The tests needed before a Flutter release depend on what changed and what customers could lose if it fails. A large suite of shallow checks can miss the one native or data-lifecycle boundary affected by the release.
Start with the change, then trace its consequences through the user workflow. Keep core regression checks stable and add focused verification where the new behaviour creates uncertainty.
Match the test level to the question
Use focused logic checks for business decisions, widget checks for interaction states, and integration or device checks for boundaries that depend on real services, storage, or native behaviour. Flutter's testing overview describes these categories and their tradeoffs.
A mocked plugin response can verify your error display without establishing that the real plugin returns that response correctly. State what each check proves so the release review does not overstate coverage.
Review a fictional attachment-release change
Suppose a messaging feature now compresses photos before upload. The change affects image selection, permissions, processing, temporary storage, queue identity, upload recovery, and recipient display.
Test a supported image, a large image, an interrupted selection, denied permission, failed compression, process termination, and reconnect after a pending upload. Check the original and compressed object lifecycle so a successful preview does not conceal an orphaned file.
Keep a normal text-message regression case because the attachment change may share queue or send controls. Test an upgrade from a supported prior app version with pending attachments, not only a fresh installation.
Choose representative devices deliberately
Include the devices and operating-system versions that materially affect the changed capability. A simulator can help repeat interface cases, while physical devices reveal some native, performance, storage, and permission behaviour that a fake cannot reproduce.
Record the artefact, environment, fixtures, executed cases, and unresolved limits. Separate a known cosmetic issue from a failure that loses unsent work or crosses account boundaries. The release decision should make that consequence visible.
The tradeoff is confidence versus execution cost. Run the meaningful checks appropriate to the change, then stop broadening the suite without a reason. Repeatedly rerunning unrelated cases can consume time while adding little evidence about the actual risk.
The Social App case study documents Flutter and native messaging work. This attachment release is fictional and does not describe its test results or release process.
Questions about Flutter release tests
Is a coverage percentage a release decision?
It is one signal about executed code, not proof of meaningful scenarios. Review the behaviours and failure boundaries actually covered.
Must every release test every device?
Use a supported matrix informed by the change and audience. Explain the sampling limits and prioritise devices relevant to the affected capability.
What should the review contain?
Include change impact, core regressions, native checks, upgrade state, remaining failures, and owners. Bring gaps to a Flutter release review.