A simulator can verify useful interface behaviour, but it cannot establish every hardware, permission, storage, and lifecycle outcome of a native integration. Test the capabilities that depend on the real device in the environment customers will use.
Choose devices deliberately and record the limits. One successful run is evidence about that configuration, not proof of support across every phone or laptop.
Build the device matrix from the capability
List operating-system versions, device families, hardware requirements, native SDK versions, app artefact, and relevant settings. Include lower-capability supported devices when performance or memory could affect the task.
Flutter's plugin testing guide explains why native and integration checks complement Dart tests. Keep each result tied to the boundary it actually exercised.
Test a fictional camera-and-upload feature
Create harmless image fixtures and a controlled account. Exercise permission request, denial, later grant, capture, cancellation, backgrounding, interrupted upload, restart, and account switching. Inspect the final stored object and parent record, not only the preview.
Add low storage and an unavailable network where those conditions matter. Verify temporary data cleanup and pending-work recovery. If the feature uses external hardware, test disconnection and reconnection with the actual supported device.
Record exact steps and observed outcomes. This article proposes a matrix and claims no device test, benchmark, or certification. The value comes from making future evidence reproducible.
Verify release-like configuration
Use the relevant profile or release artefact for performance and packaging questions, with authorised signing and test services. Debug-only permissions, manual setup, or local SDK paths can conceal production configuration gaps.
Retain safe diagnostic references and version information. Avoid collecting private customer content simply because a device test failed. Controlled fixtures should let another engineer reproduce the problem without production data.
The tradeoff is coverage versus cost. A small representative matrix can reveal major platform risks while staying manageable. Expand when a new operating-system release, native SDK change, or actual failure introduces a reason, rather than repeating every device test after unrelated edits.
The Social App portfolio account documents Flutter, native iOS, and backend contributions. It does not publish results for this fictional camera workflow or guarantee behaviour on untested devices.
Include a supported older device when hardware capacity or operating-system behaviour could change the integration outcome.
Questions about physical-device checks
Are simulator tests useless?
No. They support repeatable logic and interface checks. Use physical devices for the boundaries where real hardware and native conditions matter.
Must every supported device be tested manually?
Use a justified representative matrix and disclose its limits. Prioritise the configurations most relevant to the changed capability and audience.
What should the evidence include?
Include device, OS, app and SDK versions, fixture, sequence, observed result, and remaining gaps. Bring failures to a native-device review.