Fix Pandora authentication failures (#2949)
* fix: Pandora auth token expiry during long listening sessions
Implemented automatic re-authentication mechanism for Pandora provider
to prevent auth token expiry during extended listening sessions.
- Added token expiration tracking with AUTH_TOKEN_LIFETIME (50 minutes)
- Added ensure_valid_auth() method to proactively re-authenticate before
token expiration (with 5 minute buffer)
- Updated _api_request() to call ensure_valid_auth() before all API calls
- Added retry logic for 401 errors to handle reactive re-authentication
This follows the same pattern used by the Tidal provider for maintaining
valid authentication tokens during long-running sessions.
Fixes issue where users reported authentication expiring while listening
to Pandora stations.
* improvement: Check for explicit token expiry from Pandora API
Updated authentication to check if Pandora returns expiry information
(expiresIn or expires_in fields) and use it if available, rather than
always assuming a hardcoded lifetime.
- Checks for expiry fields in authentication response
- Uses actual expiry if provided by API
- Falls back to conservative 50-minute estimate if not
- Added logging to show which approach is being used
This makes the implementation more robust and will automatically adapt
if Pandora provides expiry information in their response.
* refactor: Switch to reactive-only auth handling for Pandora
Changed from proactive time-based re-authentication to reactive
401-error handling only. This is a better approach because:
1. User reports show auth failures occurring anywhere from 5 minutes
to several hours, indicating it's NOT a simple token timeout
2. Proactive re-auth would mask the real issue and prevent diagnosis
3. Reactive handling will log actual failure patterns to help identify
the root cause (concurrent logins, IP changes, rate limiting, etc.)
Changes:
- Removed AUTH_TOKEN_LIFETIME and AUTH_REFRESH_BUFFER constants
- Removed ensure_valid_auth() proactive check method
- Removed _auth_expires_at expiry tracking
- Added _auth_time to track when authentication occurred
- Enhanced 401 error handling to log how long tokens lasted
- Simplified to reactive-only approach: retry on 401 with fresh auth
This will help diagnose the actual cause of auth expiry issues while
still preventing playback interruption by automatically re-authenticating
when needed.
* chore: Change Pandora auth log messages to debug level
Changed authentication success and token expiry messages from info/warning
to debug level to reduce log noise for routine operations.
* chore: Remove diagnostic timing code from Pandora auth
Removed _auth_time tracking and associated debug logging that calculated
token lifetime. Fix is confirmed working, diagnostic code no longer needed.
* chore: Change authentication success log back to info level
* fix: Remove duplicate type annotation to fix mypy error
* fix: Handle network errors gracefully in Pandora stream handler
Added ProviderUnavailableError handling to _handle_stream_request to
prevent crashes when network errors (DNS timeouts, connection failures)
occur during fragment fetching.
- Wrapped all _get_fragment_data calls in try-except block
- Added ProviderUnavailableError exception handler
- Returns 503 Service Unavailable instead of crashing
- Logs error for diagnostics
* refactor: Use recursive retry pattern for Pandora auth failures
Refactored _api_request to use recursive call instead of duplicating
error handling code in retry block.
- Added 'retry: bool = True' parameter to _api_request
- On 401 error, re-authenticate and recursively call with retry=False
- Eliminates ~20 lines of duplicate error handling code
- Cleaner and more maintainable
Credit: @MarvinSchenkel for the suggestion
* fix: Add type annotation to satisfy mypy no-any-return check
* style: Remove superfluous else after return (RET505)
Fixed ruff linting error - removed unnecessary else clause after
return statement in retry logic.
---------
Co-authored-by: Claude <noreply@anthropic.com>