in_app_automation 0.2.1

In-app UI automation and an interactive on-device test runner for Flutter: widget finders, tap/type/scroll actions, assertions, and a visual inspector overlay.

0.2.1

• Fix: the on-device inspector overlay no longer blocks taps on the app beneath it. The actionability check and real pointer dispatch hit-test the whole view, so the overlay (wand button / running-test panel) could sit over an app widget and make it report NotActionableException. The overlay now drops out of hit-testing while the engine is tapping, so taps reach the app. Genuine app-owned overlays still (correctly) block.

0.2.0

Real-input engine, a production-grade runner, and CI. Pre-1.0, so some APIs changed; highlights below.

Interaction (breaking behavior change)

• tap now dispatches hit-tested pointer events via GestureBinding instead of invoking widget callbacks directly. Covered / IgnorePointer / disabled widgets are reported not actionable rather than falsely passing.
• enterText drives the real IME pipeline (inputFormatters, onChanged, selection, and optional submit) instead of assigning controller.text.
• Every action runs an actionability loop (exists → enabled → visible → receives events) until a timeout, throwing the most specific reason.
• pumpAndSettle waits for frames to actually settle (param duration → timeout); scrollUntilVisible supports axis, prefers the scrollable containing the target, and animates instead of jumping.

Finders — textContaining (substring/RegExp), byWidget<T>() (subtypes), byTooltip, and first/last/at(n); fixed the byText RichText double-match.

Assertions — all auto-retry now; added hidden, count, enabled, disabled, textContaining, and SoftAssertions. visible checks real visibility; text descends into children.

Runner — new TestRunner core with beforeAll/afterAll/beforeEach/ afterEach hooks, per-test timeouts, retries + flaky detection, and tag/grep filtering. The inspector and headless controller now share it.

Errors — typed hierarchy under AutomationException.

Reports & screenshots — TestReportFormatter (JSON/JUnit/HTML). Screenshots are captured automatically on failure and attached to the result (embedded in the HTML report); TestArtifactWriter (via package:in_app_automation/io.dart) writes reports + a PNG per failed test to disk. AutomationScreenshot.capture() for manual capture.

CI — example/integration_test headless entrypoint (exit-code gated) and a reference GitHub Actions workflow.

Safety — AutomationConfig disables actuation in release builds by default; verboseLogging gates diagnostics.

0.1.0

Packaging, correctness, and honesty pass.

• Add an MIT license (previously a TODO placeholder).
• Fix package metadata: real description, topics, declared platforms (android, ios), and a corrected flutter SDK lower bound (>=3.24.0).
• Stop shipping ad-hoc analyzer dumps (lib/src/analysis.txt, analysis_final.txt) and ignore them going forward.
• Resolve all flutter analyze issues.
• Fix: Expect.visible now verifies the widget is actually visible (present, sized, on-screen, not clipped by a scroll viewport) instead of only checking that it exists in the tree.
• Fix: tapping a disabled control now returns a clear "widget is disabled" error and lets tap() fall through to an enabled ancestor/descendant, instead of failing with a confusing internal error.
• Correct README claims about CI/headless support and Expect.visible.

0.0.1+1

• Added comprehensive README documentation.
• Documented core components: AutomationInspectorWrapper, AutomationRegistry, and AutomationEngine.
• Provided setup and usage guides.