Hi,
I used the v9.1b2 version of pg_dumpall to dump a v8.4.7 database. The resulting dumpfile generated errors when I tried to read it back in under v9.1b2. The complaint is that it's referencing a field in pg_trigger that has been deleted in 9.1 ..
CREATE VIEW vtriggers AS
SELECT (current_database())::information_schema.sql_identifier AS trigger_catalog, (n.nspname)::information_schema.sql_identifier AS trigger_schema, t.tgenabled, (t.tgname)::information_schema.sql_identifier AS trigger_name, (em.text)::information_schema.character_data AS event_manipulation, (current_database())::information_schema.sql_identifier AS event_object_catalog, (n.nspname)::information_schema.sql_identifier AS event_object_schema, (c.relname)::information_schema.sql_identifier AS event_object_table, (NULL::integer)::information_schema.cardinal_number AS action_order, (NULL::character varying)::information_schema.character_data AS action_condition, ("substring"(pg_get_triggerdef(t.oid), ("position"("substring"(pg_get_triggerdef(t.oid), 48), 'EXECUTE PROCEDURE'::text) + 47)))::information_schema.character_data AS action_statement, (CASE WHEN (((t.tgtype)::integer & 1) = 1) THEN 'ROW'::text ELSE 'STATEMENT'::text END)::information_schema.character_data AS action_orientation, (CASE WHEN (((t.tgtype)::integer & 2) = 2) THEN 'BEFORE'::text ELSE 'AFTER'::text END)::information_schema.character_data AS condition_timing, (NULL::character varying)::information_schema.sql_identifier AS condition_reference_old_table, (NULL::character varying)::information_schema.sql_identifier AS condition_reference_new_table, (NULL::character varying)::information_schema.sql_identifier AS condition_reference_old_row, (NULL::character varying)::information_schema.sql_identifier AS condition_reference_new_row, (NULL::timestamp with time zone)::information_schema.time_stamp AS created FROM pg_namespace n, pg_class c, pg_trigger t, ((SELECT 4, 'INSERT' UNION ALL SELECT 8, 'DELETE') UNION ALL SELECT 16, 'UPDATE') em(num, text) WHERE ((((((n.oid = c.relnamespace) AND (c.oid = t.tgrelid)) AND (((t.tgtype)::integer & em.num) <> 0)) AND (NOT t.tgisconstraint)) AND (NOT pg_is_other_temp_schema(n.oid))) AND (((((pg_has_role(c.relowner, 'USAGE'::text) OR has_table_privilege(c.oid, 'INSERT'::text)) OR has_table_privilege(c.oid, 'UPDATE'::text)) OR has_table_privilege(c.oid, 'DELETE'::text)) OR has_table_privilege(c.oid, 'REFERENCES'::text)) OR has_table_privilege(c.oid, 'TRIGGER'::text)));
ERROR: column t.tgisconstraint does not exist at character 1877
Is this expected behavior? It was not a show-stopper, and so far I haven't seen any bad side effects.
Thanks,
Dick Wieland
