Audit logs were not recording any changes because after_flush runs
after SQL is emitted; by then session.new, session.dirty, and
session.deleted can be cleared and attribute history for updates is
often consumed, so the handler saw nothing to log.
Changes:
- Add receive_before_flush: process session.dirty (updates) and
session.deleted (deletes) while history is still valid; stash
session.new (creates) in session.info for after_flush.
- Simplify receive_after_flush: only handle pending creates from
session.info (instances now have ids), then session.flush() so
audit rows are in the same transaction.
- Register receive_before_flush for before_flush on Session,
sessionmaker class, and SignallingSession.
- Make receive_before_flush accept (session, flush_context, instances)
to match SQLAlchemy's before_flush signature.
- Remove db.session.flush() from AuditLog.log_change to avoid
nested flush; rely on main flush or explicit flush in after_flush.
- check_audit_logging.py: use entity_type='TimeEntry' to match
get_entity_type (model __class__.__name__).
- test_audit_logging: assert at least one AuditLog for create/update/
delete; use test_client for create; fix update to merge then mutate.
- Improve audit logging error messages to distinguish table missing errors from other failures
- Add warning-level logging for audit_logs table missing scenarios with migration guidance
- Update audit event listener with better error detection and logging
- Add comprehensive diagnostic script for checking audit logging setup
- Update UI templates (base.html, admin forms, user settings, profile pages)
- Extend audit logging support across routes (admin, api, permissions, reports, timer, user)
- Add extensive test coverage for admin user management functionality
- Update time tracking service and user model with audit logging integration
Dries Peeters