Tested configurations and known limitations
What ElevenClock-rs has actually been run on, and what has not been checked. Raw screenshots, logs and measurements from development are in Git history, not in the tree.
Reference machine
Windows 11 Home 10.0.26300.9711 (26H2) x64 in a virtual machine, one 1536×1106 display at
125%, release build (x86_64-pc-windows-msvc). Earlier development runs also used
Windows 11 10.0.26200 and two 1280×720 displays at 100%. The maintainer additionally used
the app for a full workday on a physical laptop and handled the multi-display, mixed-DPI
and Explorer-recovery checks on their own hardware.
Checked
| Area | Result |
|---|---|
| Settings | Opens from the taskbar and Start; presets, preview, Apply, Cancel, Reset and Exit; invalid stored preferences recovered (7/7); overflow rejected in Settings; 20/20 repeated launch/Settings cycles with flat handle/GDI/USER counts |
| Clock replacement | Primary taskbar on 26300.9711; split secondary clock (separate bell, no Show Desktop) located and replaced |
| Taskbar states | Auto-hide reveal 3/3; full-screen window hide/restore; tray re-layout (do-not-disturb bell) followed within 5 s |
| Neighbours | Native calendar opens on click; bell and Show Desktop clicks unaffected |
| Session and time | Lock/unlock; time-zone changes and clock jumps (the clock matched local time to the second afterwards) |
| Recovery | Explorer restart ×3 with the same process and flat handles; a forced kill of the app reveals the native clock; quiet --startup launch |
| Startup and install | Per-user install, Start menu launch, startup on/off/repair without UAC on a standard account; update, uninstall and rollback between two development builds |
| Managed ZIP | On the hosted Windows runner (disposable account): the real launcher deploys from a temporary source folder; the app reports Managed: yes; Startup registers the local launcher; a changed source replaces the local copy on the next launch; setup refuses while the managed copy is present; --uninstall removes it |
| Managed inventory | Local Windows x64 validation: supervised local launcher reporting with an already-running app and invalid/missing central app payload; stable installation ID; configuration revocation; configuration reads and report replacement over loopback SMB (\\localhost\C$); administrator summary command and failure-preserves-output behavior. Native launch-to-report validation used a process-local scratch HKCU view to isolate the existing installer registration |
| Support | Status, Refresh and Copy diagnostics; the report contains no user name or path |
| Templates | Local Windows x64: first/second-instance imports, draft preservation, Apply/Cancel, local preset save/load/delete, clipboard sharing, native dialog tab order and accessible button names; per-user elevenclock: enable/disable and real Shell dispatch including Unicode, ampersands and percent literals. The foreground keyboard-injection harness could not activate the dialog on this local desktop; direct native control/dispatch checks ran separately |
| Accessibility | UIA name, role and Invoke on the clock; keyboard access to Settings and the tray menu; Settings layout at 100/125/150/200% and in high contrast (CI) |
Resource use
Ten minutes per format on the reference machine (8 logical processors), release build:
| Format | CPU (% of machine) | Private working set |
|---|---|---|
Date only (%g%V.%u) | 0.0116% | 3.01 MiB |
Minutes (%H:%M) | 0.0110% | 3.39 MiB |
Seconds (%H:%M:%S) | 0.0080% | 2.71 MiB |
No growth within any run. To measure again, choose the format in Settings with one app instance running, then:
./scripts/measure-rust-resources.ps1 -DurationSeconds 600 -IntervalSeconds 10 -Label minutes -Workload '%H:%M'The script cannot see timer-wake or discovery rates.
Not checked, or known limitations
- Other Windows builds. Only 26300.9711 (and 26200 during development) were run. Other stock Windows 11 x64 builds are assumed to work. If discovery cannot confirm the clock, the native clock is left alone.
- Narrator speech. The UIA data and events it relies on are verified; nobody has listened to it.
- Managed ZIP. The launcher flow (deploy, managed mode, launcher startup entry, update on the next launch, installer refusal, uninstall) runs on the hosted Windows runner only. Inventory configuration/publication and summary reads were checked over loopback SMB; this does not verify another server or per-user share/NTFS isolation. Not tried: a remote file share, domain-account resolution, the Local intranet zone setup, a VPN or offline start, AppLocker or Software Restriction Policies, or Group Policy or Intune deployment. The managed deployment guide says where it relies on this.
- Actual sign-out/sign-in startup and a clean first-run machine without developer runtimes.
- Secondary taskbars have no
TrayNotifyWnd, so a tray re-layout there waits for the next discovery. - Tests disturb a live desktop.
cargo test --workspace --lockedopens UI Automation fixture windows that take focus. Do not run it during interactive checks. - External minimize. A tool that minimizes the overlay from another process (some screen-control tools do) used to latch an overflow; the overlay now restores itself, but only this one tool was tried.