matching.

_{Triaged by the Playwright bot}

_{Triaged by the Playwright bot - agent run}

[Skn0tt]

ditto

[github-actions]

Test results for "tests 1"

3 flaky

⚠️ [chromium-library] › library/video.spec.ts:664 › screencast › should capture full viewport `@frozen-time-library-chromium-linux`
⚠️ [chromium-library] › library/video.spec.ts:356 › screencast › should work for popups `@chromium-ubuntu-22.04-node20`
⚠️ [firefox-library] › library/heap.spec.ts:223 › should not leak workers `@firefox-ubuntu-22.04-node20`

49682 passed, 1156 skipped

---

Merge workflow run.

[github-actions]

Test results for "MCP"

7760 passed, 1249 skipped

---

Merge workflow run.