[PM-39361] fix: Restart card scanner after camera interruption or reopen

[fedemkr]

🎟️ Tracking

https://bitwarden.atlassian.net/browse/PM-39361

📔 Objective

Fixes a blank camera screen when opening the card scanner after the Add Card view was closed while the scanner was still open, or after the device locks and unlocks while the app is on the Add Card screen.

Changes in CardScannerView.swift:

• Replace try? startScanning() with explicit error handling; on failure a new
  onScannerUnavailable callback is fired (dispatched after the current SwiftUI
  update pass to avoid re-entrant state mutations).
• Implement dataScanner(_:becameUnavailableWithError:) in the Coordinator to
  forward runtime camera-interruption events through the same callback.
• Add @Environment(\.scenePhase) observation in CardScannerWrapperView; when
  the scene becomes .active while scanning is supposed to be running, a
  stop-then-restart cycle is triggered.
• Retry up to 2 times with a 300 ms delay before auto-dismissing the sheet, so
  the user is never left with a permanently blank screen.

[github-actions]

🤖 Bitwarden Claude Code Review

Overall Assessment: REQUEST CHANGES

Reviewed the card scanner restart logic in CardScannerView.swift and its new unit tests. The change adds explicit error handling around startScanning(), a becameUnavailableWithError delegate forward, scene-phase observation, and a retry-then-dismiss flow. The error-forwarding and dispatch-after-update-pass approach are sound. One correctness concern remains around how the retry budget is shared between failure and scene-phase restarts.

Code Review Details

• ⚠️ : Benign foreground transitions consume the retry budget and can auto-dismiss a working scanner; counter never resets on a successful restart.
  • BitwardenShared/UI/Vault/VaultItem/AddEditItem/AddEditCardItem/CardScannerView.swift:65-87

[claude]

⚠️ IMPORTANT: Benign foreground transitions consume the retry budget and can auto-dismiss a working scanner.

Details and fix

scannerRetryCount is shared between two very different triggers: genuine scanner failures (onScannerUnavailable) and routine scenePhase == .active transitions (line 66). It only resets on .onAppear, which does not re-fire when the app is backgrounded and foregrounded while the sheet stays presented.

Trace:

• background → foreground Dependency Dashboard #1 → restartScanning() → scannerRetryCount = 1
• background → foreground re-open Update bitwarden/gh-actions digest to 67ab95d #2 → scannerRetryCount = 2
• background → foreground Update dependency prettier to v3 #3 → guard fails → dismiss()

After two normal app-switches the sheet closes on the next foreground even though the camera is fine. A successful restart also never decrements the counter, so each recovery permanently shrinks the budget available for a later real failure.

Consider resetting scannerRetryCount = 0 on a successful restart (e.g. after isScanning = true settles), or only counting onScannerUnavailable-driven restarts toward the cap rather than scene-phase restarts.

[matt-livefront]

❓ Is the dispatch here just to avoid updating the view in the middle of a view update?

[matt-livefront]

🤔 Does the change of scenePhase trigger for you? I think the scene phase might only work if you're using SwiftUI navigation, not UIKit. You might need to observe the notification center notifications instead.