🐛 - Bugbot overlooks the stale trigger guard and reports a browser test failure that does not occur #902

Is there an existing issue for this?

• I have searched the existing issues.

Which actions are affected?

Pull Request

Which platforms are affected?

No response

Description

The finding says the first call to propose() opens the replacement picker and then emits a valid request, causing its zero-calls assertion to fail. The first click is synchronous and reaches the old trigger before the new state renders. #show() rejects it because the rendered state differs from the assigned state. The subsequent click on the newly rendered option is also rejected because the popup was never opened. The next deliberate proposal opens the current popup and emits exactly one request.

Reproducing the issue

1. Check out ha-ui commit fbab810406b7c6b286987ce40cf5c1d76376ab37 in a clean directory, install the frozen lock and provision Chromium.
2. Run pnpm exec vitest run --project browser tests/browser/select.test.ts. All 22 tests pass in real Chromium, including the stated zero-calls and one-call assertions. The hosted pnpm test step also passed on this commit.
3. Read propose() and #show() together: click first, then await render; mismatched rendered state prevents opening. #choose() requires the popup to be open.

Expected: check the guard and run the asserted test before publishing a claimed failure. Reconcile the report without altering the passing test or production guard.

copilot Version

v3 (3e7ceea)

Relevant Log Output

Additional context and comments

Exact review · Exact CI. The separate static catalog raster difference is unrelated to the source browser test.