V1.5.13

[flarco]

New Features

• OAuth2 authorization code flow with configurable redirect URL and port: The authorization code flow now supports a configured redirect_uri, using its host, port, and callback path instead of always binding to a random local port.
• Render OAuth2 scopes before configuration: OAuth2 scopes are now rendered (variable substitution) before being passed to the OAuth2 config, enabling dynamic scope values.
• CLIENT_SECRET support for Azure and OneLake connections: CLIENT_SECRET and client certificate properties are now properly passed through for Microsoft Fabric ABFS and Azure file system authentication.
• Unicode column name support: CSV header cleaning now preserves Unicode letters and numbers (e.g. Chinese, Japanese, accented characters) instead of stripping them.
• String length inference for file sources with column_typing: column_typing string max_length rules now apply to non-database sources (CSV/files), enabling auto-inference of varchar() precision when loading into databases like Oracle.

Bug Fixes

• BigQuery GetSchemata table retrieval: Fixed schema value not being cast to string, which caused table listing to fail when discovering BigQuery schemas.
• Oracle SQL*Loader out-of-memory handling: SQL*Loader now runs with context cancellation support (CommandContext) and surfaces actionable SQL*Loader-700 / out-of-memory errors instead of generic failures.
• Oracle SQL*Loader pipe error handling: Closed-pipe errors during CSV writing to SQL*Loader stdin are now caught and propagated correctly instead of being silently ignored.
• DuckDB connection leak during chunk loading: Added missing defer dbConn.Close() calls in ClearTableForChunkLoadWithRange and ProcessChunks to prevent connection leaks during chunked replication.
• Chunking error message without update_key: Fixed a spurious nil error being passed to g.Error when update_key is missing for stream chunking.
• Compact CDC State: Fixed CDC state so that GTID messages are stored compact