sanitizedb fails due to triggers attempting to modify mutating tables

[Stuart Bishop]

sanitizedb.py turns all our foreign key constraints into ON DELETE CASCADE foreign key constraints to allow us to delete private data and all its dependencies.

Unfortunately, on some tables we have triggers. One limitation of triggers is that they should not modify a mutating table, which for us only becomes apparent with the cascading deletes.

psycopg2.IntegrityError: insert or update on table "bug" violates foreign key constraint "bug_duplicateof_fk"
DETAIL: Key (duplicateof)=(42) is not present in table "bug".
CONTEXT: SQL statement "UPDATE Bug SET date_last_message = max_datecreated FROM ( SELECT BugMessage.bug, max(Message.datecreated) AS max_datecreated FROM BugMessage, Message WHERE BugMessage.id <> $1 AND BugMessage.bug = $2 AND BugMessage.message = Message.id GROUP BY BugMessage.bug ) AS MessageSummary WHERE Bug.id = MessageSummary.bug"
PL/pgSQL function "set_bug_date_last_message" line 7 at SQL statement

[Stuart Bishop (stub)]

High priority since lots of people seem to want this for I don't know what reasons.

[Stuart Bishop (stub)]

I think we will need to special case these as the script trips over them on the production data - I don't think we can handle this programatically and there shouldn't be many cases to worry about.