Changes between Initial Version and Version 1 of Ticket #36779, comment 9
- Timestamp:
- Dec 23, 2025, 9:36:05 PM (6 days ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #36779, comment 9
initial v1 1 1 I'm not against a feature flag at all if we have a use for it but I can't reproduce the test failures you are alluding to by running the full SQLite test suite and it seems to be a propery required by third party backends to pass the full schema and migrations test suites. 2 2 3 Given `sqlite3.DatabaseSchemaEditor.delete_model` completely ignores the `cascade` argument by when calling `base.DatabaseSchemaEditor.delete_model` though `super()` in the proposed patch and that only the later makes use of `sql_delete_table_cascade` I don't see how any code path can issue a `DROP TABLE {table} CASCADE` on SQLite.3 Given `sqlite3.DatabaseSchemaEditor.delete_model` completely ignores the `cascade` argument when calling `base.DatabaseSchemaEditor.delete_model` though `super()` in your proposed alternative changes in the PR (and the quoted patch uses a different approach by redefining `sql_delete_model_cascade`) and that only the later makes use of `sql_delete_table_cascade` I don't see how any code path can issue a `DROP TABLE {table} CASCADE` on SQLite in either versions of the patch. 4 4 5 5 > This approach ensures the API is extensible and backward-compatible without breaking SQLite or third-party backends.