windows-desktop-e2e

End-to-end testing for Windows native desktop applications using pywinauto with the UIA (Windows UI Automation) backend, covering WPF, WinForms, Win32/MFC, and Qt (5.x and 6.x). All supported frameworks expose a tree of UIA elements — a built-in Windows accessibility API — whose properties tests can read and act on, with each framework shipping its own UIA provider. A reliability table rates automation quality by framework (WPF and Qt 6.x excellent, WinForms and Qt 5.15+ good, Win32/MFC and older Qt fair), noting mappings such as WPF x:Name to AutomationId and WinForms AccessibleName. Setup installs pywinauto with pytest, pytest-html, Pillow, and pytest-timeout on Windows with Python 3.8+, optionally ffmpeg for screen recording, and recommends Accessibility Insights for Windows for inspecting the UIA tree before writing tests. Its most emphasized practice is testability: giving every interactive control a stable AutomationId before writing tests, with per-framework instructions (WPF x:Name, WinForms AccessibleName, Win32/MFC resource IDs, and a dedicated Qt section). It lays out a Page Object Model project structure with conftest.py app-launch fixtures, config, page classes, and an artifacts directory, and provides a reusable BasePage with prioritized locators (AutomationId first, then name, then class plus index as a fragile fallback), waits (wait_visible, wait_gone, wait_window for new dialogs, and a generic polling wait_until for when UIA events are unreliable), actions (click, type_text with a keyboard fallback for incomplete Qt 5.x Value Pattern support, and get_text), and screenshot capture. Example LoginPage and conftest fixtures show the pattern in practice, including setting QT_ACCESSIBILITY. It clarifies when not to use it — web apps (use Playwright), Electron/CEF/WebView2, and mobile — and covers integrating desktop E2E into CI/CD on GitHub Actions windows-latest runners, plus diagnosing flaky automation.