I've a patch that has only one bug that I'm aware of left: in autocommit mode, a PRAGMA connect_triggers=1; will fire any connect triggers and any before commit triggers, but not any after begin triggers -- mildly annoying, but tolerable.
Writing tests, however, I discovered something subtle and obnoxious: - If you have any begin triggers referencing non-existent tables, this cannot be cleaned up because the begin trigger always fires before any statements that could clean this up can execute, and this results in an exception that prevents anything from being done to clean up the schema. The way I'm dealing with this is by having a pragma enable/disable all DB triggers (connect, begin, commit) -- if one gets into trouble, disable DB triggers, fix the DB, re-enable DB triggers. (I could also just ignore DB triggers with parse errors in their trigger bodies, that's easy, but then how would one communicate the error to the user? So I reject this alternative.) Any alternative advice regarding this? Nico -- _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
