Skip to content

validation.md

Latest commit

 

History

History
67 lines (54 loc) · 3.62 KB

validation.md

File metadata and controls

67 lines (54 loc) · 3.62 KB
Raw

Development validation

Run make check for formatting, unit tests and the Apple Development signed bundle. The tests cover grid bounds and gutters, asymmetric display traversal, constrained windows, display disconnection, rule migration, modifier combinations, setup persistence, permission-card placement and conditional AX resize correction.

Controlled window benchmark

Build and open the dedicated test fixture:

scripts/build-fixture.sh

The script prints its generated app path. Open that app, then run the installed Encaje binary with ENCAJE_QA=1 and --benchmark-pid <fixture-pid>. The command rejects any process outside the fixture's exact bundle/window allowlist.

The benchmark alternates old and optimized AX command paths on one fixture, compares final geometry using independent scalar reads, and restores its original frame. It also exercises Z → Q → E → Z → Q with one command per step. Any mismatched final rectangle or failed restoration fails acceptance, regardless of how quickly the AX calls return.

Initialize NSApplication.shared before any headless NSScreen query. Read visibleFrame fresh for the current test; an uninitialized probe can report the wrong usable area on a secondary display.

The measured duration covers synchronous AX work and bounded verification. It excludes physical keyboard delivery, event-queue delay before the command, and compositor presentation. A fast failed resize is not an improvement; AX success alone does not establish the intended window size. AppKit may clip an initial resize by its previous position even on the same display; the engine conditionally retries when the observed size proves that correction is needed.

Interactive checks

  • Record Command+W without closing Settings; record Control+Option+key; cancel and clear recording; reject exact duplicates while allowing distinct modifiers.
  • Use the full navigation row, edit a named temporary zone, drag a selection and change its numeric bounds. Restore or remove only that test zone.
  • Save/restore/delete an owned test workspace; preserve unrelated windows.
  • Add/remove a temporary exclusion; check startup preference without retaining a changed login setting after QA.
  • Open/close the menu panel by icon, outside click and Escape; Configuration/About must retain focus, and closed views must release monitors/rendering work.
  • Check the app icon in both its UI and the system's representation.
  • Click empty painted edges of pause/resume and padded buttons; verify the action, not only Accessibility activation. Confirm the language picker aligns with switches.
  • Open Settings and Welcome in both languages; confirm integrated traffic lights, no visible titlebar text or separator, unclipped content and native close behavior.
  • After closing the UI and settling for at least one minute, sample process CPU and memory over multiple intervals. Report bundle size separately from ignored development caches and rollback artifacts.

Unequal-display return regression

Exercise right → left → neighboring right → neighboring left → right → original left → original right with one physical keypress per step. Verify complete usable height in both directions after settling; an outward-only test misses stale AppKit screen limits on the return. Corrections run only for unchanged clipped frames and use per-window tickets so a workspace batch does not cancel earlier windows.

Permission image layout is verified against a 256-point source image: the hosted native drag view must remain 58 × 58 points. App-owned before/after rendering can verify this without resetting Accessibility or requesting screen recording.