> Should the same precautions be taken for UPDATE statements ? UPDATES only change the WHERE clause (by appending the product=...), so they should not be problematic.
> PS: Notice that tests have always passed even when product was > duplicated in INSERTs. Indeed SQLite always used to work. At the > moment direct tests against an instance running PostgreSQL or MySQL > are still needed . I have tried setting the TRAC_TEST_DB_URI to point to postgres db and running the tests, but it did not work (every test failed). The problem is, that reset_db tries to perform untranslated queries to the converted database, which cause validation errors (product cannot be null when inserting into the system table).
