Google Sign-In Works in Debug but Fails from Google Play

This article documents how I approach the feature in a real application: the implementation decision, the failure cases I check, and the release test I use before considering the work complete.

Why release is different

My Dairy is a salary and overtime management project with attendance, custom salary periods, Google authentication, PHP/MySQL APIs, report export and server-controlled application behavior. Because the dashboard, history and salary calculations depend on the same records, a small mismatch in user identity or date logic can make correct database data appear missing in the app. When I work on why release is different, I first decide which layer owns the responsibility. Flutter should handle presentation and interaction, while protected business decisions are checked by the server or platform configuration that actually controls them. This separation prevents a UI workaround from hiding a backend or release problem.

I deliberately test failure paths for why release is different. Network timeout, empty data, invalid input, cancelled authentication, unavailable advertising, or a rejected API request should result in a controlled screen state. The application should not show an endless loader or silently save incomplete information simply because the successful path was the only one tested.

Finding SHA fingerprints

For finding sha fingerprints, I capture the exact state before editing code: input values, user identifier, date range, HTTP status or native error, raw response where safe, and whether the failure happens in debug, locally signed release, or a build installed from Google Play. That evidence usually narrows the problem much faster than changing several files at once.

Another important check is consistency. If the backend calculates a salary period one way but a Flutter history screen sends another month range, both pieces of code may be individually valid while the user sees the wrong result. I therefore compare the parameters and business rules end to end before changing the visual layer.

cd android
./gradlew signingReport
# Windows PowerShell:
.\gradlew signingReport

Debug and upload certificates

I deliberately test failure paths for debug and upload certificates. Network timeout, empty data, invalid input, cancelled authentication, unavailable advertising, or a rejected API request should result in a controlled screen state. The application should not show an endless loader or silently save incomplete information simply because the successful path was the only one tested.

After a fix, I repeat the original failing workflow rather than accepting a successful compilation as proof. For Play-specific behavior I test the Play-delivered artifact; for API problems I verify the real production endpoint; and for data problems I compare the returned JSON with the model used by the screen.

Play App Signing certificate

My Dairy is a salary and overtime management project with attendance, custom salary periods, Google authentication, PHP/MySQL APIs, report export and server-controlled application behavior. Because the dashboard, history and salary calculations depend on the same records, a small mismatch in user identity or date logic can make correct database data appear missing in the app. When I work on play app signing certificate, I first decide which layer owns the responsibility. Flutter should handle presentation and interaction, while protected business decisions are checked by the server or platform configuration that actually controls them. This separation prevents a UI workaround from hiding a backend or release problem.

Firebase configuration

For firebase configuration, I capture the exact state before editing code: input values, user identifier, date range, HTTP status or native error, raw response where safe, and whether the failure happens in debug, locally signed release, or a build installed from Google Play. That evidence usually narrows the problem much faster than changing several files at once.

Testing the Play build

My Dairy is a salary and overtime management project with attendance, custom salary periods, Google authentication, PHP/MySQL APIs, report export and server-controlled application behavior. Because the dashboard, history and salary calculations depend on the same records, a small mismatch in user identity or date logic can make correct database data appear missing in the app. When I work on testing the play build, I first decide which layer owns the responsibility. Flutter should handle presentation and interaction, while protected business decisions are checked by the server or platform configuration that actually controls them. This separation prevents a UI workaround from hiding a backend or release problem.

I deliberately test failure paths for testing the play build. Network timeout, empty data, invalid input, cancelled authentication, unavailable advertising, or a rejected API request should result in a controlled screen state. The application should not show an endless loader or silently save incomplete information simply because the successful path was the only one tested.

My troubleshooting workflow

What this project taught me

The main lesson from My Dairy is that production behavior depends on more than Dart code. Signing certificates, package names, API endpoints, database filters, SDK configuration and store settings are all part of the application. I keep those values in a release checklist and verify the distribution that users will actually install.

Version note: Flutter packages, Android tools, Firebase, Google Play and advertising services change. Check current official documentation before applying version-specific configuration to a production release.