Fix Windows-only failures in the Electron backend integration tests

[tcobbs-bentley]

Fix Windows-only failures in the Electron backend integration tests

Two causes of the failures were introduced by #9627 and reproduced on Windows in the Node 20 and Node 24 CI legs. The fractional scaling problem has probably always existed.

Should save main window size, position and maximized flag.

Three separate problems in ElectronHost.test.ts:

• Resize issued mid-restore. fix(core-electron): harden window state and RPC shutdown #9627 replaced the unconditional 250 ms sleep after unmaximize() with a waitUntil that returns immediately, because savedMaximized() and window.isMaximized() flip in the same tick. setSize then landed while Windows was still restoring, and the restored bounds won the last debounced write. Added waitForStableBounds() to wait for the window to settle first.
• Fractional display scaling. On a display at non-integer scale (e.g. 250%), bounds round through physical pixels, so the realized size/position can differ from what was requested. The test now asserts that the saved state converges on whatever the window actually reports, plus a separate check that the window changed at all — instead of requiring exactly 250×251 at (50, 75).
• Timeout too tight. waitUntil budget raised from ~1.25 s to ~5 s.

Should ignore a response after its request was removed during shutdown.

Exited with 0x80000003 (STATUS_BREAKPOINT) — a Chromium abort, not an assertion failure. This test never calls ElectronHost.startup(), so it could finish and tear down the main process while Chromium was still initializing. RunSingleTest.ts now awaits app.whenReady() and exits via app.exit() rather than process.exit().

Notes

• Test-only; no shipped behavior change.
• Targets master for backport to release/5.13.x. The affected files are identical on both branches, so the cherry-pick should be clean.
• The Windows leg was run multiple times to verify the fix — the 0x80000003 failure was intermittent and passed on the Node 24 leg.

[Copilot]

Pull request overview

Stabilizes Windows Electron backend integration tests around window restoration, fractional scaling, and Chromium startup.

Changes:

• Waits for Electron readiness before running tests.
• Adds window-bound stabilization and longer polling.
• Records a test-only Rush change entry.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
RunSingleTest.ts	Uses Electron-managed startup and exit.
ElectronHost.test.ts	Makes window-state tests scaling-aware.
tcobbs-electron-race-fixes_2026-08-27-22-34-10.json	Adds the package change record.

Suppressed comments (1)

core/electron/src/test/backend/ElectronHost.test.ts:224

• The position check has the same stale-baseline/all-fields issue: it compares with bounds captured before the maximize cycle and requires both coordinates to change. A prior window-manager adjustment can make this true before setPosition is reflected, while a legitimate move where one coordinate was already equal to the target will time out. Compare against bounds captured immediately before the move and accept either coordinate changing.

  assert(await waitUntil(() => window.getBounds().x !== expectedBounds.x && window.getBounds().y !== expectedBounds.y));

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[GytisCepk]

These size tests are constant problem and have been fixed multiple times. I'll need to review them and figure out what is a long term solution here.

Approving for release

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

core/electron/src/test/backend/ElectronHost.test.ts:228

• The initial BrowserWindow position is chosen by the OS, so it can already be (50, 75). In that case setPosition is a no-op and the new “window changed” assertion waits five seconds and fails even though persistence is working. Choose a target that is guaranteed to differ from the sampled position.

  window.setPosition(50, 75);

[hl662]

Thanks for doing this, I ran into these flaky windows problems all the time months ago in my first MVP for vitest certa. Could never nail down the problem but you got it

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

[tcobbs-bentley]

Thanks for doing this, I ran into these flaky windows problems all the time months ago in my first MVP for vitest certa. Could never nail down the problem but you got it

Don't I know it. The last person to make fixes to this file before this PR was me. Those fixes were due to failures in my ARM64 Linux VM.

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request

[tcobbs-bentley]

@Mergifyio backport release/5.13.x

[mergify]

backport release/5.13.x

✅ Backports have been created

Details

• #9673 Fix Windows-only failures in the Electron backend integration tests (backport #9668) [release/5.13.x] has been created for branch release/5.13.x